Gröbner Bases, Multipolynomial Resultants

Grbner bases, multipolynomial resultants
and the Gauss-Jacobi combinatorial algorithms

-adjustment of nonlinear GPS/LPS observations-
Von der Fakultt fr

Bauingenieur- und Vermessungswesen der Universitt Stuttgart
zur Erlangung der Wrde eines Doktor-Ingenieurs (Dr.-Ing.)
genehmigte Abhandlung
vorgelegt von
M.Sc. Joseph L. Awange
aus Gem-Kathomo, Siaya / Kenya
Examiner - Hauptberichter:
Co-examiners - Mitberichter:
Tag der mndlichen Prfung:
Prof. Dr.-Ing. habil. Dr. tech. h.c. mult., Dr.-Ing. E.h. E. W. Grafarend
Prof. Dr.-Ing. habil. D. Lelgemann, Berlin
Prof. Dr. rer. nat. W. Freeden, Kaiserslautern
26-11-2001
Geodtisches Institut der Universitt Stuttgart

2002
Dissertation (D93)
Geodtisches Institut der Universitt Stuttgart
2002
An online colourfull version of this dissertation is availlable at homepages of the University Library and Department
of Geodesy and geoInformatics of the University of Stuttgart under the following addresses:
http://elib.uni-stuttgart.de/opus/
http://www.uni-stuttgart.de/gi/research
iii
Abstract
The algebraic techniques of Grbner bases and Multipolynomial resultants are presented as efficient algebraic tools
for solving explicitly the nonlinear geodetic problems. In particular, these algebraic tools are used to provide symbolic
solutions to the problems of GPS pseudo-ranging four-point P4P, Minimum Distance Mapping and the threedimensional resection. The various forward and backward substitution steps inherent in the classical closed form solutions
for these selected geodetic problems are avoided. Similar to the Gauss elimination technique in linear systems of
equations, the Grbner bases and Multipolynomial resultants eliminate several variables in a multivariate system of
nonlinear equations in such a manner that the end product normally consists of univariate polynomial equations whose
roots can be determined by existing programs such as the roots command in MATLAB.
The capability of Grbner bases and Multipolynomial resultants to solve explicitly nonlinear geodetic problems enables us to use them as the computational engine in the Gauss-Jacobi combinatorial algorithm proposed by C. F.
Gauss (published posthumously e.g Appendix A.4) and C. G. I. Jacobi (1841) to solve the nonlinear Gauss-Markov
model. With the nonlinear geodetic observation equations converted into algebraic (polynomial) form, the GaussJacobi combinatorial algorithm is carried out in two steps.
In the first step all combinations of minimal subsets of the observation equations are formed, and the unknowns of
each subset are rigorously solved by means of the Grbner bases and the Multipolynomial resultants. The solution
. In a second step the solution of the overdetermined
sets can be represented as points in an -dimensional space
Gau-Markov-model is constructed as weighted arithmetic mean of those solution points. Hereby the weights are
obtained from the Error propagation law/variance-covariance propagation. Using the developed Gauss-Jacobi combinatorial algorithm, the overdetermined three-dimensional resection problem based on the test network Stuttgart
Central is solved.
The algorithms are finally applied successfully to two case studies; transforming in a closed form geocentric coordinates to Gauss ellipsoidal coordinates (geodetic) and to obtain the seven datum transformation parameters from two
sets of coordinates. By means of Grbner basis, the scale parameter (in the seven datum transformation parameters
problem) is shown to fulfill be a univariate algebraic equation of fourth order (quartic polynomial) while the rotation
parameters are functions in scale and the coordinates differences.
Zusammenfassung
Die Methode der Grbner-Basen (GB) und der Multipolynomialen Resultante (MR) wird als wirksames algebraisches
Hilfsmittel zur expliziten Lsung nichtlinearer geodtischer Probleme vorgestellt. Insbesondere wird diese Methode dazu benutzt, um das Vierpunkt-pseudo-ranging-Problem beim GPS, das Problem des minimalen Abstands eines
vorgegebenen Punktes zu einer gekrmmten Flche sowie das Problem des dreidimensionalen Rckwrtsschnittes
analytisch zu lsen. Die verschiedenen Schritte der Vorwrts- und Rckwrts-Einsetzung, die bei den klassischen
geschlossenen Lsungen dieser ausgewhlten Probleme unumgnglich sind, werden dabei vermieden. In hnlicher
Weise wie bei der Gauschen Elimination bei linearen Gleichungssystemen werden durch die Methode der GB
und der MR die Variablen eines multivariaten nichtlinearen Gleichungssystems so eliminiert, dass eine univariate
Gleichung hherer Ordnung entsteht, deren Lsungsmenge mit existierenden Formelmanipulationsprogrammen wie
MATLAB bestimmt werden kann.
Wir nutzen die GB und MR als Rechenhilfsmittel bei der Lsung des nichtlinearen Gau-Markov-Modells mit Hilfe
des kombinatorischen Gau-Jacobi-Algorithmus, der von C.F. Gau (posthum verffentlicht, Anhang A.4) und von
C.G.I. Jacobi (1841) vorgeschlagen wurde. Sind die nichtlinearen geodtischen Beobachtungsgleichungen in algebraische (polynomiale) Form gebracht, wird der Gau-Jacobi-Algorithmus in zwei Schritten durchgefhrt. Im ersten
Schritt werden alle
Kombinationen von minimalen Untermengen der Beobachtungsgleichungen gebildet, daraus
werden mit Hilfe von GB- und MR-Algorithmen jeweils streng die Unbekannten des Problems bestimmt. Die
Lsungsmengen knnen als Punkte in einem -dimensionalen Raum
dargestellt werden. Im zweiten Schritt wird
die Lsung des berbestimmten Gau-Markov-Modells als gewichtetes arithmetisches Mittel dieser Lsungspunkte
konstruiert. Dabei ergeben sich die Gewichte aus dem Fortpflanzungsgesetz der Varianzen/Kovarianzen. Mit Hilfe
des so erweiterten kombinatorischen Gau-Jacobi-Algorithmus wird das Testnetz Stuttgart Stadtmitte als berbestimmter Rckwrtsschnitt ausgeglichen.

Die Algorithmen werden schlielich auf zwei Fallstudien erfolgreich angewandt: es werden zum einen geozentrische
Koordinaten in geschlossener Form in Gausche ellipsoidische (geodtische) Koordinaten transformiert, zum anderen
werden aus zwei entsprechenden Koordinaten-Datenstzen, die sich auf Systeme unterschiedlicher Datumsfestlegung
iv
beziehen, die sieben Transformationsparameter bestimmt. Mit Hilfe der GB wird gezeigt, dass bei dem letztgenannten Problem der Mastabsfaktor als eine univariates algebraische Gleichung vierter Ordnung erfllt, whrend die
Rotationsparameter Funktionen des Mastabsfaktors und der Koordinatendifferenzen sind.
Acknowledgments
First and foremost, let me begin by quoting from the book of Psalms 121:1-2 (NIV) I lift up my eyes to the hills,
where does my help come from? My help comes from the Lord, the Maker of heaven and earth. To you my God shall
I offer my praise, for it was your hand that led me through. The precious gift of life that you have granted unto us
manifests your true love, grace and mercy. We are truly a living proof of you grace and mercy. May your name be
lifted high both in heaven and on earth. Glory and honour be to you, my most holy God, for through Jesus Christ, we
have been made sons and part takers of the kingdom.
I wish to acknowledge gratefully the financial support from the Deutsche Akademische Austauschdienst DAAD which
enabled me to undertake my Ph. D. studies at the Department of Geodesy and Geoinformation, University of Stuttgart,
Germany. Your financial support was like fresh cool water for a thirsty fellow in the desert.
My special gratitude to the staff of the Department of Geography and Earth Sciences, Moi University Kenya, as well
as the administration of Moi University for approving my study leave to study in Germany. The support you gave to
my family during my absence is highly appreciated.
I wish to express my sincere gratitude to Prof. Dr.-Ing. habil. Erik W. Grafarend whose tireless efforts and patience
as my supervisor greatly contributed towards the completion and success of this study. The advice, suggestions,
discussions and encouragement he gave inspired me greatly. Of particular importance, he fitted me into his bed
of tight schedule, but unlike the bed of Procrustes (the Greek legend who fitted his visitors to their beds by either
stretching or cutting their legs, in which case they died), this particular bed provided vast knowledge. For that, I am
grateful.
I am indebted to Prof. B. Buchberger and Prof. D. A. Cox who readily discussed the topic of Grbner basis, Prof. B.
Sturmfels and Prof. D. Manocha for their discussions on the topic of Multipolynomial resultants, Prof. W. Keller who
was always willing to discuss the topic and Professors F. W. O. Aduol of Dept. of Surveying, University of Nairobi
and P. Omondi of Dept. Geography and Earth Sciences, Moi University for their encouragement: The suggestions you
gave, materials you sent and the encouragement from you are highly appreciated.
Special thanks to Dr.-Ing. F. Krumm whose door was always opened for consultations, advice and discussions. I
recognize the friendship and sacrifice you made especially when things were tough. I salute Dr.-Ing. J. Engels and
Dr.-Ing. V. Schwarze for their contributions towards the success of the study. I would have wished to mention each
of the members of the Department of Geodesy and Geoinformation, University of Stuttgart by name, for each of them
contributed in one way or another towards the success of my research work: Your encouragement, tolerance and love
provided a conducive working environment. May the almighty God greatly reward each of you.
To my friends A. Kahi, J. B. K. Kiema, J. B. Miima, M. Ngigi, C. Gaya, J. Omollo, E. Mugoye, V. Mabu, N. M. Otiato,
and A. Osire, I want you to know that true friends are like rare jewels and I am grateful for your friendship.
Last, but not least, I wish to acknowledge the patience of my wife Naomi Awange and two daughters Mellissa Lucy
Awange and Ruth Mich Awange. Your cheerful faces and smiles brightened the entire study.
For those who contributed in one way or another and have not been named, you have not been forgotten. I salute you.
To you all, I thank you from the bottom of my heart. May the blessings of the almighty God be with you richly.
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iv
Introduction
1-1 Statement of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1-2 Solution of the problem
Nonlinear Adjustment
2-1 Linear and nonlinear Gauss-Markov models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2-2 The Gauss-Jacobi combinatorial algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2-3 Solution of nonlinear Gauss-Markov model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2-31
Construction of the minimal combinatorial subsets . . . . . . . . . . . . . . . . . . . . . . .
13
2-32
Closed form solution of the minimal combinatorial subsets . . . . . . . . . . . . . . . . . . .
13
2-321
Closed form solution of nonlinear equations by Grbner bases
. . . . . . . . . . .
13
2-322
Closed form solution of nonlinear equations by Multipolynomial Resultants . . . .
19
Adjustment of the combinatorial minimal subsets solutions . . . . . . . . . . . . . . . . . . .
23
2-4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
Solution of selected Geodetic problems
27
3-1 LPS and GPS positioning systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
2-33
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3-11
GPS Positioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
3-12
LPS positioning
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
3-121
Local Positioning Systems (LPS) . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
3-122
Local datum choice in an LPS 3-D network . . . . . . . . . . . . . . . . . . . . .
28
3-13
Relationship between Global and Local Level Reference Frames
. . . . . . . . . . . . . . .
29
3-14
Observation Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
3-2 Selected geodetic problems
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3-21
Threedimensional resection problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3-22
The minimum distance mapping problem . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
3-23
GPS positioning with observations of type pseudo-ranges
. . . . . . . . . . . . . . . . . . .
51
3-231
The pseudo-ranging four-point problem in GPS positioning . . . . . . . . . . . . .
51
3-232
Overdetermined GPS positioning with observations of type pseudo-ranges . . . . .
57
3-233
Linearized least squares versus Gauss-Jacobi combinatorial algorithm . . . . . . .
59
3-234
Outlier diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
CONTENTS
vi
4 Test network Stuttgart Central
63
4-1 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
4-2 Closed form solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
4-21
Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
4-3 Overdetermined solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
4-31
Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
4-4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
4-41
Distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
4-42
Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
5 Case studies
83
5-1 Geocentric Cartesian to Gauss ellipsoidal coordinates . . . . . . . . . . . . . . . . . . . . . . . . . .
83
5-2 7-parameter datum transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
6 Summary and Conclusions
95
Appendices
97
Appendix A-1:
Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix A-2:
Simplexes for Stuttgart Central . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Appendix A-3:
Error propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Appendix A-4:
C. F. Gauss combinatorial approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
References
97
113
List of Figures
2.1
Levelling Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
3.1
Tetrahedron for closed form 3d-resection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.2
Computed PDOP for respective combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
58
3.3
Scatter of the 15 Gauss-Jacobi combinatorial solutions ( ) around the adjusted value ( ). . . . . . . .
59
3.4
Magnification of the scatter of 14 Gauss-Jacobi combinatorial solutions ( ) around the adjusted value
( ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
3.5
Deviation of positional norm from the mean value . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
4.1
Graph of the Test network Stuttgart Central . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
4.2
Deviation of computed position of station K1 in Tables (4.18) and (4.20) from the real value in Table
(4.1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
4.3
Scatter of combinatorial solutions for data sets 1 and 2 in Tables (4.3) and (4.4) . . . . . . . . . . . .
77
4.4
78
4.5
79
4.6
80
4.7
Scatter of combinatorial solutions for data sets 9 and 10 in Tables (4.11) and (4.12) . . . . . . . . . .
81
4.8
Scatter of combinatorial solutions for data set 11 Table (4.13) . . . . . . . . . . . . . . . . . . . . . .
82
5.1
Minimum distance mapping of a point on the Earths topographic surface to a point on the International Reference Ellipsoid
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
5.2
Scatter of the computed 36 minimal combinatorial values of scale around the adjusted value . . . . .
93
5.3
Scatter of the 36 computed translations and rotations around the adjusted values . . . . . . . . . . . .
94

A1
Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
A2
Combinatorial simplexes for the for test network Stuttgart Central . . . . . . . . . . . . . . . . . 104
A3
A4
A5
A6
vii
viii
LIST OF FIGURES
List of Tables
3.1
Cartesian coordinates of topographic points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
3.2
Computed polynomial coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
3.3
Computed ellipsoidal cartesian coordinates and the Lagrange factor . . . . . . . . . . . . . . . . . .
52
3.4
Geocentric coordinates of four GPS satellites and the pseudo-ranging observations . . . . . . . . . .
56
3.5
Geocentric coordinates of six GPS satellites and the pseudo-range observations . . . . . . . . . . . .
57
3.6
Possible combinations and the computed PDOP . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
3.7
Computed coefficients of the combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
58
3.8
Computed combinatorial solution points in a polyhedron . . . . . . . . . . . . . . . . . . . . . . . .
58
3.9
Computed receiver position and stationary receiver range bias . . . . . . . . . . . . . . . . . . . . .
60
3.10 Computed root-mean-square errors and the difference in solution between the two approaches . . . .
60
3.11 Computed residuals, squares of residuals and error norm . . . . . . . . . . . . . . . . . . . . . . . .
60
3.12 Computed positional norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
, "!#$%!&!'()+*-,./1020303(4
4.1
GPS Coordinates in the Global Reference Frame
. . . . .
64
4.2
Ideal spherical coordinates of the relative position vector in the Local Horizontal Reference Frame :
Spatial distances, horizontal directions, vertical directions . . . . . . . . . . . . . . . . . . . . . . .
65
Randomly generated spherical coordinates of the relative position vector: horizontal directions and
vertical directions
root-mean-square errors of individual observations, differences
with respect to
ideal data
of Table 4.2, first data set: set 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.4
Second data set: set 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.5
Third data set: set 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.6
Fourth data set: set 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.7
Fifth data set: set 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
4.8
Sixth data set: set 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
4.9
seventh data set: set 7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
4.10 Eigth data set: set 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
4.11 Ninth data set: set 9
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
4.12 Tenth data set: set 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
4.13 Eleventh data set: set 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
4.14 Distances computed by Grbner basis or Multipolynomial resultants
. . . . . . . . . . . . . . . . .
74
4.15 Distances computed by Gauss-Jacobi combinatorial algorithm . . . . . . . . . . . . . . . . . . . . .
74
4.16 Root-mean-square-errors of distances in Table (4.15) . . . . . . . . . . . . . . . . . . . . . . . . . .
74
4.3
7 ! #)*8,$.91030203:;4/
<=61!?>@*A61!CBD61!$FEG2H9G3IJLK$G2M9NO<P7 !?>@*Q7R!CBS7R!#EG2H9G3IJLK$G2M9
ix
5
6!
61!T7R!U
LIST OF TABLES
4.17 Deviations of distances in Table (4.15) from values in Table (4.2) . . . . . . . . . . . . . . . . . . . .
75
4.18 K1 computed by Grbner basis or Multipolynomial resultants . . . . . . . . . . . . . . . . . . . . .
75
4.19 Position of K1 computed by Gauss-Jacobi combinatorial algorithm . . . . . . . . . . . . . . . . . .
75
4.20 Deviation of K1 in Table (4.19) from the real value in Table (4.1)
. . . . . . . . . . . . . . . . . . .
75
5.1
Cartesian coordinates of topographic point (Baltic Sea Level Project) . . . . . . . . . . . . . . . . . .
85
5.2
Polynomial coefficients
86
5.3
5.4
V(WXV2YLXV
ZVN[OZV(\ of the univariate polynomial of order four in ];\ . . . . . .
Computed Cartesian coordinates F]^Y_#] ]%[&*`]C#a^$bc and Lagrange multiplier ]%\ . . . . . . . . .
86
Geodetic Coordinates computed from ellipsoidal Cartesian coordinates in closed form (Baltic Sea
Level Project) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
5.5
Coordinates for system A (Local system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
5.6
Coordinates for system B (WGS 84) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
5.7
Computed 7-parameter datum transformation by Grbner basis . . . . . . . . . . . . . . . . . . . . .
91
5.8
Transformed Cartesian coordinates of System A (Table 5.5) into System B (Table 5.6) using the 7datum transformation parameters of Table (5.7) computed by Grbner basis . . . . . . . . . . . . . .
91
Computed 7-parameter datum transformation using Gauss-Jacobi combinatorial algorithm . . . . . .
92
5.9
5.10 Transformed Cartesian coordinates of System A (Table 5.5) into System B (Table 5.6) using the 7datum transformation parameters of Table (5.9) computed by the Gauss-Jacobi combinatorial algorithm 92
5.11 Transformed Cartesian coordinates of System A (Table 5.5) into System B (Table 5.6) using the 7datum transformation parameters computed by Least Squares Solution . . . . . . . . . . . . . . . . .
93
5.12 Computed residual norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
93
Chapter 1
Introduction
1-1
Statement of the problem
In Geodesy, Photogrammetry and Computer Vision, nonlinear equations are often encountered in several applications
as they often relate the observations (measurements) to the unknown parameters to be determined. In case the number
), the unknown parameters may be obtained by
of observations and the number of unknowns are equal (
solving explicitly or in a closed form the system of equations relating observations to the unknown parameters. For
example, D. Cox et al. (1998, pp.28-32) has illustrated that for systems of equations with exact solution, the system
become vulnerable to small errors introduced during root findings and in case of extending the partial solution to the
complete solution of the system, the errors may accumulate and thus become so large. If the partial solution was
derived by iterative procedures, then the errors incurred during the root-finding may blow up during the extension of
the partial solution to the complete solution (back substitution).
d*
In some applications, symbolic rather than numerical solution are desired. In such cases, explicit procedures are
usually employed. The resulting symbolic expressions often consists of univariate polynomials relating the unknown
parameters (unknown variables) to the known variables (observations). By inserting known values into these univariate
polynomials, numerical solutions are readily computed for the unknown variables. Advantages of explicit solutions
have been listed by E. L. Merritt (1949) as; provision of satisfaction to the users (Photogrammetrist and Mathematicians) of the methods, provision of data tools for checking the iterative methods, desired by Geodesist whose task of
control densification does not favor iterative procedures, provision of solace and the requirement of explicit solutions
rather than iterative by some applications. In Geodesy for example, the Minimum Distance Mapping problem considered by E. Grafarend and P. Lohse (1991) relates a point on the Earths topographical surface uniquely (one-to-one)
to a point on the International Reference Ellipsoid. The solution of such an optimization problem requires that the
equations be solved explicitly.
The draw back that was experienced with explicit solutions was that they were like rare jewel. The reason for this
was partly because the methods required extensive computations for the results to be obtained and partly because the
resulting symbolic expressions were too large and required computers with large storage capacity. Until recently, the
computers that were available could hardly handle large computations due to lack of faster Central Processing Unit
(CPU), shortage of Random Access Memory (RAM) and limited hard disk space for storage. The other set back
experienced by the explicit procedures was that some of the methods, especially those from algebraic fields, were
formulated based on theoretical concepts that were hard to realize or comprehend without the help of computers. For a
long time, these setbacks hampered progress of the explicit procedures. The advances made in computer technology in
recent years however has changed the tides and led to improvements in explicit computational procedures which hitherto were difficult to achieve. Apart from the improvements in existing computational procedures, new computational
techniques are continously being added to the increasing list of computational methods with an aim of optimizing
computational speed and efficiency. In this category are the algebraic methods of Grbner bases and Multipolynomial
resultants.
The present study examines the suitability of algebraic computational techniques of Grbner bases and Multipolynomial resultants in solving explicitly nonlinear systems of observation equations that have been converted into algebraic
(polynomial) form in Geodesy. The algebraic techniques of Grbner bases and Sylvester resultant (resultant of two
polynomials) for solving polynomial equations in Geodesy have been mentioned and examples of their applications
to the two-dimensional case given in the work of P. Lohse (1994, pp.36-39, 71-76). The present study considers
1
CHAPTER 1. INTRODUCTION
the Grbner bases and Multipolynomial resultants (resultant of more than two polynomials) in the solution of threedimensional problems. E. Grafarend (1989) has already suggested the use of Grbner bases approach in solving the
perspective center problem in photogrammetry.
Other than revolutionizing computation procedures, the advances made in computer technology have also led to improvement in instrumentation for data acquisition as exemplified in the case of GPS positioning satellites. Since its
inception as a positioning tool, the Global Positioning System (GPS) -referred to by E. Grafarend and J. Shan (1996)
as the Global Problem Solver- has revolutionized geodetic positioning techniques and maintained its supremacy as a
positioning tool. These improvements on instrumentation for data acquisition have led to improved data collection
procedures and increase in accuracy. Lengthy geodetic procedures such as triangulation that required a lot of time,
intervisibility between stations and a large manpower are rapidly being replaced by satellite positioning techniques
which require shorter observation periods, no intervisibility requirement, weather independent and less manpower
leading to optimization of time and money. Whereas improved instrumentation is applauded, it comes a long with its
own difficulties. One of the difficulties is that a lot of data is collected than required to determine unknown parameters
leading to redundancies. In positioning with GPS for example, due to its constellation that offer a wider coverage,
more than four satellites can be observed at any point of the earth. In the minimal case, only four satellites are required
to fix the receiver position and the receiver clock bias assuming that the transmitter clock bias and the transmitterreceiver bias have been neglected for the pseudo-range type of observations. More than four satellites therefore lead
, the explicit solutions give way to optimization procedures
to superfluous observations. In such cases, where
such as least squares solution which work very well for linear models under specified conditions.
de
In Geodesy however, the observation equations are normally nonlinear thus requiring the use of nonlinear GaussMarkov model which is normally solved either by first linearizing the observation equations using Taylor series expansion to the second order terms about approximate values of the unknowns then applying linear models estimation
procedures or by using iterative procedures such as the Gauss-Newton approach. The linearization approach has the
disadvantage that the linearized approximation of the nonlinear models may still suffer from nonlinearity and thus
resulting in the estimates of such models being far from the real estimates of the nonlinear models. This can easily be
checked by re-substituting the estimates from the linearized model into the original nonlinear model.
For the iterative procedures, the greatest undoing may be the requirement of the initial approximate values to start off
the iterations which may not be available for some applications. For simper models, the approximate initial values may
be computed, for others however, the approximate values may be impossible to compute. Apart from the problem of
getting the initial approximate values, there also exists the problem that poor choice of approximate values may lead to
lack of convergence or if the approximate value be far from the real solution, then a large number of iterations may be
required to get close to the real solution thus rendering the whole procedure to be quite slow, especially where multiple
roots are available. For other procedures such as the 7-parameter datum transformation that requires linearization and
iterative methods, it is not feasible to take into account the stochasticity of both systems involved. Clearly, a procedure
for solving nonlinear Gauss-Markov model that can avoid the requirement of initial approximate starting values for
iteration and linearization approaches and also take into consideration the stochasticity of the systems involved is the
desire of modern day Geodesist and Photogrammetrist.
With this background, the present study aims at answering the following questions:
For geodetic problems requiring explicit solutions, can the algebraic tools of Grbner bases and Multipolynomial resultants that have found applications in other fields such as Robotics (for kinematic modelling of robots),
Visions, Computer Aided Design (CAD), Engineering (offset surface construction in solid modelling), Computer Science (automated theorem proving) e.t.c. be used to solve systems of nonlinear observation equations
of algebraic (polynomial) type?
Is there any alternative for solving the nonlinear Gauss-Markov model without resorting to linearization or
iterative procedures that require approximate starting values?
To answer the first question, the present study uses the Grbner bases and Multipolynomial resultants to solve explicitly the problems of GPS four-point pseudo-ranging, Minimum Distance Mapping and the threedimensional resection.
The answer to the second problem becomes clear once the first question has been answered. Should the algebraic
techniques of Grbner bases and Multipolynomial resultants be successful in solving explicitly the selected geodetic
problems, then they are used as the computational engine of the combinatorial algorithm that was first suggested by
C. F. Gauss (Published posthumously e.g. in Appendix A.4) and later on by C. G. I. Jacobi (1841) and extended by P.
Werkmeister (1920). We refer to this algorithm as the Gauss-Jacobi combinatorial algorithm. In attempting to answer
the questions above, the objectives of the present study are formulate as:
1-2. SOLUTION OF THE PROBLEM
(1)
Analyze the algebraic computational procedures of type Grbner bases and Multipolynomial resultants
with the aim of establishing their suitability in solving explicitly (in closed form) geodetic nonlinear
problems. In this respect, the Grbner bases and Multipolynomial resultants techniques are used to solve
explicitly (symbolically) geodetic problems of GPS pseudo-ranging four-point P4P, Minimum Distance
Mapping and the threedimensional resection. By converting the nonlinear observation equations of these
selected geodetic problems into algebraic (polynomial), the study aims at using the Grbner bases and
Multipolynomial resultants techniques to eliminate several variables in a multivariate system of nonlinear
polynomial equations in such a manner that the end product from the initial system of nonlinear observation equations normally consist of a univariate polynomial. The elimination procedure is similar to the
Gauss elimination approach in linear systems.
(2)
From the principle of weighted arithmetic mean and using the Gauss-Jacobi combinatorial lemma (C.
G. I. Jacobi 1841), develop an adjustment procedure that neither linearizes the nonlinear observation
equations nor uses iterative procedures to solve the nonlinear Gauss-Markov model. Linearization is
permitted only for nonlinear error propagation/variance-covariance propagation. Such procedure is to
use the univariate polynomial generated by the algebraic computational procedures of type Grbner bases
or Multipolynomial resultants as the computing engine for its minimal combinatorial set.
(3)
Test the procedures in solving real geodetic problems of determining the 7 transformation parameters and
transforming in a closed form the geocentric Cartesian coordinates to Gauss ellipsoidal coordinates
(geodetic).
1-2
Solution of the problem
The current known techniques for solving nonlinear polynomial equations can be classified into symbolic, numeric
and geometric methods (D. Manocha 1994c). Symbolic methods, which we consider in this study for solving closed
form geodetic problems, apply the Grbner bases and the Multipolynomial resultants techniques to eliminate several
variables in a multivariate system of equations in such a manner that the end product often consist of univariate
polynomials whose roots can be determined by existing programs such as the roots command in MATLAB. The
current available programs however are efficient only for sets of low degree polynomial systems consisting of upto
three to four polynomials due to the fact that computing the roots of the univariate polynomials can be ill conditioned
for polynomials of degree greater than 14 or 15 (D. Manocha 1994c).
Elaborate literature on Grbner bases can be found in the works of B. Buchberger (1965, 1970), J. H. Davenport et
al. (1988, pp.95-103), F. Winkler (1996), D. Cox et al. (1997, pp.47-99), H. M. Mller (1998), W. V. Vasconcelos
(1998), T. Becker and V. Weispfenning (1993,1998), B. Sturmfels (1996), G. Pistone and H. P. Wynn (1996), D. A.
Cox (1998) and D. Cox et al.(1998, pp.1-50), while literature on Multipolynomial resultants procedure include the
works of G. Salmon (1876), F. Macaulay (1902, 1921), A. L. Dixon (1908), B. L. van Waerden (1950), C. Bajaj et al.
(1988), J. F. Canny (1988), J. F. Canny et al. (1989), I. M. Gelfand et al. (1990), J. Weiss (1993), D. Manocha (1992,
1993, 1994a,b,c, 1998), D. Manocha and J. F. Canny (1991, 1992, 1993), I. M. Gelfand et al. (1994), G. Lyubeznik
(1995), S. Krishna and D. Manocha (1995), J. Guckenheimer et al.(1997), B. Sturmfels (1994, 1998), E. Cattani et al.
(1998) and D. Cox et al. (1998, pp.71-122). Besides the Grbner bases and resultant techniques, there exists another
approach for variable elimination developed by WU Wen Tsn (W. T. Wu 1984) using the ideas proposed by J. F.
Ritt (1950). This approach is based on Ritts characteristic set construction and was successfully applied to automated
geometric theorem by Wu. This algorithm is referred by X. S. Gao and S. C. Chou (1990) as the Ritt-Wus algorithm
(D. Manocha and F. Canny 1993). C. L. Cheng and J. W. Van Ness (1999) have presented polynomial measurement
error models.
Numeric methods for solving polynomials can be grouped into iterative and homotopy methods. For homotopy we
refer to A. P. Morgan (1992). Also in this category are geometric methods which have found application in curve and
surface intersection whose convergence are however said to be slow (D. Manocha 1994c). In general, for low degree
curve intersection, the algebraic methods have been found to be the fastest in practice. In Sections (2-321) and (2-322)
of Chapter 2, we present in a nut shell the theories of Grbner bases and Multipolynomial resultants.
The problem of nonlinear adjustment in Geodesy as in other fields continues to attract more attention from the modern
day researchers as evidenced in the works of R. Mautz (2001) and L. Guolin (2000) who presents a procedure that tests
using the F-distribution whether a nonlinear model can be linearized or not. The solution to the minimization problem
of the nonlinear Gauss-Markov model unlike its linear counter part does not have a direct method for solving it and
as such, always relies on the iterative procedures such as the Steepest-descent method, Newtons method and the
Gauss-Newtons method discussed by P. Teunissen (1990). In particular, P. Teunissen (1990) recommends the GaussNewtons method as it exploits the structure of the objective function (sum of squares) that is to be minimized. P.
Teunissen and E. H. Knickmeyer (1988) considers in a statistical way how the nonlinearity of a function manifests
itself during the various stages of adjustment. E. Grafarend and B. Schaffrin (1989, 1991) while extending the work
of T. Krarup (1982) on nonlinear adjustment with respect to geometric interpretation have presented the necessary
and sufficient conditions for least squares adjustment of nonlinear Gauss-Markov model and provided the geometrical
interpretation of these conditions.
Other geometrical approaches include the works of G. Blaha and R. P. Besette (1989) and K. Borre and S. Lauritzen
(1983) while non geometrically treatment of nonlinear problems have been presented by K. K. Kubik (1967), T. Saito
(1973), H. J. Schek and P. Maier (1976), A. Pope (1982) and H. G. Bhr (1988). A comprehensive review to the
iterative procedures for solving the nonlinear equations is presented in the work of P. Lohse (1994). M. Gullikson and
I. Sderkvist (1995) have developed algorithms for fitting surfaces which have been explicitly or implicitly defined to
some measured points with negative weights being acceptable by the algorithm.
Our approach in the present study goes back to the work of C. F. Gauss (Published posthumously e.g. Appendix A.4)
and C. G. I. Jacobi (1841). Within the framework of arithmetic mean, C. F. Gauss (Published posthumously e.g.
Appendix A.4) and C. G. I. Jacobi (1841) suggest that given linear(ized) observation equations with unknowns
combinations, each consisting of
equations be solved for the unknown elements and the weighted
arithmetic mean be applied to get the final solution. Whereas C. F. Gauss (Published posthumously e.g. Appendix A.4)
proposes weighting by using the products of the square of the measured distances from the unknown point to known
points and the distances of the side of the error triangles, C. G. I. Jacobi (1841) proposed the use of the square of the
determinants as weights. In tracing the method of least squares to the arithmetic mean, A. T. Hornoch (1950) shows
that the weighted arithmetic mean proposed by C. G. I. Jacobi (1841) leads to the least squares solution only if the
weights used are the actual weights and the pseudo-observations formed by the combinatorial pairs are uncorrelated.
Using a numerical example, S. Wellisch (1910, pp. 41-49) has shown the results of least squares solution to be identical
to those of the Gauss-Jacobi combinatorial algorithm once proper weighting is applied.
Ffe
(;g
P. Werkmeister (1920) illustrated that for planar resection case with three directional observations from the unknown
point to three known stations, the area of the triangle (error figure) formed by the resulting three combinatorial coordinates of the new point is proportional to the determinant of the dispersion matrix of the coordinates of the new station.
In the present study, these concepts of the combinatorial linear adjustment are extended to nonlinear adjustment. In
chapter 2, the linear and nonlinear Gauss-Markov models are introduced in Section (2-1). Section (2-2) presents
and proves the Gauss-Jacobi combinatorial lemma which is required for the solution of the nonlinear Gauss-Markov
model. We illustrate using a leveling network and planar ranging problem that the results of Gauss-Jacobi combinatorial algorithm are identical to those of linear Gauss-Markov model if the actual variance-covariance matrices are
used.
To test the algebraic computational tools of Grbner bases and Multipolynomial resultants presented in Chapter 2,
geodetic problems of threedimensional resection, Minimum Distance Mapping and GPS pseudo-ranging four-point
P4P have been used in Chapter 3. In general, the search towards the solution of the three-dimensional resection
problem traces its origin to the work of a German mathematician J. A. Grunert (1841) whose publication appeared
in the year 1841. J. A. Grunert (1841) solved the threedimensional resection problem -what was then known as the
Pothenots problem- in a closed form by solving an algebraic equation of degree four. The problem had hitherto been
solved by iterative means mainly in Photogrammetry and Computer Vision. Procedures developed later for solving
the three-dimensional resection problem revolved around the improvements of the approach of J. A. Grunert (1841)
with the aim of searching for the optimal means of distances determination. Whereas J. A. Grunert (1841) solves
the problem by substitution approach in three steps, the more recent desire has been to solve the distance equations
in lesser steps as exemplified in the works of S. Finsterwalder and W. Scheufele (1937), E. L. Merritt (1949), M. A.
Fischler and R. C. Bolles (1981), S. Linnainmaa et al. (1988) and E. Grafarend, P. Lohse and B. Schaffrin (1989).
Other research done on the subject of resection include the works of F. J. Mller (1925), E. Grafarend and J. Kunz
(1965), R. Horaud et al. (1989), P. Lohse (1990), and E. Grafarend and J. Shan (1997a, 1997b). An extensive review
of some of the procedures above are presented by F. J. Mller (1925) and R. M. Haralick et al. (1991, 1994).
R. M. Haralick et al. (1994) reviewed the performance of six direct solutions (J. A. Grunert 1841, S. Finsterwalder
and W. Scheufele 1937, E. L. Merritt 1949, M. A. Fischler and R. C. Bolles 1981, Linnainmaa et al. 1988, and E.
Grafarend, P. Lohse and B. Schaffrin 1989) with the aim of evaluating the numerical stability of the solutions and the
effect of permutation within the various solutions. All the six approaches follow the outline adopted by J. A. Grunert
(1841) with the major differences being the change in variables, the reduction of these variables, and the combination
of different pairs of equations. The study revealed that the higher the order of the polynomials, the more complex
the computations became and thus the less accurate the solutions were due numerical instabilities. Consequently, S.
1-2. SOLUTION OF THE PROBLEM
Finsterwalders (SF) procedure which solves a third order polynomial is ranked first, J. A Grunert (JG), Fischler and
Bolles (FB), and Grafarend et al. (GLS) solutions are ranked second, Linnainmaa et al. solution which generates
an eighth order polynomial is ranked third. Though it does not solve the eight order polynomial, the complexity of
the polynomial is still found to be higher than those of the other procedures. An amazing result is that of Merritts
procedure which is ranked last despite the fact that it is a fourth order polynomial and is similar to Grunerts approach
except for the pairs of equations used. R. M. Haralick et al. (1994) attributes the poor performance of Merritts
procedure to the conversion procedure adopted by E. L Merritt (1949) in reducing the equations from forth to third
order. For planar resection problem, solutions have been proposed e.g. by D. Werner (1913), G. Brandsttter (1974)
and J. van Mierlo (1988).
In Sections (3-21), we present a solution of the three-dimensional resection problem by solving the Grunert distances
equations using Grbner bases and Multipolynomial resultants techniques. The resulting fourth order univariate
polynomial is solved for the unknown distance and the admissible solution substituted in other elements of Grbner
basis to determine the remaining two distances. Once we have obtained the spatial distances, the position is computed using any of the four approaches of the three-dimensional ranging problem (Bogenschnitt); Grbner basis,
Multipolynomial resultants, elimination by substitution and elimination using matrix approach. For the orientation
step which concludes the solution of the three-dimensional resection problem, we refer to the works of J. L. Awange
(1999) and E. Grafarend and J. L. Awange (2000) who solved the threedimensional orientation problem by using the
simple Procrustes algorithm. Here the simple Procrustes problem or partial Procrustes problem, a special case of the
general Procrustes problem, is understood to mean the fit of the rotation matrix which transforms a set of Cartesian
coordinates into another set of Cartesian coordinates. The general Procrustes problem includes besides the rotational
elements also translation, dilatation and reflection. A list of references is I. Borg and P. Groenen (1997), F. B. Brokken
(1983), T. F. Cox and M. A. Cox (1994), F. Crosilla (1983a,1983b), B. Green (1952), M. Gullikson (1995a, 1995b), S.
Kurz (1996), R. Mathar (1997), P. H. Schnemann (1996), L. N. Trefethen and D. Bau (1997), and I. L. Dryden (1998).
In order to relate a point on the Earths topographical surface uniquely (one-to-one) to a point on the International
Reference Ellipsoid, E. Grafarend and P. Lohse (1991) have proposed the use of the Minimum Distance Mapping.
Other procedures that have been proposed are either iterative, approximate closed, closed or higher order equations.
Iterative procedures include the works of N. Bartelme and P. Meissl. (1975), W. Benning (1987), K. M. Borkowski
(1987, 1989), N. Croceto (1993), A. Fitzgibbon et al. (1999), T. Fukushima (1999), W. Gander et al. (1994), B. Heck
(1987), W. Heiskannen and H. Moritz (1976), R. Hirvonen and H. Moritz (1963), P. Loskowski (1991), K. C. Lin and
J. Wang (1995), M. K. Paul (1973), L. E. Sjoeberg (1999), T. Soler and L. D. Hothem (1989), W. Torge (1976), T.
Vincenty (1978) and R. J. You (2000).
Approximate closed procedures include B. R. Bowring (1976, 1985), A. Fotiou (1998), B. Hofman-Wellenhof et al.
(1992), M. Pick (1985) and T. Vincenty (1980), while closed procedures include W. Benning (1987), H. Frhlich and
H. H. Hansen (1976), E. Grafarend et al. (1995) and M. Heikkinen (1982). Procedures that required the solution
of higher order equations include M. Lapaine (1990), M. J. Ozone (1985), P. Penev (1978), H. Snkel (1976), and P.
Vaniceck and E. Krakiwski (1982). In Section (3-22), the Minimum Distance Mapping problem is solved using the
Grbner bases approach. A univariate polynomial of fourth order is obtained together with 13 other elements of the
Grbner basis. The obtained univariate polynomial and the linear terms are compared to those of E. Grafarend and P.
Lohse (1991). Other reference include E. Grafarend and W. Keller (1995) and Mukherjee, K. (1996)
The GPS four-point pseudo-ranging problem is concerned with the determination of the coordinates of a stationary
receiver together with its range bias. Several closed form procedures have been put forward for obtaining closed form
solution of the problem. Amongst the procedures include the vectorial approach as evidenced in the works of L. O.
Krause (1987), J. S. Abel and J. W. Chaffee (1991), P. Singer et al. (1993), J. W. Chaffee and J. S. Abel (1994), H.
Lichtenegger (1995) and A. Kleusberg (1994, 1999). E. Grafarend and J. Shan (1996) propose two approaches; one
approach is based on a closed form solution of the nonlinear pseudo-ranging equations in geocentric coordinates while
the other approach solves the same equations in barycentric coordinates.
S. C. Han et al. (2001) have developed an algorithm for very accurate absolute positioning through Global Positioning
System (GPS) satellite clock estimation while S. Bancroft (1985) provides an algebraic closed form solution of the
overdetermined GPS pseudo-ranging observations. In Section (3-231) we solve using Grbner basis and Multipolynomial resultants GPS four-point pseudo-ranging problem. Chapter 3 ends by using the Gauss-Jacobi combinatorial
algorithm in Section (3-232) to solve the overdetermined GPS point pseudo-ranging problem. The LPS and GPS
Reference Frames relevant for these selected geodetic problems have been considered in Section (3-1). For literature
on three-dimensional positioning models, we refer to E. W. Grafarend (1975), V. Ashkenazi and S, Grist (1982), G. W.
Hein (1982a, 1982b), J. Zaiser (1986), F. W. O. Aduol (1989), U. Klein (1997) and S. M. Musyoka 1999).
Chapter 4 extends Chapter 3 by solving the three-dimensional resection problem based on the test network Stuttgart
Central. In Section (4-2), using the computed symbolic solutions obtained in Section (3-21), the coordinates of
Dach K1 are determined explicitly. Three stations Liederhalle, Eduardpfeiffer and Haussmanstr. are used. In Section
(4-3) all the 7 network points are used to determine the coordinates of Dach K1, thus leading to the solution of
overdetermined three-dimensional resection problem.
Overdetermined planar resection problem has been treated graphically by E. Hammer (1896), C. Runge (1900), P.
Werkmeister (1916) and P. Werkmeister (1920). E. Gotthardt (1940,1974) dealt with the overdetermined two dimensional resection where more than four points were considered with the aim of studying the critical configuration that
would yield a solution. The work was later to be extended by K. Killian (1990). A special case of an overdetermined
two-dimensional resection has also been considered by H. G. Bhr (1991) who uses six known stations and proposes
the measuring of three horizontal angles which are related to the two unknown coordinates by nonlinear equations.
By adopting approximate coordinates of the unknown point, an iterative adjustment procedure is performed to get the
improved two-dimensional coordinates of the unknown point. It should be noted that the procedure is based on the
coordinate system of the six known stations. K. Rinner (1962) has also contributed to the problem of overdetermined
two-dimensional resection.
Once we have developed and tested the algorithms in Chapters 2 to 4, Chapter 5 considers two case studies; the
conversion of the geocentric GPS points for the Baltic Sea level Project into the Gauss-Jacobi ellipsoidal coordinates
and the determination of the seven datum transformation parameters. We have used the skew symmetric matrix to
construct the orthogonal matrix
in Chapter 5. Other approaches have been presented in G. H. Shut (1958/59) and
E. H. Thompson (1959 a, b). Datum transformation models have been dealt with e.g. in E. Grafarend, F. Okeke (1998),
F. Krumm and F. Okeke (1995), E. Grafarend and F. Okeke (1998) and E. Grafarend and R. Syffus (1998).
hi[
Chapter 6 summarizes and concludes the study.
Chapter 2
Nonlinear Adjustment
In the present Chapter, we depart from the traditional iterative procedures for estimating the unknown fixed parameters
of the nonlinear Gauss-Markov model and present a combinatorial approach that traces its roots back to the work of
C. F. Gauss that was published posthumously (Appendix A.4). C. F. Gauss first proposed the combinatorial approach
using the products of squared distances (from unknown point to known points) and the square of the perpendicular
distances from the sides of the error triangle to the unknown point as the weights. According to W. K. Nicholson (1999,
pp. 272-273), the motto in Gauss seal read pauca des matura meaning few but ripe. This belief led C. F. Gauss not
to publish most of his important contributions. For instance, W. K. Nicholson (1999, pp. 272-273) writes Although
not all his results were recorded in the diary (many were set down only in letters to friends), several entries would have
each given fame to their author if published. Gauss new about the quartenions before Hamilton.... The combinatorial
method, like many of his works, was later to be published after his death. Several years later, the method was to be
developed further by C. G. I. Jacobi (1841) who used the square of the determinants as the weights in estimating the
unknown parameters from the arithmetic mean. P. Werkmeister (1920) later established the relationship between the
area of the error figure formed by the combinatorial approach and the standard error of the determined point. We will
refer to this combinatorial approach as the Gauss-Jacobi combinatorial algorithm.
First we define both the linear (both fixed and the random effect model) and nonlinear Gauss-Markov models in
section (2-1). While the solution of the linear Gauss-Markov model by Best Linear Uniformly Unbiased Estimator
(BLUUE) is straight forward, the solution of the nonlinear Gauss-Markov model has no straight forward procedure
to
. We therefore introduce in
owing to the nonlinearity of the injective function (or map function) that maps
Section (2-2) the Gauss-Jacobi combinatorial algorithm which we propose to use in solving nonlinear Gauss-Markov
model. In Section (2-3), we demonstrate how the procedure can be used to solve the nonlinear Gauss-Markov model.
j
2-1
Linear and nonlinear Gauss-Markov models
Presented in this Section are both the linear and nonlinear Gauss-Markov models. We start by the definition of the
linear Gauss-Markov model as follows
npo
Definition 2-0a (Special linear Gauss-Markov model ):

Given a real
random vector
of observations, a real
parameters over a real
coefficient matrix
a real
the functional model
lkm,
rk
sto %u9j
s"q*zyv{2n}|O$yv{2n+|or~v's(#_9s*
rkv
kd,
qio j
vector
of unknown fixed
positive-definite dispersion matrix
wp*ir{n&|O_;wf*z
wx
(2-1)
is called the a special linear Gauss-Markov model with full rank.
The unknown vector of fixed parameters in the special linear Gauss-Markov model (2-1) is normally estimated by
Best Linear Uniformly Unbiased Estimation BLUUE, defined in E. Grafarend and B. Schaffrin (1993, p. 93) as
Definition 2-0b ( Best Linear Uniformly Unbiased Estimation BLUUE):
An
vector
is
for (Best Linear Uniformly Unbiased Estimation respectively
k,
"q *Xn
mB7y
CHAPTER 2. NONLINEAR ADJUSTMENT
8
the
-BD
) in (2-1) when on one hand it is uniformly unbiased in the sense of
y"{ ^q | *zyv{LXnl|R*iqZ^qro j
(2-2)
and on the other hand in comparison to all other linear uniformly unbiased estimators give the minimum variance
and therefore the minimum mean estimation error in the sense of
where
Lr{ 1q |R*zyv{ vq Bq;#'yv{ vq Bq%|R*

*g
RwPz*p3=
*iv H
k matrix and an k, vector.
is a real
(2-3)
q in (2-1) leads to
Y
xq *`s w s Y s w" Y n
Using (2-3) to estimate the unknown fixed parameter vector
(2-4)
(2-5)
r{ 1q |*f's w Y s Y 0
The dispersion matrix (variance-covariance matrix) w is unknown and is obtained by means of estimators of type
MINQUE, BIQUUE or BIQE as in C. R. Rao (1967, 1971, 1973 and 1978), C. R. Rao and J. Kleffe (1979), B.
Y
Schaffrin (1983), and E. Grafarend (1985). In the event that s w s is not regular (i.e. s has a rank deficiency), the
with its regular dispersion matrix
rank deficiency can be overcome by procedures such as those presented by E. Mittermayer (1972), E. Grafarend and
B. Schaffrin (1974), E. Grafarend and B. Schaffrin (1993, pp. 107-165), F. K. Brunner (1979), A. Peremulter (1979),
P. Meissl (1982), E. Grafarend and F. Sanso (1985) and K. R. Koch (1999, pp.181-197) among others.
Definition 2-1a ( Gauss-Markov model with random effects):
The model
n*vS9RB/
vy {n&| *yv{|o %u9
r{n&|R*zr{2nD` |dr{|L o
{2n2| *
(2-6)
y"{|O$yv{2n&|Ow n (w CX
) A }*_9*i
of observations, a real ?km, vector of unknown random effects
with a real mkm, random vector npo
%u/ of rank L;*A is called the Gauss-Markov
(zufallseffekte), a non stochastic real valued matrix o
model with random effects.
2kR,
The
random effect vector of the model (2-6) can be predicted by the Best Linear Uniformly Unbiased Prediction
BLUUP defined by (E. Grafarend and B. Schaffrin 1993, p.304 ) as
Definition 2-1b ( Best Linear Uniformly Unbiased Prediction BLUUP):
An
vector is called h omogenious BLUUP (Best Linear Uniformly Unbiased Prediction) of
special linear Gauss-Markov model with random effects (2-6) if
(1st) is a homogenious linear form
}kl,
and uniformly unbiased in the sense
in the
*Xn
(2-7)
yv{1 |R*zyv{n&| *yv{|
(2-8)
(2nd) in comparison to all other linear uniformly unbiased predictors give the minimum mean square predictor
error in the sense of
2 X
>@*
>*yv{cx BD1N BDT|R* Lr{1x BD| *
* _ r{2n}| +r{C| n 1C *

*p3
w n Syv{2 B132 B1 |R*zP H
with MSPE symbolizing Mean Square Predictor Error and is a real k
(2-9)
matrix .
2-2. THE GAUSS-JACOBI COMBINATORIAL ALGORITHM

Using (2-9) to predict the unknown random parameter vector
in (2-6) leads to
r *`' w n Y Y w n Y n
(2-10)
with its regular dispersion matrix
{1 |R*`' wn Y Y
r
%u9 exist.
if r{n&|R*zr{2nDf|dr{|L o
(2-11)
Definition 2-2 (Nonlinear Gauss-Markov model):

The model
(2-12)
yv{2n}|R*nBD*sq%Nr{n&| *w
of observations, a real k, vector q-o j of unknown fixed
with a real ikm, random vector no
parameters, SkS, vector of random errors (with zero mean and dispersion matrix w ), s being an injective
C and y the expectation operator is said
function from an open domain into +B dimensional space
to be a nonlinear Gauss-Markov model (E. Grafarend and B. Schaffrin 1989).
The difference between the linear and nonlinear Gauss-Markov models defined above lies on the injective function
In the linear Gauss-Markov model, is linear and thus satisfies the algebraic axiom
sY}S
&*zsY2Cs
N#o #YL#
o j
sT0
(2-13)
For the nonlinear Gauss-Markov

q j being
flat.tracing
to
an -dimensional manifold
with the -dimensional manifold traced by

for varying values of
is a nonlinear vector function that maps
model on the other hand,
that is curved.
sT0
In Geodesy, many nonlinear functions are normally assumed to be moderately nonlinear thus permitting linearization
by Taylor series expansion and then applying the linear models to estimate the unknown fixed parameters (e.g. K.
R. Koch 1999, pp.155-156). Whereas this may often hold, the effect of the nonlinearity of the models may still be
significant on the outcome and as such, we revert in the next Section to the Gauss-Jacobi combinatorial algorithm.
2-2
The Gauss-Jacobi combinatorial algorithm
In this Section we present the Gauss-Jacobi combinatorial algorithm which is neither iterative nor requires linearization of the nonlinear observation equations for the solution of nonlinear Gauss-Markov model. Linearization is permitted only for the nonlinear error propagation/variance-covariance propagation in order to generate the dispersion
matrixn (i.e. the second moments). We start by stating the Gauss-Jacobi combinatorial lemma in Box (2-1) and
refer to S. Wellisch (1910, pp. 46-47) and T. Hornoch (1950) for the proof that the results are equivalent to those
of least squares solution. The levelling network is presented to illustrate the Gauss-Jacobi combinatorial approach.
We conclude the Section by Theorem (2-1) which allows the applicability of the Gauss-Jacobi combinatorial algorithm in solving nonlinear Geodetic observation equations once they have been converted into algebraic (polynomial)
equations.
We state the C. F. Gauss and C. G. I Jacobi (1841) combinatorial lemma as follows:
10
For
Box 2-1 (Lemma 2-1: Gauss-Jacobi combinatorial):

algebraic observation equations in unknowns
Y ]vd Y a=Ba Y
*

]vd
a=Ba

*
[ ]vd [ a=Ba [
*
0302030
(2-14)
aO! )}o{O,./2030302#|
stor~P %u/j
for the determination of the coordinates and of the unknown point ,

being the observable and
being the elements of the design matrix
there exist no set of solution
from
any combinatorial pair in the equations above that satisfy the entire system of equations. This is because the solution
obtained from each combinatorial pair of equations differs from the others due to the unavoidable random measuring
and
errors. If the solutions from the pair of the combinatorial equations are designated
with the subscript indicating the combinatorials, then the combined solution is the sum of the weighted arithmetic mean
!#$(!$ )o{c,$.93020303|
]1Y
#]
[03020

^

Y
^
]
Y

N
[
]

[
z
3
0
2
0
3
0
0
^

Y
c
a
Y

N
[
a
}
[
i
2
0
3
0
2
0
0
]*
Y

[ i
0203020 /av*
Y

[ i
0203030
{]C#a|
acY
#a
[_102030
(2-15)
Y
'
[ 102030 being the weights of the combinatorial solutions given by the square of the determinants as
Y
*`' Y
BD
Y

[X*`'
([ZBD[2

(2-16)
0302030
The results (2-15) coincides with the center of the error figure (see Figure in Appendix A.4 for *Q ) formed by the
with
coordinates of the combinatorial solutions and are identical to those of least squares solution for linear observation
equations.
We illustrate the Gauss-Jacobi combinatorial approach using the levelling network below.
Levelling network: Consider a levelling network with four-points in Figure (2.1) below.
Figure 2.1: Levelling Network
Let the height of point
Y
be given as
Y and those of
and [ be sought. We have

;
.d * 9. 'BD.O( *
2-2. THE GAUSS-JACOBI COMBINATORIAL ALGORITHM
11
[
number of combinatorial routes that can be used to obtain the heights of points
and . These are levelling routes
and
These combinatorials sum up to the outside loop
The observation equations are written as
of the network
+YB

B \XBYL_

B [XBDC\ B
Y?B
BD [ZBD+YL0
[B+YBC\BD[_0
]

B Y *ia Y
] [ B Y *ia
] [ B]
*ia [
(2-17)
] \ B Y *ia \
] \ B]
*ia
] \ B] [ *ia
which can be expressed in the form of the special linear Gauss-Markov model (2-1) in page (7) as
acY+lY ,
a lY ,
]

a[
y a_\lY * B , , , ]%
[
(2-18)
a B, , ];\
a
B,,
where aY#a 302030203#a are the observed height differences, ] #]%[#]%\ are the unknown heights of points [_\
respectively. Let the dispersion matrix r{2n}|P*pw be chosen such that the correlation matrix is unit (i.e. w*f[P*
w Y _3) )UZB L)U1) cL%wv Y *z*z ), the decomposition matrix and the normal equation matrix s w Y s
be given respectively by

a Y d Y
a
l Y
B,B,
Y
(2-19)
8* BFa \ l Y a l Y s w s* B, B, 0

B
,

B
,
a
BZa

aO
BZaO
We compute the heights of points and [ using the combinatorial procedure as follows:
1st, 2nd and 3nd Combinatorials (levelling along routes: L c#,?>@*i+YBr BrC\B+YLL .>@*z B[&B
C\XBD
and L O>@*[ZBD+YB\XBD[ ):
(2-19) and (2-4) leads to the partial solutions
a aO[
acY& Y B a B a_\
. B .
. . .
aY B a\
;q

* Y a [ B aO
;q Y
*
Y
. .
. .
aO aO
a Y Y a
. B .
. B . . Ba_\
(2-20)
_
a
\
a
. B .
_
a
\
a
;q [
*
Y . . B . Ba

^Y a a
. B .
B . a \
The heights of the stations ] ]%[#];\ would then be given by the summation of the combinatorial solutions
a Y Y B a [ B a

. . .
q * q + q q *

Y a Y a [ B Y Ba
0
(2-21)
. . .
aY B a
. .
12
+YB
D
B [XB+Y ) and with (2-4) we have
Y
[
a
a
aY&d^Y
a Y . B . B .
BFa
l Y
[
a
Y
Y
Y
Y
a Y a [ B Y Ba
q Z*f's w sx s w
*

. . .
acY B a
. .
which are identical to the Gauss-Jacobi combinatorial solution in (2-21) "0
Levelling along route
(2-22)
Having stated and illustrated using a levelling example the Gauss-Jacobi combinatorial lemma for linear equations,
we state below the theorem that allows the solution of nonlinear equations in Geodesy.
Box 2-2 (Theorem 2-1):
Given algebraic (polynomial) observational equations ( observations, where is the dimension of the observation
space ) of order in
variables (unknowns) ( is the dimension of the parameter space ), the application of
least squares solution (LESS) to the algebraic observation equations gives
as the order of the set of nonlinear
algebraic normal equations. There exists normal equations of the polynomial order
to be solved.
Proof: Given nonlinear algebraic equations
expressed as
L!o{2 Y 030203 j |
Y o{2 Y 1030203# j |

o{2YL1030203# j |
0
0
0
o{Y020302# j |O0
._;BS,
._^Bm,
(2-23)
and the order considered as , we write the objective function to be minimized as
3+
* Y
z0302030l
1 ! o{Y020302# j |
{2 Y 1030203 j |
(2-24)
. The
and obtain the partial derivatives (first derivatives) of (2-24) with respect to the unknown variables
order of (2-24) which is then reduces to
upon differentiating the objective function with respect to the
variables
Thus resulting in normal equations of the polynomial order
.
Y 1030203# j 0
._B-,
'.^Bl,
Example (pseudo-ranging):
For pseudo-ranging or distance equations, the order of the polynomials in the algebraic observational equations is
. If we take the pseudo-ranges squared or distances squared, a necessary procedure in order to make the
observational equations algebraic or polynomial, and implement LESS, the objective function which is of order
reduces by one to order
upon differentiating once. The normal equations are of order
as expected.
*t.
*
The significance of the theorem above is that by using the Gauss-Jacobi combinatorial approach to solve the nonlinear
Gauss-Markov model, all observation equations of geodetic interest are successfully converted to algebraic or
polynomial equations.
2-3
Solution of nonlinear Gauss-Markov model
Having presented and proved the Gauss-Jacobi combinatorial lemma using simple linear levelling and linearized
ranging examples, we proceed to solve the nonlinear Gauss-Markov model in two steps:
Step 1: Combinatorial minimal subsets of observations are constructed and rigorously solved by means of the
Multipolynomial resultant or Groebner basis (J. L. Awange and E. Grafarend 2001).
2-3. SOLUTION OF NONLINEAR GAUSS-MARKOV MODEL
13
Step 2: The combinatorial solution points of step 1 are reduced to their final adjusted values by means of an
adjustment procedure where the Best Linear Uniformly Unbiased Estimator (BLUUE) is used to estimate the
vector of fixed parameters within the linear Gauss-Markov model with the dispersion matrix of the real valued
random vector of pseudo-observations from Step 1 generated via the nonlinear error propagation law also
known in this case as the nonlinear variance-covariance propagation.
2-31 Construction of the minimal combinatorial subsets
de
we construct the minimal combinatorial subsets comprising equations solvable in closed form using
Since
either Grbner bases or Multipolynomial resultants which we present in Section (2-32). We begin by the following
elementary definitions:
Definition 2-3 (Permutation):
Given a set with elements
, the arrangement obtained by placing !
in some sequence
is called permutation. If we choose any of the elements say first, then each of the remaining elements
can
be put in the second position, while the third position is occupied by the unused letter either or . For the set
, the following permutations can be made:
{) O$|o
{2) |o
#)
) $O)
) )# %O)
"
O$
(2-25)
From (2-25) there exist three ways of filling the first position, two ways of filling the second position and one
. In general, for
way of filling the third position. Thus the number of permutations is given by
different elements, the number of permutation is equal to
kr.vk,*:
Dk02030Zkrvkr.Pk,*i+
Definition 2-4 (Combination):

If for elements only elements are used for the permutation, then we have a combination of the th order.
If we follow the definition above, then the first position can be filled in ways, the second in
ways and
the th in
ways. In (2-25) above the combinations are identical and contain the same elements in
different sequences. If the arrangement is to be neglected, then we have for elements, a combination of th
order being given by
B Bl,
Bi,
& FBl,^030202FB z, 0

+
*
(2-26)
B N
kp03020Zkvk.Pk,

Given nonlinear equations to be solved, we first form
minimal combinatorial subsets each consisting of
elements (where is the number of the unknown elements). Each minimal combinatorial subset is solved using the
algebraic procedures discussed in (2-32). In geodesy, the number of elements normally consist of the observations
in the vector n , while the number of elements normally consist of the unknown fixed parameters in the vector q .
*
2-32 Closed form solution of the minimal combinatorial subsets

In this Section, we present two algebraic algorithms that are used to solve in closed form the minimal combinatorial
subsets constructed in Section (2-31). We first present the approach based on the Grbner bases and thereafter consider
the Multipolynomial resultants approach.
2-321
Closed form solution of nonlinear equations by Grbner bases
As a recipe to what Grbner bases can do, consider that most problems in nature, here in Geodesy, Photogrammetry,
Machine Vision, Robotics, Surveying etc. can be modelled by sets of nonlinear equations forming polynomials. These
nonlinear systems of equations that have to be solved can be used to form linear combinations of other polynomials
called Ideals by being multiplied by arbitrary polynomials and summed up. In this case, a collection of these nonlinear algebraic equations forming Ideals are referred to us the set of polynomials generating the Ideal and forms
the elements of this Ideal. The B. Buchberger algorithm then takes this set of generating polynomials and derive,
using a procedure that will be explained shortly, another set of polynomials called the Grbner basis which has some
special properties . One of the special properties of the Grbner bases is that its elements can divide the elements of
14
the generating set giving zero remainder. This property is achieved by the B. Buchberger algorithm by canceling the
leading terms of the polynomials in the generating set and in so doing deriving the Grbner basis of the Ideal (whose
elements are the generating nonlinear algebraic equations). With the lexicographic type of ordering chosen, one of
the elements of the Grbner basis is often a univariate polynomial whose roots can be obtained by the roots command
in MATLAB. The other special property is that two sets of polynomial equations will generate the same Ideal if and
only if their Grbner bases are equal with respect to any term ordering. This property is important in that the solution
of the Grbner basis will satisfy the original solution required by the generating set of nonlinear equations.
With this brief outline, we now define the term Grbner basis. In the outline above and definitions below, we have
used several terms that require definitions. In Appendix A.1 (p.97), we present the definitions of the terms as follows;
monomial (Definition A-1), polynomial (Definition A-2), linear algebra (Definition A-3), ring (Definition A-4), polynomial ring (Definition A-5), Ideal (Definition A-6), monomial ordering (Definition A-7) and leading coefficients (LC),
leading terms (LT), leading monomials (LM) and multideg (Definition A-8). Indeed what we present here is just a
definition of the term Grbner basis and as such, any reader interested on the subject may consult text books such as
J. H. Davenport et al. (1988), T. Becker and V. Weispfenning (1993, 1998), B. Sturmfels (1996), F. Winkler (1996), D.
Cox et al. (1997), B. Buchberger and F. Winkler (1998), and W. V. Vasconcelos (1998). We start by first defining an
Ideal and the Hilbert Basis Theorem that guarantees that every Ideal has a finite set of Grbner basis before defining
the term Grbner basis and briefly looking at the B. Buchberger algorithm which is the engine behind the computation
of Grbner bases.
Definition 2-5 (J. H. Davenport et al. 1988, p.96, D. Cox et al. 1997, p.29):
An Ideal is generated by a family of generators as consisting of the set of linear combinations of these genera) then
%& be polynomials in ('
tors with polynomial coefficients. Let
]^Y_02030#]
Y02030$
Y02030$%&e *+*
&
! ! YL03020 -&or.' ]^Y_103020 ] /) 00

(2-27)
! Y
Y 0203030$ & e is an Ideal and if a subset 1 2A.' ] Y C03030#] ) is an Ideal, it must satisfy the following conditions;
(a) v3
o 1.
(b) If 4x3
o 1% then P5
46
o 1 (i.e. 1 is an additive subgroup of the additive group of the field ).
o 1 and or. ' ]^Y_03020 ] ) , then %7
o 1 (i.e.1 is closed under multiplication ring element).
(c) If 6
,
The definition of an Ideal can be presented in terms of polynomial equations

by expressing the system of polynomial equations
Y C03030$ & o.' ] Y 103030#]

)
. One begins
*
Y

*
0
0
&Z*i9
%
(2-28)
!
! o
YL03020 e
YL03020 oQ ]1YL02030#]
which can be used to derive others by multiplication of the individual equation

by another polynomial
) and adding to get
('
-& %&
%&
(cf. 2-27). The Ideal
thus con) , then
%&
%&
8'
sists of a system of equations
thus indicating that if
%&
%& (thus forming the basis of the Ideal 1 ), being the basis of the Ideal
is an Ideal generated by
1 .
^] Y_02030#]
YL03020 Xe
^ YNYZ 03020cA ( *t

Yr*8
*
03
020* *
YL03020
Perhaps a curious reader may begin to wonder why the term Ideal was selected. To quench this curiosity we refer to
W. K. Nicholson (1999, p.220) and quote from T. Becker and V. Weispfenning (1993, p.59 ) on how the term Ideal and
ring came to be. They write:
On the origin of the term Ideal, the concept is attributed to Dedekind who introduced it as a set theoretical
version of Kummers Ideal number to circumvent the failure of unique factorization in certain natural
extension of the domain 9 . The relevance of Ideal in the theory of polynomial rings was highlighted by the
Hilbert Basis Theorem. The systematic development of Ideal theory in more general rings is largely due
to E. Noether. In the older literature the term module is sometimes used for Ideal (cf. F. Macaulay
1916). The term ring seems to be due to D. Hilbert; Kronecker used the term order for ring.

Example 2-1: Given the polynomials in
' ^] Y_#]
]
Y
]
15
]%[ ) as
l._ Y
] Y ]

]
S z
*
l.
[ ]
] [ ]
[
l
*
l._V [(Y ] [ ] Y S]
Y d
V *i
(2-29)
* ]
Y l._ Y
] Y ]
S
]
S #]
l.
[ ]
] [ ]
[ d #]
[ d.V [Y ] [ ] Y

Example 2-2: Given the polynomials in ' ]^YL#] ]%[O#];\ ) as
]
Y B._ W ] Y S]
B. W ]
S]
[ BD.V W ] [ B] \
l. W ] \ S W
l W
dV W
]
Y B._ Y ] Y S]
B. Y ] S]
[ BD.V Y ] [ B] \
l. Y ] \ S
Y l
Y dV
Y
]
Y B._
] Y S]
B.
]
S]
[ BD.V
] [ B] \
l.
] \ S
l
dV
]
Y B._ [ ] Y S]
B. [ ]
S]
[ BD.V [ ] [ B] \
l. [ ] \ S
[ l
[ dV
[
then
Ideal 1
then Ideal 1 is given by
BD.OWN]
]
[ B._V(WN][ZB] \
l. W3]%\d W
l W
dV W
BD.O3Y$]
]
[ B._V2Y$][ZB] \
l. Y]%\d
Y l
Y dV
Y
BD.O
]
]
[ B._V
][ZB] \
l.
]%\d
dV
BD.O([(]
]
[ B._VN[(][ZB] \
l. [N]%\d
[ l
[ dV
[

Example 2-3: Given the polynomials in ' ] Y #] ] [ #] \ ) as

BFB]^Y2Cd
]^YN];\R*
B-B] Cl
] ] \ *z
BmB] [
Cl
] [
] \ *z

]
Y l
]
d
]
[ BD

*
The
BF Br]^Y2%
]^YN];\2BiBr]
] ];\O3BUDBr]%[%
]%[2];\
]
Y
1 c:1*
]
Y
]
Y
]
Y
]
Y
BD.OW2]^Y&S]
BD./Y(]^Y&S]
BD.
]^Y&S]
BD.c[3]^Y&S]
S]
Y SV e
W

*

Y *

*

[ *
W

Y e

[ 0
(2-30)
(2-31)
(2-32)

]
[ B

e
Having defined and given examples of an Ideal, we present the definitions of the division algorithm exploited by
B. Buchberger algorithm before defining the Grbner basis (Standard Basis) of an Ideal. We begin by stating the
Hilbert Basis Theorem, which gives the assurance of the existence of the Grbner basis. In general, we will denote
/) a collection of polynomials
%& with variables in
by .'
and coefficients in any field .
)
.'
forms a polynomial ring (Definition A-5, Appendix A.1 ).
1] YL102030#]
] Y C03030#]
YL103030$
OY02030$
W
Lemma 2-2 (Division Algorithm, D. Cox et al. 1997, theorem 3, p.61, theorem 4, p.47):
;
) .
%& be an ordered S-tuple polynomial in 8'
Fix a monomial order > on 9
, and let <
)
)
& &
Then every
.'
can be written as
where
.'
and either
or a linear combination with coefficients in of monomials, none of which is divisible by any
) has a finite
&
of
The Hilbert Basis Theorem assures that every Ideal 1=2 .'
generating set, that is 1
for some 4
4
4 &
>4 & 71
or ] Y 103020 ]
=*
&6' Y N03020 $&6= (0
* Y 03020 e
-*`'Y_103020 (
]1Y02030#]
*z Y Y

02030 L
!#o ] Y C03030#]
t ] Y 03020 ]
Y C03030 o ;0
The finite generating set ? in Hilbert Basis Theorem is what is known as a basis. Suppose every non-zero polynomial
is written in decreasing order of its monomials:
,
! Y
!]%!$!A*z
@ 9/]!&ed]!CB Y
(2-33)
if we let the system of generators of the Ideal be in a set ? , a polynomial is reduced with respect to ? if no leading
monomial of an element of ? (LM ( ? )) divides the leading monomial of (LM( )). The polynomial is said to be
completely reduced with respect to ? if no monomials of is divisible by the leading monomial of an element of ?
(J. H. Davenport et al. 1988, pp. 96-97).
The introduction of the division algorithm given in Lemma 2-6 above fits well to the case of univariate polynomials
as the remainder can uniquely be determined. In the case of multivariate polynomials, the remainder may not be
16
{YL$
|
uniquely determined as this depends on the order of the divisors. The division of the polynomial < by
may
not necessarily give the same remainder as the division of < by
in whose case the order has been changed.
This problem is overcome if we pass over to the Standard basis having been assured of its existence for every Ideal
by the Hilbert Basis Theorem in Lemma 2-6.
{
$Y|
The Standard Basis ? which completely reduces the polynomial

known as the Grbner basis and is defined as follows:
and uniquely determines the remainder
is also
Definition 2-6 (J. H. Davenport et al. 1988 p.97, D. Cox et al. 1997, Definition 1, p.100):
A system of generators (or basis) ? of an Ideal 1 is called a Standard Basis or Grbner basis (with respect to
the order <) if every reduction of D1 to a reduced polynomial (with respect to ? ) always gives zero as a
remainder. The above definition is a special case of a more general definition given as: Fix a monomial order
) Given
) then reduces to zero Modulo ? ,
and let ?
4
>4 & E2 8'
8'
written as
(2-34)
3FHG
zo
*{ Y 10302030 | f ] Y 0302030#] 0
if
mod ] Y 0302030#]
D9
can be written in the form
* Y 4 Y 02030d 4
such that whenever !I4!J*
@ 9 we have multideg( ) K multideg( !L4!
(2-35)
).
Suggested by W. Grbner in 1949 and developed by his student at the time B. Buchberger in 1965, B. Buchberger
decided to honour his thesis supervisor W. Grbner(1899-1980) by naming the Standard basis for Ideals in polyno/) as Grbner basis (B. Buchberger 1965). Grbner bases has become a household name in
mial ring .'
algebraic manipulations and finds application in fields such as Statistics and Engineering. It has found use as a tool
for discovering and proving theorems to solving systems of polynomial equations as elaborated in publications by B.
Buchberger and F. Winkler (1998). Grbner bases also give a solution to the Ideal membership problem. By reducing
a given polynomial with respect to the Grbner basis ? , is said to be a member of the Ideal if the zero remainder
) and let
)
is obtained. Thus let ?
4
4 & be a Grbner basis of an Ideal 1M2 .'
.'
be a polynomial, then 31 if and only if the remainder on division of by ? is zero. Grbner bases can also be used
to show the equivalence of polynomial equations. Two sets of polynomial equations will generate the same Ideal if
and only if their Grbner bases are equal with respect to any term ordering. This implies that a system of polynomial
%&
equations
will have the same solution with a system arising from any
%& with respect to any term ordering. This is the main property of the Grbner bases that is
Grbner basis of
used to solve a system of polynomial equations as will be explained below.
]1Y02030#]
Q*-{ Y 03020 |
o
i ] Y C03030#]
o ] Y 03020302#]
YL]^Y_030203] X*Q;030203 F]^Y_020303] X*f

OY030203
B. Buchberger algorithm
Given polynomials 4
>4 & N1 , the B. Buchberger algorithm seeks to derive the standard generators or the
Groebner basis of this Ideal. Systems of equations 4
to be solved in practise are normally
>4 &
formed by these same polynomials which here generating the Ideal. The B. Buchberger algorithm computes the
>4& O1 and computes
Groebner basis by making use of pairs of polynomials from the original polynomials 4
the subtraction polynomial known as the
explained in D. Cox et al. (1997; p. 81) as follows:
Y 03020302
Y * 902030203
cY02030203 o
BLFa )U
BLFa )U ):

d
z
o
] Y 02030] ) be two non-zero polynomials. If multideg 1=* and multideg I49=* , then let
P
* P Y 103020 P , where P !*i"JRQ {UT !T%!#| Tfor each ) T . ]-S is called the Least Common Multiple (LCM) of LM(
) and LM(4 ) expressed as ] S *z
{2 '1N #4/| . The B%LFa )c of and 4 is given as
]S

]S
X'>49&*
(2-36)
}6='1 B &6=# 4/ 40
] S , -] S with polynomial coefficients

The expression above gives as a linear combination of the monomials
&6=1 &6I 49
and 4 and thus belongs to the Ideal generated by and 4 .
Definition 2-7 (
Let 4
.'
B%La )
* { cY 102030 2|
Definition 2-8 (Grbner basis in terms of

):
>4 reduces to zero
A basis ? is Grbner basis if and only if for every pair of polynomials and 4 of ? ,
V4
>4& for an Ideal 1 is a Grbner basis if and only if
with respect to ? . More generally a basis ?

>4 WF G
XY
@ .
Z 9
9) *
X' 9
17
xo
UT T 1($ T I49#
Z' /
9 0 The concept
The implication
of Definition 2-8 is the following: Given two polynomials 4 3? such that
T
T
I4 then the leading monomials of and 4 are relatively prime leading to
4 WF G
of prime integer is clearly documented in K. Ireland and M. Rosen (1990 pp. 1-17).
'1N0
9
BLac ) ):
Example 2-4 (
Consider the two polynomials
Y *i]
Y l._ Y
] Y ]
]
S
(2-37)
*i]
l.
[ ]
] [ S]
[ l
of Example A-1 in Appendix A.1, the XB&%Lac ) can then be computed as follows: First we choose a lexicographic
ordering {2] Y ed] el] [ | then

T I 4 Y }*i]
Y U; T T # 4 T
&*z]
T &6I 4 Y }*i]

Y ;&6=# 4
}*]
I 4 Y N # 4
&*i]
Y ]
]
]
F]
Y d. Y
] Y ]
S]
d B ] ]
Y ]

]
l.
[ ]
] [ ]
[ l
(2-38)
XI4 Y 4 }*
Y
[
\
*z]
]
Y d./Y
]1Y]
S]
d ]
B]
Y ]
B.

[N]
Y [ ]
][Z\ B]
Y ]
[ B ]
Y
*`B ]
Y BD.O
[N]
Y ]
]%[ZB]
Y ]
[ l._9Y
]1Y]
]
d ]
Example 2-5 ( BLac ) ):

"
4
4
Consider Example A-2 in Appendix A.1 for pseudo-ranging problem. The first two polynomials equations are given as
BD.OWN]
]
[ B._V(W2]%[ZB] \
l. W3]%\d W
d W
dV W
W
BD.O3Y$]
]
[ B._V2Y(]%[ZB] \
l. Y]%\d
Y d
Y dV
Y
Y 0
By choosing the lexicographic ordering {] Y em] el] [ em] \ | , the ?B polynomial is computed as follows
T
T
UT I 4 [ &T *z]
Y T I4 \ &*z]
Y }6=I4 [ *]
Y }6=I4 \ *]
Y
I 4 [ N # 4 \ #*]
Y

]
Y I4 [ +B ]
Y #4 \ &*=4 [ BZ4 \
XI4 [ 4 \ }*
XI 4[_ 4_\2} *.9]
Y 9YBD T] ]^
Y Y&l.93YBD ]
l.9VY BV ][
.; B YNT];\ZS BD/Y}d B3Y+dV BDV2Y& WXB Y
"
[*]
Y D
B .cWN]^Y&S]
4_\ *]
Y D
B .9Y$]^Y&S]
4
BLa )
(2-39)
(2-40)
Example 2-6 (
):
polynomial, let us consider the minimum distance mapping
As an additional example on the computation of
problem of Example A-3 in Appendix A.1. The last two polynomials equations are given as
?B
*z]%[l
]%[N]%\XBd
(2-41)
*
]
Y l
]
S
]
[ BD

By choosing the lexicographic ordering {3] Y el] el] [ ed] \ | , the ZB polynomial is computed as follows
T
T
UT #43}T *] [ $ T # 43&*i]
Y $&6=# 4N}*] [ $&6=# 4N}*z
]
Y
' I 42( I 43}*z]

Y ] [
]
Y ]%[
]
Y ]%[
(2-42)
ZI4>4&* ] [ F] [ l
] [ ] \ B B
]
Y '
]
Y l
]
d
]
[ BD

[
ZI4 >4 &*-B X]
&l
]
Y ][3]%\XB]
][ZB ] [ d
]%[
Example 2-7 (computation of Grbner basis from the B%La )' ):

"
4
4
By means of an Example given by J. H. Davenport et al. (1988, pp. 101-102) , we illustrate how the B. Buchberger algorithm works. Let us consider the Ideal generated by the polynomial equations
Y *i] [ abB]%b
4
*i];a
bB];a/b
4 [ *i]%
2a
?BDb
Z cY [
(2-43)
] e8aQe b adopted. The -polynomials to be considered are Z#4cY4

,
Z
> 4[2 and show that the result is used to supress 4cY upon which any
with the lexicographic ordering

I4 4
and #4 >4 . We consider first I4
Z
[2
18
Z cY !
]
a

Z cYL and XI4cY4[3 ) containing 4cY will not be considered. }6=I4

*z];a
b%Z&6=#4[N*

[3}
*i]
a
b respectively
]
a
b 4 B ]
a
b 4[
XI 4 4 [ }*
(2-44)
*`]
a
bB] ]%
aa b
b BF]
a ]
b
a BD
b

*`BZ]
abdb
We immediately note that the leading term of the resulting polynomial LT( Z# 4 4 [ ) is not divisible by any of
the leading terms of the elements of ? , and thus the remainder upon the division of XI4 4 [ by the polynomials
in ? is not zero (i.e. when reduced with respect to ? ) thus ? is not a Grbner basis. This resulting polynomial
after the formation of the ZB polynomial is denoted 4_\ and its negative (to make calculations more reliable)
added to the initial set of ? leading to
[
4 Y *z] a/bB]b
4
*z];a/
3bB]%ab
(2-45)
4
[*z]
a
Bb
4_\ *z]
a/bBb
0
The ?B polynomials to be considered are now ZI 4 Y > 4 N
Z# 4 Y 4 [ ( ZI 4 Y > 4 \ N Z# 4

4 \ and ZI 4 [ > 4 \ N0 In the
set of ? , one can write 4cYZ*z]4_\ leading without any change in ? to the suppression of 4cY leaving only ZI4 >4_\
and Z# 4[_ 4_\2 to be considered. Then

"
ZI 4 > 4 \ &*z
BZ]
a/b ] S4
aBb
a 4 \
(2-46)
*Q
I4 >4
pair #4 >4 U
(e.g.
T
I4 4
then
v*pab
Bdb
, a non zero value thus the set ? is not a Grbner
*]%a
bBD];a/b%
4 [ *]
a
BDb;
(2-47)
4 \ *]
a/bBb

4*a/b
Bb

the ZB polynomials to be considered are now ZI 4 [ > 4 \ N% ZI 4 > 42N%ZI4 [ >42 and Z#4 \ >4(0 We then compute
"
Z# 4[_ 4_\2}*i
b 4[?Ba 4_\
(2-48)
*za/b
BDb
which upon subtraction from 4 reduces to zero. Further,
ZI 4 > 43}*z%
BZ];a/b
b S4
];Ba/b ];
a 4
*`
(2-49)
*z
which can be reduced by adding 4 to give 4

basis. This value is added to the set of ? to give
and
Z \ >42&*b4 \ B[ ]
a4
*i]
2bc
Bb
which is added to ? as 4 giving
4
*]%a
bBD];a/b%
4 [ *]
a
BDb;
4 \ *]
a/bBb

4*a/b
Bb

[
4*]
a
BDb
The polynomials to be considered are now ZI 4 [ > 42N
Z# 4
> 42( Z# 4 [ 43(XI4 \ 42
illustrate that all this B%LFa )c reduces to zero as follows
Z#4 [ 42&*b
4 [ B]
a 4*i]
a/b
Bb [ Bb [ 4 \ *z
Z# 4 42&*z]%
b 4
Ba
4*QBZ]
[ a
b
[ Sa
b a
4 \ *z
Z# 4[_ 4 &*b

4[XBa
4 *za
b
Bb Bma/bBb! 4 *
Z# 4_\ 4 &*%b 4_\?Ba 4 *za/b [ BDb [ BDb 4 *
Z# 4 4 &*z]
4 B
a 4 *QBZ]
b
a/b [ [
4 BDb 4 *90
"
I4
(2-50)
(2-51)
X _ 2N0 We now
and I4 4
(2-52)
19
Thus equation (2-52) comprise the Grbner basis of the original set in (2-43).
LFa )'
The importance of the Sis that they lead to the cancellation of the leading terms of the polynomial pairs
involved. In so doing the polynomial variables are systematically eliminated according to the polynomial ordering
chosen. For example if the lexicographic ordering
is chosen, will be eliminated first, followed by
and the final expression may consist only of the variable . D. Cox et al. (1998, p.15) has indicated the advantage of
lexicographic ordering as being the ability to produce Grbner basis with systematic elimination of variables. Graded
lexicographic ordering on the other hand has the advantage of minimizing the amount of computational space needed
to produce the Grbner basis. The procedure is thus a generalisation of the Gauss elimination procedure for linear
systems of equations. If we now put our system of polynomial equations to be solved in a set ? ,
combinations
can be formed from the set of ? as explained in the definitions above. The theorem, known as the Buchbergers
polynomial criterion, gives the criterion for deciding whether a given basis is a Grbner basis or not. It

suffices to compute all the Sand check whether they reduce to zero. Should one of the polynomials
not reduce to zero, then the basis fails to be a Grbner basis. Since the reduction is a linear combination of the elements
of ? , it can be added to the set ? without changing the Ideal generated. B. Buchberger (1979) gives an optimisation
criterion that reduces the number of the Salready considered in the algorithm. The criterion states
that if there is an element of ? such that the leading monomial of (LM( )) divides the LCM( >4 \? ), and if

>4 have already been considered, then there is no need of considering
>4 as this reduces to zero.
]e l
a eb
b
BZ%c)U
lBc)
Z1 Z@ 9
LFa )U
LFa )U
Z 9
o
The essential observation in using the Grbner bases to solve a system of polynomial equations is that the variety
(simultaneous solution of system of polynomial equations) does not depend on the original system of the polynomials
<
%& but instead on the Ideal 1 generated by < . This therefore means that the variety
1 and one
makes use of the special generating set (Grbner basis) instead of the actual system < . Since the Ideal is generated by
< , the solution obtained by solving for the affine variety of this Ideal satisfies the original system < of equations. B.
Buchberger (1970) proved that ]1 is void, and thus giving a test as to whether the system of polynomial < can be
solved , if and only if the computed Grbner basis of polynomial Ideal 1 has
as its element. B. Buchberger (1970)
further gives the criterion for deciding if 1 is finite. If the system has been proved to be solvable and finite then F.
Winkler (1996, theorem 8.4.4, p.192) gives a theorem for deciding whether the system has finitely or infinitely many
solutions. The theorem states that if ? is a Grbner basis, then a solvable system of polynomialT equations has finitely
there is a polynomial 4 6? such that
is a pure power
many solutions if and only if for every
U^ _^
#4
of . The process of addition of the remainder after the reduction by the
and thus expanding the
generating set is shown by B. Buchberger (1970), D. Cox et al. (1997 p.88) and J. H. Davenport et al. (1988 p.101) to
terminate.
p>*Q{YL03020 2|
`*x
c
x
]!#, d) l+
]%!
{,|
!o
DBLFa )'
!
The B. Buchberger algorithm, more or less a generalization of the Gauss elimination procedure, makes use of the
subtraction polynomials known as the
in Definition 2-8 to eliminate the leading terms of a pair
of polynomials. In so doing and if lexicographic ordering is chosen, the process end up with one of the computed

being a univariate polynomial which can be solved and substituted back in the other
using the Extension Theorem (D. Cox et al. 1998, pp.25-26) to obtain the other variables. The Grbner bases approach adds to the treasures of methods that are used to solve nonlinear algebraic systems of equations in Geodesy,
Photogrammetry, Machine Vision, Robotics and Surveying.
mBD%La )'
?B+LFa )'
?B+%La )'
Having defined the term Grbner basis and illustrated how the B. Buchberger algorithm computes the Grbner bases,
we briefly mention here how the Grbner bases can be computed using algebraic softwares of Mathematica and Maple.
In Mathematica Versions 2 or 3, the Grbner basis command is executed by writing In[1]:=GroebnerBasis[{polynomials},
{variables}, {options}] (where In[1]:= is the mathematica prompt) which computes the Grbner basis for the ideal
generated by the polynomials with respect to the monomial order specified by monomial order options with the variables specified as in the executable command giving the reduced Groebner basis. Without specifying the options part,
one gets too many elements of the Grbner basis which may not be relevant. In Maple Version 5 the command is
accessed by typing > with (grobner); (where > is the Maple prompt and the semicolon ends the Maple command).
Once the Grbner basis package has been loaded, the execution command then becomes > gbasis (polynomials, variables, termorder) which computes the Grbner basis for the ideal generated by the polynomials with respect to the
monomial ordering specified by termorder and variables in the executable command. Following suggestions from B.
Buchberger (1999), we Mathematica software is adopted in the present study.
2-322
Closed form solution of nonlinear equations by Multipolynomial Resultants
Whereas the resultant of two polynomials is well known and algorithms for computing it are well incorporated into
computer algebra packages such as Maple, the Multipolynomial resultant, i.e. the resultant of more than two polynomials still remain an active area of research. In Geodesy, the use of the two polynomial resultant also known as
20
the Sylvester resultant is exemplified in the work of P. Lohse (1994, pp.72-76). The present study therefore extends
on the use of Sylvester resultants to resultants of more than two polynomials of multiple variables (Multipolynomial
resultant) and illustrates in Chapter 3 how the tool can be exploited in Geodesy to solve nonlinear system of equations.
The necessity of Multipolynomial resultant method in Geodesy is due to the fact that many geodetic problems involve
the solution of more than two polynomials of multiple variables. This is true since we are living in a threedimensional
world. We shall therefore understand the term multipolynomial resultants to mean resultants of more than two polynomials. We treat it as a tool besides the Grbner bases and perhaps more powerful to eliminate variables in solution of
polynomial systems. Publications on the subject can be found in the works of G. Salmon (1876), F. Macaulay (1902,
1916), A. L. Dixon (1908), C. Bajaj et al. (1988), J. F. Canny (1988), J. F. Canny et al. (1989), I. M. Gelfand et al.
(1990, 1994), D. Manocha (1992, 1993, 1994a,b,c), D. Manocha and J.F. Canny (1991, 1992, 1993), G. Lyubeznik
(1995), S. Krishna and D. Manocha (1995), J. Guckenheimer et al.(1997), B. Sturmfels (1994, 1998) and E. Cattani
et al. (1998). Text books on the subject have been written by I. Gelfand et al. (1994) and, more recently, by D. Cox et
al. (1998, pp.71-122) who provides interesting material.
In order to understand the Multipolynomial resultants technique, we first present the simple case; the resultant of two
polynomial also known as the Sylvester resultant.
Resultant of two polynomials
Definition 2-10 (Homogeneous polynomial): If monomials of a polynomial
same total degree, the polynomial is said to be homogeneous.
with non zero coefficients have the
Example 2-8 (Homogeneous polynomial equation):

A homogeneous polynomial equation of total degree 2 is
all have the sum of their powers (total degree) being 2.
{2]#a^b%#]%a^#]%b%#a/b;|
*i]
ra
b
r]%ar]%b+ra/b
=oa' ]
To set the ball rolling, we examine next the resultant of two univariate polynomials `
since the monomials
of positive degree as
*z W ] ! 0302030lO!#; W *i
@ ;9)&e
(2-53)
`*z W c
] bz0303020d b 9 W *z
@ ;"eed
?kDF)1Z determinant
the resultant of and ` , denoted Res @ `_( is the F)1Z
W cY
0 0 0 !
_W Y
0 0 0 !
W Y
0 0 0 !
W Y
0 0 0 O!
W Y
0 0 0 O!
~L}C `_&* W Y 0 0 0 W Y
0 0 0 O !
(2-54)
W
Y 0 0 b0
W
Y 0 0 b0
W
Y
0 0 0b b
W Y
0 0 0 b
W Y
0 0 0 b
where the coefficients of the first polynomial of (2-53) occupies rows while those of the second polynomial `
occupies ) rows. The empty spaces are occupied by zeros as shown above such that a square matrix is obtained. This
resultant is also known as the Sylvester resultant and has the following properties (B. Sturmfels 1998, D. Cox et al.
1998, 3.5)
@ _ is a polynomial in W020302$ ! ; WO030203 b with integer coefficients

2. Res @ `_}* if and only if F]^ and `/] have a common factor in fg' ] )
3. There exist a polynomial Lo fM' ] ) such that Pl`*~LOC`_
1. Res `
Sylvester resultants can be used to solve two polynomial equations as shown in the example below
21
Example 2-9 (D. Cox et al.1998, p.72):

Consider the two equations
>@*]%aBm,*z
=>@*i]
Sa
Bh*90id
(2-55)
In order to eliminate one variable, we use the hide variable technique i.e. we consider one variable say
(of degree zero). We then have the Sylvester resultant from (2-54) as
as a constant
a B , \
B, *ia BOOa

z,
~L}C`]}* a
, a
Bh
(2-56)
which can be readily solved for the variable and substituted back in any of the equations in (2-55) to get the values
of the other variable . For two polynomials, the construction of resultant is relatively simpler and algorithms for the
execution are incorporated in computer algebra algorithms. For the resultant of more than 2 polynomials of multiple
variables, we turn to the Multipolynomial resultants.
Multipolynomial Resultants
In defining the term multipolynomial resultant, D. Manocha (1994c) writes:
Elimination theory, a branch of classical algebraic geometry, deals with conditions for common solutions of a system
of polynomial equations. Its main result is the construction of a single resultant polynomial of homogeneous
polynomial equations in unknowns, such that the vanishing of the resultant is a necessary and sufficient condition
for the given system to have a non-trivial solution. We refer to this resultant as the multipolynomial resultant and use
it in the algorithm presented in the paper.
We present here two approaches for the formation of the design matrix whose determinant we need; first the approach
based on F. Macaulay (1902) formulation and then a more modern approach based on B. Sturmfels (1998) formulation.
Approach based on F. Macaulay (1902) formulation: With polynomials, the construction of the matrix whose
entries are the coefficients of the polynomials
can be done in five steps as illustrated in the following
approach of F. Macaulay (1902) :
OY1030203
Y*9020303 *t
are considered to be homogeneous equations in the variables

Step 1: The given polynomials
and if not, they are homogenized. Let the degree of the polynomial be . The first step involves
the determination of the critical degree given by C. Bajaj et al. (1988) as
]^YC030302#]
*-,
,
!Bm,N0
(2-57)
Step 2: Once the critical degree has been established, the given monomials of the polynomial equations are multiplied with each other to generate a set whose elements consists of monomials whose total degree equals the
critical degree. Thus if we are given the polynomial equations
, then each monomial of
is multiplied by those of
, those of are multiplied by those of
until those of
are
multiplied by those of . The set of monomials generated in this form is
030203$

Y *;1030302$ *
[ 030203
Oj*Q{2]kj * Y l

0_0O0S |
] i
* ]mlY 30 0_0] ln
and the variable j
Y
(2-58)
Step 3: The set

containing the monomials each of total degree
criteria (J. F. Canny 1988, p.54)
is now partitioned according to the following
22
*
Yj *
0
j 0
0
0
0
0
j * 2{ ] l o
2{ ] l
o j L +Y(K Y|
{] l o j _
K
c Y Y|
0
0
0
j L K c ! !%^)+*-,O030203#Bm,_|O0
The resulting sets of ! j are disjoint and every element of j is contained in exactly one of them.
Step 4: From the resulting subsets
defined as follows
!j
2
j
, a set of polynomials <
<
(2-59)
which are homogeneous in
! * ] j! j o !
!
variables are
(2-60)

! ! Y T

l
B
,

l
B
,
is of the order
k
and is such that for a given polynomial <! in (2-60), the row is
made up of the symbolic coefficients of each polynomial. The square matrix s has a special property that the non
trivial solution of the homogeneous equations <! which also form the solution of the original equations _! are in its

null space. This implies that the matrix must be singular or its determinant, 'sx , must be zero. For the determinant
to vanish therefore, the original equations L! and their homogenized counterparts <! must have the same non trivial
from which a square matrix is now formed with the row elements being the coefficients of the monomials of the
q>p pp
polynomials <
and the columns correspond to the monomials of the set j . The formed square matrix
solutions.
Step 5: After computing the determinant of the square matrix above, F. Macaulay (1902) suggests the computation
of extraneous factor in order to obtain the resultant. D. Cox et al. (1998, Proposition 4.6, p.99) explains the ex%< r
traneous factors to be integer polynomials in the coefficients of < r
, where < r <
and is related to the determinant via
W020#03 Y
! * ! ];Wc1030203#] Y_c
)Uc *z~RLc< Y C030302< (0 y]
(2-61)
with the determinant computed as in step 4, ~Lc< Y 020303 < being the Multipolynomial resultant and y] the
extraneous factor. This expression was established as early as 1902 by F. Macaulay (1902) and this procedure
of resultant formulation named after him. F. Macaulay (1902) determines the extraneous factor from the submatrix of the
square matrix and calls it a factor of minor obtained by deleting rows and columns of
the
matrix . A monomial l of total degree is said to be reduced if j o divides l for exactly one
. The extraneous factor is obtained by computing the determinant of the sub-matrix of the coefficient matrix
obtained by deleting rows and columns corresponding to reduced monomials l
kr
]!
] 0
From the relationship of step 5, it suffices for our purpose to solve for the unknown variable hidden in the coefficients
of the polynomials by obtaining the determinant of the
square matrix and equating it to zero neglecting
the extraneous factor. This is because the extraneous factor is an integer polynomial and as such not related to the
variable in the determinant of . The existence of the non-trivial solution provides the necessary and sufficient
conditions for the vanishing of the determinant. The procedure becomes clear when we consider the example of
the threedimensional resection problem in Chapter 3. It should be pointed out that there exists several procedures
for computing the resultants as exemplified in the works of D. Manocha (1992, 1993, 1994a,b,c) who solves the
Multipolynomial resultants using the eigenvalue-eigenvector approach, J. F. Canny (1988) who solves the resultant
using the characteristic polynomial approach and B. Sturmfels (1994, 1998) who proposes a more compact approach
for solving the resultants of a ternary quadric using the Jacobian determinant approach which we present below.
Approach based on B. Sturmfels (1998, p.26) formulation: Given three homogeneous equations of degree two as
follows
<
<
<
Y >@*i
[ >@>@*i
*i
Y Y ]
[YY ]]
Y
a
$[
aa
#Y [ b
[$[[ bb
TY \ ]%ad
[\\ ]%]%aad
d
Y 2]%b d
Y 3a/b *
2]%b d
3a/b *
[ 2]%b d
[ 3a/b *
(2-62)
23
we compute the Jacobian determinants of (2-62) by
Y
a
b
]
t < t < t <
* t ]
t a
t b

(2-63)
t <[ t <[ t <[
t
] ta tb
s
which is a cubic polynomial in the coefficients {2]/a^;b9| . Since the determinant polynomial in (2-63) is a cubic
polynomial, its partial derivatives will be quadratic polynomials in variables {2]Ca^b;| and can be written in the form
<
<
<
t s
Y
a
d
YT[ b
l
YT\ ];al Y 3]%b l
Y Na/b *
] >@* YY ]
l
s

\ ];al
3]%b l

Na/b *
a >@*
Y ]
l
$
a
d
[ b
l
t
t
t
(2-64)
t s
b >@*N[Y(]
l([
a
dN[[2b
l([\3];al([ ]%b l([ a/b * 90
The coefficients ( ! in (2-64) are quadratic polynomials in ! of equation (2-62). The final step in computing the
b
b
resultant of the initial system (2-62) involves the computation of the determinant of a :vk: matrix given by
Y$Y Y
YT[ YT\ Y Y

Y
$

[
\

(2-65)
~RL
$

] < Y <
< [ }* 3Y$[(YY 2Y[
3Y#[[[ 2Y[\\ 3Y[ 3Y[ 0

Y

[
\

([(Y N[
([$[ N[#\ ([ [
The resultant (2-65) vanishes if and only if (2-62) have a common solution {]C/a^;b;|O where {2]9a^;b9| are complex
t
numbers or real numbers not all equal zero. In Chapter 3, we use the procedure to solve the Grunert equation and the
GPS pseudo-range equations.
2-33 Adjustment of the combinatorial minimal subsets solutions

Once the combinatorial minimal subsets have been solved using either the Grbner bases or the Multipolynomial
resultant approach, the resulting sets of solution are considered as pseudo-observations. For each combinatorial, the
obtained minimal subset solutions considered as pseudo-observations are used as the approximate values to generate
the dispersion matrix via the nonlinear error propagation law/variance-covariance propagation (e.g. E. Grafarend and
B. Schaffrin, 1993, pp.469-471) as follows:
From the nonlinear geodetic observation equations that have been converted into its algebraic (polynomial) form, the
) with
combinatorial minimal subsets will consist of polynomials
being
a'
the unknown variables (fixed parameters) to be determined and
the known variables comprising the
observations or pseudo-observations. We write the polynomials as
Y @> *D41F] Y 020302#] j

@> *D41F] Y 020302#] j
0
0
0
j >*=41F] Y 030203] j
which is expressed in matrix form as
Y C030302$ j o ] Y 1030203] j
{a Y 020302#a |
#a Y 020303a }*i
#a Y 020303a }*i
#a Y 030203#a &*
>@*Yv}w?n&*
{2] Y 020303] j |
(2-66)
(2-67)
24
{2]^YL030203] j |
n
{]^YC030302#] j |
"y { | *
yv{n&|R*
r{ |*8w
r{2n}|*w
{2acY030203#a |
{2aY1030203a |
where the unknown variables

are placed in a vector w and the known variables
are placed
in the vector . Next, we implement the error propagation from the observations (pseudo-observations)
to the parameters
that are to be explicitly determined namely characterized by the first moments, the
x , as well as the second moments, the variance-covariance matrices/dispersion
expectation w xAy and
y and
matrices w
. From E. Grafarend and B. Schaffrin (1993, pp.470-471), we have upto
nonlinear terms
w
=z y{z
Rz
]z y
(2-68)
r{ |R* Y w= `
Y
with z y z being the partial derivatives of (2-67) with respect to w ,n respectively at the Taylor points xAy9 x . The
approximate values of unknown parameters {]^Y_1030302#] j |Po w appearing in the Jacobi matrices z y z are obtained
from Grbner bases or Multipolynomial resultants solution of the nonlinear system of equations (2-66).
Y
Y
Given zZ!*ez y z from the ) combination and z
|z y }~zZ } from the combinatorials, the correlation
*
b
o
o
between the ) and combination is given by
w !b
* z b w } o z !
(2-69)
wvY(w
w[O1030302w
i (where is the
The sub-matrices variance-covariance matrix for the individual combinatorials
number of combinations ) obtained via (2-68) and the correlations between combinatorials obtained from (2-69) form
the variance-covariance/dispersion matrix
w Y w Y
0
w
Y w
0
0
w [
wp*
0
0
w 3Y 0 0
0 0 w Y
0 0 w

0
0
w
(2-70)
for the entire combinations. The obtained dispersion matrix is then used in the linear Gauss-Markov model (2-4) to
obtain the estimates of the unknown parameters with the combinatorial solution points in a polyhedron considered
as pseudo-observations in the vector of observations while the design matrix comprises of integer values 1 being
the coefficients of the unknowns as in (2-73). In order to understand the adjustment process, we consider the following
example.
Example 2-10: Consider that one has four observations in three unknowns in a nonlinear problem. Let the observations be given by
leading to four combinations giving the solutions and the a
,
,
and
. If the solutions are placed in a vector
) the adjustment model is then defined as

'
{2aY#a #a[#a_\|
cFacYL#a
#a[
F a
#a [ #a \ a Y a
[ a \ a Y a
a \
#
"*
(2-71)
yv{^|R*i Y
u [ q [ u Y $r{a^| c)UcVN% cc)OV3;%4 )+0
Let

(2-72)
q * mZ >cV m>@* o Y
u Y

such that the postulations Lr{2q |*

)U i.e. best and y"{3q |*8q^q o j i.e. uniformly
unbiased holds. We then have from (2-70), (2-71) and (2-72) the result
q *` [ u Y
w 1Y
u [2# [ u Y
w Y
*zO 4{ Lr{2q |R* +Rw * U) yv|
{ ^q |
(2-73)
The dispersion matrix

of the estimates is obtained via (2-5). The shift from arithmetic weighted mean to the
use of linear Gauss Markov model is necessitated as we do not readily have the weights of the minimal combinatorial
subsets but their dispersion which we obtain via error propagation/Variance-Covariance propagation. If we employ
the equivalence theorem of E. Grafarend and B. Schaffrin (1993, pp. 339-341 ), an adjustment using linear Gauss
markov model instead of weighted arithmetic mean in Box (2-1) is permissible. In Appendix A.3 in page (108), we
present the error propagation based on the nonlinear random effect model (multivariate) where we illustrate the effect
of the biased term.
25
Example 2-10 (error propagation for planar ranging problem):

Let us consider a simple case of the planar ranging problem. From an unknown point
, let distances
and be measured to two known points
and
respectively. We have the

distance equations expressed as
"
"For
Y Y Y o

F
$
Xo
*Q Y B
z' Y BP
*Q
B
z'

B P
(2-74)
which we express in algebraic form (2-66) as
OY > *`FxYB

1Y?BD=
[
B Y

* 0

> *`F
B

BD=
[
B

*
"
(2-75)

Y > *.9F Y B Y B.9 F Y B d.; Y BD= Y B
B .9' Y BD -B. Y Y *z
(2-76)

> *.9F
B
B.9 F
B d.;
BD=
B
B.9'
D
B -B.

*z
which on arranging the differential vector of the unknown terms {2$v|v*{2] Y ] |xo5w on the left hand side
and that of the known terms {2 Y $ Y # $ Y |*f{a Y #a #a [ #a \ #aO_#aO|on on the right hand side leads
to

CY
xY
"

|
z y
* z Y
(2-77)

On taking total differential of (2-75) we have
z y
tY
* t
"
t
t
t
with
Y
*
t Y
t
CY
X*
z
t
t
"
B.9 Y B B .; Y BP

B.9

B B .;

B Pg
and
Y
Y
t
t
Y
1Y
t
t
t
t
(2-78)
(2-79)
.CY B.;FxYB B.9'^YBD=
.
B.9F
B B.9'
BDE

If we consider that
{w|R*w y *
g

g
r{2n&|R*w * gg
g
g
and
g
g
"

g
g

(2-80)
g
g
g
g
g

g
g

(2-81)
{2v| *Q{2] Y #]

|o
we obtain with (2-77), (2-78) and (2-79) the dispersion (2-68) of the unknown variables
6w .
The unknown values of
6w appearing in the Jacobi matrices (2-78) and (2-79) are obtained
from Grbner bases or Multipolynomial resultants solution of the nonlinear system of equations (2-74).
{v|R*Q{2] Y ]
|o
26
2-4
Concluding remarks
The Chapter has presented the closed form algebraic tools of Grbner Bases and the Multipolynomial resultants that
are used in the next Chapters to solve closed form geodetic problems. Using these algebraic closed form tools, the
Gauss-Jacobi combinatorial algorithm is presented as an alternative to estimate the fixed parameters in the nonlinear
Gauss-Markov model. In Chapter 3, we consider the solution of selected geodetic problems using the algebraic tools of
Grbner Bases and the Multipolynomial resultants while in Chapter 4, we use the Gauss-Jacobi combinatorial algorithm to estimate the fixed parameters in the nonlinear Gauss-Markov model in the overdetermined three-dimensional
resection case.
Chapter 3
Solution of selected Geodetic problems

In this Chapter, we solve in a closed form the geodetic problems of; three-dimensional resection, minimum distance
mapping and GPS point positioning with observations of type pseudo-range. The Grbner bases and the multipolynomial resultants techniques presented in Chapter 2 are used as the computational tools with the Gauss-Jacobi
combinatorial algorithm employed in the solution of the overdetermine GPS code pseudo-ranging problem. We start
in Section (3-1) by giving an overview of the GPS and LPS systems before considering the solution of the selected
geodetic problems in Section (3-2).
3-1
LPS and GPS positioning systems
In this section we highlight the main positioning systems; the Global Positioning System GPS and the Local Positioning
System LPS and in particular, the reference frames upon which they operate. An exposition on their relationship is
given and the term three-dimensional orientation problem defined.
3-11 GPS Positioning
{ $|
of
The results of the three-dimensional positioning by GPS are the three-dimensional geodetic coordinates R
a point. When positioning with GPS, the outcome is the geocentric position for an individual receiver or the relative
positions between co-observing receivers. The mode of operation and techniques are well-documented in GPS related
publications and books (e.g. B. Hofman-Wellenhof et al. (1994), A. Leick (1995), V. S. Schwarze (1995), G. Strang
and K. Borre (1997), A. Mathes (1998), R. Dach (2000) and M. S. Grewal et al. (2001) among others).

Y #
[
q
The Global Reference Frame
upon which the GPS observations are based is defined by the base vectors
with the origin being the center of mass. The fundamental vector is defined by the base vector q and coincides with
the mean axis of rotation of the Earth and points to the direction of the CIO. is oriented such that the plane
completes the right handed sysformed by and q points to the direction of the Greenwich in England.
tem by being perpendicular to and . The Geocentric Cartesian coordinates of a positional vector is given by
Y
R[
Y
[
Y

R[
h *iY
S[
where { } are the components of the vector h in the system Y # # [q
(3-1)
.
3-12 LPS positioning

3-121
Local Positioning Systems (LPS)
E. W. Grafarend (1991) defines the Local Level System as a three-dimensional reference frame at the hand of an
experimenter in an engineering network. When one is positioning using an electronic theodolite or a total station, one
first centers the instrument. When the instrument is properly centered and ready for operation, the vertical axis of the
instrument at this moment coincides with the direction of the local gravity vector at that particular point, hence the term
direction of the local gravity vector, and points in the direction opposite to that of the gravity vector at the theodolite
27
CHAPTER 3. SOLUTION OF SELECTED GEODETIC PROBLEMS
28
6!
station (i.e. to the zenith). The system can now be used to measure observations of the type horizontal directions
, vertical directions
and the spatial distances with the triplet
being measured in the Local Level
Reference Frame. These systems as opposed to GPS are only used within the local networks and are referred to as
the Local Positioning Systems (LPS). In Section (3-122) below, we consider the choice of the local datum for such
systems in a three-dimensional network.
7 !
{ ! #6 ! $7 ! |
3-122 Local datum choice in an LPS 3-D network

In positioning with the Local Positioning System (LPS), one has two options, namely; using a known orientation
or defining the orientation arbitrarily. When the first option is adopted, in which case the azimuths are known, one
operates in the Local Level Reference Frame of type
discussed in (a) below. Should the second approach be opted
for, then one operates in the Local Level Reference Frame of type
discussed in (b).
(a)
Local Level Reference Frame of type

The origin of the
system is a point
5
5
whose coordinates are

h * *

Y
[
Y
(3-2)
defined by base vectors , , q of type south, east, vertical. points to the direction opposite to
points south, while
completes the system by pointing
that of the local gravity vector at point .
east. The datum spherical coordinates of the direction point in the Local Level Reference Frame
are
(b)
!
kB F
! kB F
5Z *P5 *R5Z*
5
VN_~!V3_27R!
(3-3)
P5 _ *
! 2)UU!_VN_7R!
R5
3)U7R!
with azimuths ! , vertical directions 7 ! , and spatial distances ! .
Local Level Reference Frame of type 5
Defined by the base vectors Y , , Rq[ , with Y within the local horizontal plane spanned by the
base vectors Y and directed from to ! in vacuo. The angle between the base vectors Y and
Y is the "unknown orientation parameter" in the horizontal plane. Y ,

, [ is related to Y
, ,[ by a "Karussel-Transformation" as follows
Y * Y - #(H

*-BZ Y H DS

-
(3-4)
[ * q[
or

-
(
B6#H
-

Y #
[q ) *' Y
[ ) H
(3-5)
thus the Local Level Reference Frame of type 5 is related to the Local Level Reference Frame of type
5 by
' Y _
* ' Y L
#[ )~g[ ?(0
(3-6)
[)
The datum spherical coordinates of point ! in the Local Level Reference Frame 5 are given as
F 5X *5X *AR5X*z
/6 ! 97 !
5
(3-7)
! F 5 J
* ! #HZ61! 97 !
R5
HX7R!
'
where
6!
and
7 !
are the horizontal and vertical directions respectively, while
are the spatial distances.
3-1. LPS AND GPS POSITIONING SYSTEMS
29
The local cartesian coordinates of a point whose positional vector is in the
5
system is given by
*i Y ]PS
a S [q b
(3-8)
]Ca^b ) are the components of the vector in the system {2 Y #
# [ |&0 More insight on the topic can be found
in B. Richter (1986, p.28) and E. Grafarend and B. Richter (1977).
3-13 Relationship between Global and Local Level Reference Frames

In positioning within the LPS threedimensional positioning framework , one is interested not only on the geometrical
quantities
being the direcG for position of the new point but also on the physical quantities q_
tion of the local gravity vector at the positioned point. When positioning with LPS within the Global Reference
Frame, the direction q_ of the local gravity vector is obtained by solving the threedimensional orientation
problem. This is achieved by the transformation of coordinates from the Local Level Reference Frame to the Global
Terrestrial Reference Frame (e.g. ITRF97). It is conventionally solved by a means of a Q rotation matrix, which is
represented by a triplet q of orientation parameters called the astronomical longitude , astronomical
latitude , and the "orientation unknown" in the horizontal plane. With respect to the local gravity vector the
triplets
are its spherical coordinates, in particular q its direction parameters. The threedimensional orientation problem therefore solves the problem of determining (i) the Q rotation matrix and (ii) the
triplet _ _ q_ of orientation parameters from GPS/LPS measurements. As soon as the astronomical longitude
_ and astronomical latitude _ are determined - no astronomical observations are needed anymore - the vertical
deflections with respect to a well-chosen reference frame, e.g. the ellipsoidal normal vector field can be obtained (e.g.
E. Grafarend and J. L. Awange (2000)). The three-dimensional orientation problem is formulated by relating the
frame to the
frame as follows
{2|
{ 1 ^|
{ 1 ^|
{
{ *` R|
{ 1 1 ^|
5

'
Y #
_#[q ) *' Y _#
#[q )E q
(3-9)
where the Euler rotation matrix is parameterised by
[ _;

BH_ [ ];
(3-10)
i.e. the threedimensional orientation parameters, the astronomical longitude astronomical latitude and the
"orientation unknown" in the horizontal plane. In-terms of Cartesian co-ordinates ]Ca^b of the station point, the
target points ]%!#aO!Tb! in the Local Level Reference Frame @5 and Cartesian co-ordinates $}( of the station point,
the target points "!#$%!#+! in the Local Level Reference Frame one writes
]%!1B]
v!B
a!1Ba J * ] 1q _1 _; %!1BD _
(3-11)
b!^Bb
+!B
]1q_1_;>@*
with
]!1B]
3V _61!_VN_7R!
a ! Ba J *1! 2) 6 ! VN_7 ! >^!o{O,./^030203|
b ! BDb
2 )7 !
The question then arises: How can the parameters relating the Local Level Reference Frame
Frame
be obtained?

(3-12)
5
to the Global Reference

1.
The conventional approach has been to determine the direction of the local gravity vector at
the origin of the network and the orientation unknown in the horizontal plane by the use of steller
astronomical observations.
2.
E. W. Grafarend, P. Lohse and B. Schaffrin (1989) have proposed an approach based on the solution
of three-dimensional resection problem. In the approach, directional measurements are performed to
the neighbouring 3 points in the Global Reference Frame and used to derive the distances by solving the Grunert equations. From these derived distances, a closed form solution of the six unknowns
q
by means of the Hamilton-Quaternion procedure is achieved.
{2X
|
30
3.
J. L. Awange (1999) and E. Grafarend and J. L. Awange (2000) solved the overdetermined form of the
problem by using the simple Procrustes algorithm to determine the three-dimensional orientation parameters q and the deflection of the vertical for a point whose geometrical positional quantities
G are known.
{
{3|
4.
|
{2|
The fourth approach would be first to determine the geometrical values

G of the unknown
point through the three-dimensional resection approach presented in the present study and then back
substitute the obtained geometrical values
in (3-11) to obtain the Euler rotation matrix. The
Euler rotation angles can then be deduced via an inverse map presented by E. Grafarend and J. L. Awange
(2000, lemma 2.3, p.286).
{|
Here the three-dimensional orientation problem is understood as the fundamental problem to determine the rotation
matrix as well as its three parameters called {astronomical longitude, astronomical latitude, horizontal orientation
unknown in the horizontal plane} which relates the Local Level Reference Frame to the Global Terrestrial Reference
Frame (e.g. ITRF 97) from GPS position measurements both at target points as well as the station point and LPS
direction measurements (horizontal directions, vertical directions by a theodolite) from the station point to the target
points (at least three). If a reference direction parameterized in terms of surface normal {ellipsoidal longitude ,
ellipsoidal latitude } is subtracted from the local vertical parameterized in terms of {astronomical longitude ,
astronomical latitude }, namely [ - H we have access to the vertical deflections.
B |
3-14 Observation Equations

In this section, we consider the equations with respect to a stationary theodolite for the purpose of point positioning.
For the case where the theodolite moves from point to point, i.e. moving horizontal triad, we refer to E. Grafarend
(1975, 1988 and 1991). In general a more elaborate literature on observations in three-dimensional positioning is
given by E. Grafarend (1981). Stationed at the point
and with the theodolite properly centred, we sight the
target points,
where = 1,2,3,.....,n. There exist three types of measurements that will be taken from
to
in the LPS system (Local Level Reference Frame
). These are; the horizontal directions
, vertical
directions
and spatial distances whose equations are given as
!+o [
[
!oS
7R!
to [
1!
< a!
6 ! *zJI K$J_H <]%!
5
61!
BH
po [
(3-13)
<b !
(3-14)
<]
! l<a !m

! * <]
! l<=a !
l<Pb !

(3-15)
where <]%!* F]%!}B](`<a!=* Fa!+Ba9(Q<Pb!=*b!}Bmb denote the coordinate difference in the Local Level
Reference Frame 5 and _ is the "unknown orientation" in the horizontal plane after setting the zero reading of
the theodolite in the direction F ! . The relationship between the Local Level Reference Frame 5 and the Global
Reference Frame is then given by
<]!
<="!
<a! * _1 _1$O <!
(3-16)
<Pb !
<" !
7R!C*zJI K$J_HZ
1 1$O&*
_ _
with
H
_
6 _

-_

B H
H
#H
#H
_X _
_
-_X _
_
#H
(3-17)
(_
we now express the observations (3-13), (3-14) and (3-15) in the Global Reference Frame as
B6 HX_;<="!% _;<!

(3-18)
< ! H
(H H <P ! B - <" ! d B\_
! - H <P ! H H( <" !
7 ! *iJ_I KJ_H* H( k < <! H
(3-19)
H H <P ! B - <" !
S
0
with
* < ! BD H < !

, < ! * F ! BmN%<P ! * ' ! Bi(<v ! * ! Bz in the Global
Reference Frame and { '( '(q | are the three orientation unknowns at the unknown point
61!C*zJI K$J_H # H

3-2. SELECTED GEODETIC PROBLEMS
3-2
31
Selected geodetic problems
In this section, we consider the application of the techniques discussed in Chapter 2 to solve;
(1)
the three-dimension resection problem
(2)
the minimum distance mapping problem
(3)
the four-point GPS pseudo-ranging equations.
3-21 Threedimensional resection problem

Here we consider the closed form solution of the threedimensional resection problem using Grbner basis and Multipolynomial resultant techniques. The closed form threedimensional resection procedure is carried out in three steps
namely: distance-derivation step (solution of the Grunert equations), the position-derivation step and the orientationderivation step. One may argue that the distance derivation step is irrelevant in light of the modern distance measuring
equipments such as EDM. We should however not forget that in forest areas, one may be lucky to measure angles
but the blockage of EDM signals by tree leaves and branches may hamper accurate distance measurements. In photogrammetry, the distances have to be derived from the image coordinates in order to obtain the perspective center
coordinates and the orientation parameters, i.e. the elements of exterior orientation. These two examples validate the
necessity of still having procedures for deriving the spatial distances despite the existence of Electromagnetic Distance
Measuring (EDM) equipments.
In the closed form threedimensional resection problem, we are interested in determining the position and orientation
and vertical directions
to three
of the point connected by angular observations of type horizontal directions
other known points
as in Fig (3.1) below. From the angular measurements, the distances are derived in the
distance derivation step by solving the Grunert equations. Once the distances have been established, the unknown
position
is determined in the position derivation step. The closed form threedimensional resection problem is
completed in the orientation derivation step by solving the orientation parameters that relate the Global Reference
System
to the Local Level Reference System of type .

6!
+YL$
[
7 !
5
!
a
Figure 3.1: Tetrahedron for closed form 3d-resection
(a)
The distance-derivation step (solution of the Grunert equations):

We begin in Box (3-1) by presenting the derivation of the distance equations (also known the Grunert
equations) relating the known distances
!
X\
@ between the known GPS stations in
b
the Global Reference Frame
the unknown distances
between the unknown station
and the known stations
and the spatial angles

\
@ The spatial
b
angles
and

\
@ are obtained from observations of type horizontal directions
b
vertical directions
as shown in Box (3-1) in the Local Level Reference Frame
The relationship
between observations in the Local Level Reference Frame and the Global Reference Frame
in Box (31) is presented in (3-22) where
are GPS Cartesian coordinates of known points

are spherical coordinates of type: spatial distances,
1! ) l* , $.9 ) *
1!)v* , ./
o [
!=om [
!
) D*,O$.9 ) * O0
+! r) -* ,./$ ) *
7R!
P0 5

c
Z
)
S
o
c
{

,

/
.

9
|
!
!
!
! o [ )=o-{O,O$.9/|c ! #6 ! $7 ! 1)PoA{O,./$/|
P
61!
32

fo [
horizontal and vertical directions respectively linking the old and new GPS points, while
are the
required GPS coordinates of the unknown point
(Figure 3.1). is the rotation matrix containing
the three-dimensional orientation parameters (see (3-11) in Section (3-13)). Multiplying (3-22) by ( 3-24)
leading to (3-25) derive the relationship between the spherical coordinates of type horizontal directions ,
vertical directions and the space angles . After manipulations of (3-26),(3-27) and (3-28), the spatial
b
angle
can be written in terms of the spherical coordinates
of points and
with
b
b
b
b
respect to a theodolite orthogonal Euclidean frame
as in equation (3-29). The Grunert equations for the
three unknown distances can now be written in terms of the known distances
and

the space angles
illustrated in (Figure 3.1) as in (3-31). In photogrammetry, the relationship

between the space angles and the measured image coordinates with respect to orthogonal Euclidean frame
centred at the perspective centre is given by E. Grafarend and J. Shan (1997a, equation (1.1), p.218) as
+!
7R!
6!
5
Y [
Y

[ [ Y
c
!b *
{361!T7R!#|&N{26 7 |
!
Y

[ [(Y
] !] b
a!'a b d

]
! Sa !
d
] b
a b
l
(3-20)
Box 3-1 (derivation of the Grunert equations):
GPS and LPS coordinate systems"
]%!B]
"!1B
a!^BDb Ja * !1BD _
b ! BDb
! B
/61! 97 !
v!B
1! #HZ61! 97 !
* %!1BD
HX7R!
&!1B
u

[
[
o Wv'O>@*-{h
o
Lh=h * [ h`*A,_|
#B.m ' /6 b 97 b H6 b 97 b HZ7 b ) b
V3_361!'VN_7 !
TB.O' VN_26 b VN_7 b $2)U16 b V3_27 b $2)U7 b ) 1!! b 2)U161!FVN_7 ! *
2)U7 !
! B
TB.V ' b B( b BDN2U b B ) ! B
! BS
(3-21)
(3-22)
(3-23)
(3-24)
(3-25)
33
Box 3-1 (derivation of the Grunert equations continued):
b
B 3F ! B+*i b ! B b B !

! B b 3F ! B b &*z !
B.L ! b b
F"!B3F"!^B&*i !
B.L"!S
F b B3F b B&*i b

B.L b S
F ! B b
BF ! B
Bm b B
*`B.9 b BN ! B
VN_261!VN_7R!
)
3)U161!FVN_7R! *
#

B

.

'
N
V
_
6
N
V
_
7

2
)
1

6
N
V
_
7
$

2
U
)

7
1

]
!

b
b
b
b
b
b
2)7R!
F"!1B b
!BD b
&!1B b
B
BF ! B
Bm' ! B=
B ! B
B
B b B
B b B
B b BS
B.V{ HX7 #HZ7 !% 97 97R! 2F6 B61!'N~| 1]! *

F ! B b
z ! B b
z ! BS b
B b
b
=
B"!1B

B!1BDb
BmU+!1BSb
B (3-28)
B b B
Bm' b BD
BmU b B
c ! * 97 ! ;7
36
B6 ! C\
#HX7 ! #HX7 b
b
b
b
"
B. / ! b ! b
* !
5
B !
5
B
!

*b !
Y
b
B. 1!b b c+! b b
Y

* Y
\

BD. CY
&Y

[
*

\
[
BD
. [ [
*
[
\
[(
Y
Y
BD.
[ aY
[(Y
(3-26)
(3-27)
(3-29)
(3-30)
(3-31)
CYL
1[
Several procedures have been put forward to solve the Grunert equations for the unknown distances in
(3-31) above. These include J. A. Grunert (1841), S. Finsterwalder and W. Scheufele (1937), E. L. Merritt (1949), M.
A. Fischler and R. C. Bolles (1981), S. Linnainmaa et al. (1988), E. Grafarend et al. (1989), P. Lohse (1990) and E.
Grafarend and J. Shan (1997a, 1997b). F. J. Mller (1925) and R. M. Haralick et al. (1991, 1994) present an extensive
review of these procedures. The present study considers the use of algebraic computational tools (Grbner bases
and Multipolynomial resultants) discussed in Chapter 2 to solve the Grunert equations for the unknown distances

in (3-31).
Y
[
In order to understand the usefullness of Grbner basis in the solution of the three-dimensional resection problem,
we present first the hand computation of the Grunert distances for a regular tetrahedron using Grbner basis before
considering the general case of the Grunert equations for distances in the general threedimensional resection problem
whose Grbner basis is computed using Mathematica software.
(i)
Grbner basis solution of the Grunert equations for a regular tetrahedron
We begin by expressing the Grunert distances equations (3-31) -whose geometrical behaviour has been studied by E.
Grafarend, P. Lohse and B. Schaffrin (1989)- in the form
cW z
* ]
Y S
]
BD./Y
]1Y]
( W *]
]
[

B .
[(]
]%[
VNW *i]
[
]
Y
B ._VN[(Y][3]1Y
(3-32)
where
]
BD. Y
] Y ]
S]
D
B W *z
]%
Y D
B
.O
[ ]
] [ S]%
[ BD W *
]
Y BD.VN[(Y$]^YN]%[S]
[ BV(W *i
(3-33)
E. Grafarend, P. Lohse and B. Schaffrin (1989) demonstrate that for each of the quadratic equation in (3-32), there
exists an elliptical cylinder in the planes
and
for the first, second and third equations
{2]^YL#]
&| C{2]
#][_|
{2][_]^YL|
34
{3]
o |
{]%[o |
] ! )+*Q,O$./$
]^Y*Q] *Q]%[
Y
*
[ * [
Y
{2]1YRo B &|
respectively . These cylinders are constrained to their first quadrant since the distances are positive thus
B
B
and
. In Box (3-2) below, we apply the Grbner basis technique to solve for the distances
between the unknown station
and the known stations
. For a regular tetrahedron, the
distances
joining the unknown point
to the known points
are equal to the distances

between the known stations. Let us consider these distances to be equal to U . The spatial
angles are also equal (i.e.
). The task now is to compute by hand the Grbner basis of

(3-32) and use them to find the Grunert distances for the regular tetrahedron (i.e. show that the desired solutions for
B
are
U
{2] Y #]
#] [ |o
or [
! or [ [
! o
o [
Y
*
[ * ([ Y *t:
] Y *z]
i
* ][ A
* 0
Box 3-2 (Hand computation of Grbner basis of the Grunert equations for a regular tetrahedron):
]1Ye`]
ep]%[ and subtracting the left hand side of (3-32) from the right hand
Y >@*z]
Y B._9Y ]^Y(] S]
BcW *z

>@*z]
B.
[(
]
][
]
[
BW *z
(3-34)
[ >@*z]
Y B._V [(Y ] Y ] [ d]
[ BDV W *

as polynomials in ' ] Y #] ] [ ) . These equations form (following Definition A-6 in Appendix A.1) the Ideal 1
as
1
* ' Y $
$ [ 2{' ] Y #]
] [ )
(3-35)
For a regular tetrahedron, where +! *:

__. 3:O 2&*Q, and W * W *zV W * . (3-34) is then written as
b
]
Y B] Y ]
]
B *z
]
B]
] [ ]
[ B *z
(3-36)
]
Y B] Y ] [ ]
[ B *z
Upon lexicographic ordering

side we have
giving rise from (3-35) to the Ideal 1 as
* ]
Y B] Y ]
S
]
B ]

B ]
] [ ]
[ B # ]
Y B] Y ] [ S]
[ B -
]1YL]
#]%[ )
1
2
'
whose generators ?
are written as
(3-37)
B ]^YN]
S]
Y
(3-38)
Y B]^YN]%[S]
[ B
B]
] [ S]
[ B 0
Desired now are the Grbner basis for the for the generators (3-38) of the Ideal 1 in (3-35). We then proceed
to compute the pair polynomials I4 Y >4 N2#4 Y 4 [ (I 4 > 4 [ from the generators (3-38) of (3-35). From B.
Buchbergers third criterion explained in Chapter 2, we notice that LM # 4 X*`]

Y divides the LCM I 4 Y 4 [ X*
consider only I 4 Y 4 N2I4 >4 [ .

]
Y ]
0 It suffices therefore to suppress the consideration of I 4 Y > 4 [ and instead
Z#4 Y >4 gives
(3-39)
Z#4 Y 4 &*`BZ] Y ]
S] Y ] [ S]
B]
[
which is reduced with respect to ? by subtracting 4 [ to obtain
BZ] Y ]
S] Y ] [ BD._]
[ S]
] [
(3-40)
which does not reduce to zero and is added in the original list ? of the generating set of the Ideal 1 as 4 \ .
The BLFa ) pairs to Tbe consideredT next are ZI 4 4 [ N Z# 4 > 4 \ ( ZI 4 [ > 4 \ from the new generating
set ?*t{ 4 > 4 [ > 4 \ | . Since

I 4 and
I 4 [ are relatively prime, ZI 4 > 4 [ reduces to zero modulo ?
( ZI 4 4 [ U
F\G ). The pair polynomials remaining for consideration are
I 4
4 \ and I 4 [ > 4 \ . Z# 4
> 4 \
gives
BD._] Y ]
[ S] ]
B]
(3-41)
ZI 4 > 4 \ *]
Y ] [ ] Y
which is reduced with respect to ? by subtracting ] [ 4 to give
] Y B] Y ]
[ S]
]
[ B]
B] [[ S] [
(3-42)
cY?*i]
4
*i]
4 [ *i]
4
35
Box 3-2 (Hand computation of Grbner basis of the Grunert equations continued):
which does not reduce to zero and is added to the list ? of the generating set of the Ideal 1 as
sets to be considered next is I4 >4
from the new generating set
4 . The
. #4 4
gives
?
V4 4 >4 >4

zBl%Lac )
Z [ \
Q*Q{
[ \ _| Z [ \
Z# 4[_>4_\&*QBZ]1Y]
[ ]1Y d._] ]
[ B]
]%[XB]
which is reduced with respect to ? by subtracting 4 and adding ] [ 4 [

.L] [[ B.L] [
(3-43)
to give
(3-44)
which is a univariate polynomial and completes the set of the reduced Grbner basis of the set ?
summarised as follows
4
*z]
Y B] Y ] [ S]
[ B
4 [ *z]
B] ] [ S]
[ B

S] Y ] [ B.L]
[ ] ] [
4 \ *`BZ] Y ]
?f>@*
4*z] Y B
] Y ]
[ S]
]
[ B]

B] [[ S] [
[
4*.L] [ B.L] [
(3-45)
[[ B._] [ is a univariate

*
L
.
]
. We then proceed to derive the solution to the
polynomial in ] [ and readily gives the values of ] [ *Z
B , the value of [ *U . This is substituted back
Grunert distance equations (3-34) as follows: Since [ *i] [ o
and 4 *i]
Y Bx] Y ] [ ]
[ B to give ] *^Z a and ] Y
* Z( a respectively.
in 4 [ *i]
B] ] [ r]
[ B
as we had
This complets the solution of Grunert equations (3-34) for the unknown distances ] Y *A] *A] [ *Q
initially assumed.
From the computed reduced Grbner basis in (3-45) we note that the element 4
(ii)
Grbner basis solution of the Grunert equations for the general three-dimension resection problem
We next present the application of Grbner basis technique to the solution of the Grunert equations (3-31) expressed
as follows:
c4 Y>@*i]
4
>@*i]
4R
[ >@*i]
]
l
._9Y
]^YN]
ScW
*
Y S
.
[N]
][lW *i
S]
[
l
._VN[(Y][3]1Y&dV(W *z;
[ S]
Y l
(3-46)
where
* ] Y o B *z] o
Y z
B. / Y
* Y

3B.
/
B Y

*z W 2B

[ * W 3B
B [ *] [ o B
[ 2B. / [(Y
* V [Y
[(
[ Y *z
*iV
W 0
(3-47)
We then have the Ideal formed from (3-46) as
1* ]
Y
]
l._9Y
]^YN]
dcW]
S]
[ l.
[3]
]%[lW]
[ S]
Y d.VN[Y(]%[3]^Y&dV(We
(3-48)
whose Grbner basis are computed in Mathematica 3.0 after Lexicographic ordering ]^YRem] ed]%[ by the command
GroebnerBasis as
?LL(2L72)T
'@{ 4cY 4
> 4[_|ON{]^YL#]
#][_| ) 0
(3-49)
1
The execution of the Mathematica 3.0 command above gives the computed Grbner basis of the Ideal (3-48) as
expressed in Boxes (3-3a) and (3-3b) below.
36
Box 3-3a (Computed Grbner basis for the Grunert distance equations-univariate term):
Y f
* T,:B3
Y
\Y
Bc
[ .[
Y

\ [ \
[ BO Y

[ V [Y ._ [Y

[ V [(Y ._ Y
[
[ V [(Y BOV
[(Y ._
Y
V
[(Y
.
[ V
[Y d
Y

[ V
[(Y l._ Y

[ V [(Y SV [Y ]-[
cW2
Y
\ c.WBd,:
Y
W}S. \Y
(W&i,2:O\ OW
[ BD._cW2
Y

[ BhcW2 [
[ d.
Y
(W
[ B.OW \
[
+( - 32a W+H
O.V([ WBr,2:O
Y
V(W1._ Y
V([ WBr,2:c
[ V(WÊc[
Y

[ V(W.
[ VNWÊOcW39Y
[
[NVN;,cW2 Y

[3VN[([ Y^Br,.9Y
(W
[3VN[(Y;
Y
W
[ V [(Y B W Y

[ V [Y S Y
W
[ V [(Y Bl,._ Y

[ V W V [(Y SO Y

[ V W V [(Y O Y

[ V W V [Y z,2:O \ W V
[Y B
._ W
Y
V
[(Y Bi,2:O W V
[(Y Y
c
Y
W V
[(Y H
B O W
[ V
[(Y . W
Y
[
[ V
[(Y .O W
[ [ V
[(Y m
Y
W [
[ [ V
[(Y m W
[ V
[([ Y B
V W V
[(Y ._
Y V W V
[(Y .
[ V W V
[(Y "
Y
[ V W V
[(Y B W Y [ V [Y "O Y W [ V [Y W Y [ V [(Y " Y [ V W V [Y B
._ W V \[Y l. W
V \[(Y S W

[ V \[(Y ] [

\
+( 24a W
B._ W

Y B OOWW}m,.OW2
Y (W&SR. c W
BS,
Y W
D Y W
BS,2O W

[ W

Y
[ m,._cW2W
[ B
%.
Y WV(W

\
\

. W

[ `
Y
W

[ f W

[ {
B cW3VNW,._cW3 \
Y
V(WP
O(W3V(WB
O Y W2V(Wv`._OcW
[ V(WB
OcW2
Y
[ VNWXBl,.W
[ V(Wl._
Y W
[ V(WXB._cW [ V(W?d%
. OV W
Bm,2O
Y
V W
S \Y
V [ W
Bm,2c
[ V W
d
Y
[
[ V W

\ [ V W

B
._ W
9Y [3V3[Y&
H
[
OcW39Y W [3VN[(Y&dOW Y W [3V3[YB._9Y W
[2VN[(Y+[ S Y
W

[2VN[(Y+S W
9Y
[
[ V3[Y&

W Y
[ W [
[ V [(Y
E

O W Y

[ V W V [(
Y [
W [Y

[ V W V [(
Y [B._
Y
W
[ V W
V [Y
E
O Y W [ V W V [(Y B._ W Y [ V W V [(Y
Y
W
[ V W V [(Y B. Y

[ V W
V [Y R Y

[ V W
V [Y R Y

[ V W
V [Y B,2O W
V
[Y R W

Y

V
[(Y
. W W V
[(Y XB O
W

Y
W V
[(Y B
,2c W
V
[(Y SO
Y W
V
[(Y SO W

[ V
[(Y O
B c W W
[ V
[Y S W
Y W
[ V
[(Y S W

[ V
[(Y ,. W V W V
[Y Bm,.O W V W V
[Y
._
Y
W V W V
[(Y UB [
c W
[ V W V
[(Y P[
W
Y

[ V W V
[(Y P
[
Y
W
[ V W V

[(Y B=
.[V W
V
[(Y P \
Y

V W
V
[(Y v
[\ V W
V
[(Y P\ W
Y \

[ V [[(Y B
._ W Y
W
[ V [(Y d Y
W

[ V [(Y d W Y

[ V W V [Y S Y
W
[ V W V [Y d W
V [(Y B._ W W V [(Y d W
V [(Y T] [
PTB O W[ R. c W
W B._ W

Y
W BR. O W W
U c W
Y
W
c W[ B.
Y \
W[ =. W[
[ (B c W
W
[ ._ W W

[ =%. O W
V W B
._ W

Y
V W B W W V W v,2: W
Y
W V W R. W
V W BV, c
Y
W
V W ._ Y
W
V W B : \ W

[ V W W

Y

[ V W . O W W
[ V W B
. W

[ V W r
Y [
W

[ V W [ BR. O W V W
[ 3
c W
Y
V [ W
R.[ c W V W
BV, c
Y
W V W
._ Y
W V W
r:O W
[ V W
B._ W
Y [

[ V W
B
c W
[ V W

OV W B._
Y V W B.
[ V W =
Y
[ V W W Y [ V [(Y B W
Y W [ V [Y B W Y W
[ V [(Y Y W [ V [Y B
W
Y

[ V W V [(Y :O [ W
Y
W

[ V W V [(Y
W
[Y [
W
[ V W
V[ [
Y MB _ Y
W

[ V W V
[(Y [Y
W

[ V W V
[Y Bv W Y

[ V
W
V [Y B
_9Y W [2V W
V3[YL= Y W [3V W
VN[(YL9Y [3V W VN[(YL=. W V
[Y B: W
(W2V
[(Y W

Y (W2V
[(Y :OcW2 W
V
[Y B._cW3
Y W
V
[Y B
. W[ V

[(Y S

Y
W[ V
[(Y O
B
c W

V(W2V
[Y \
c
W
W2V(WV
[(Y O
B W
V(WV
[(Y d.cW3V W
V
[Y B.
(W2V W
V
[Y S
Y
WV W
V
[(Y T]
[
\ B c W[ W :OcW
(W
E
+ (aW E
B c W W[ \W
B O W[ V W ,._W
W V W Br._W

Y
\ W V W B,. W (W
V [ W 7
O W [
Y
NW
V W 7
[ W[ V W \ B
[
._
Y
W V(W^:O W
V W
B,.cW3W2V W

OcW3
Y
WV W
:c W
V W
MB c
Y
W
V W
Y
W
V W
MB cOW2V W
c(W3V W Bx.
Y
WV W V W
Box 3-3b (Computed Grbner basis for the Grunert distances equations-multivariate terms):
[
_\
4
' cW29Y [S9Y W [?B9Y [NV(W?d.OWVN[(YB.OW3VN[(YB._V(W2VN[Y2]1Y2

.W

[X
BD.OW
[d
.O
[3V(WX
BD

Y

[2V(WXB._9Y
V(W2V3[Y3]

*
[ V(WXBDcW
[ BW
[ dOW9Y

[3VN[(YBD/Y
(W
[3VN[(Y+l._c[ W2V
[(Y
B.O W V
[(Y T] [ '
[ V [(Y BOcV [Y ] Y ]

[ zc
[ B
Y

[ B
[ [
B. Y V [Y BD Y
[ V [(Y B. [ V
[(Y T] ]
[ B Y [ V [Y l._V
[Y ] [
' Y W
BD W Y
d
Y V W T] Y
d
V W
] ' W
[
B W [ l [ V W B
Y
* W
Y
V [Y d Y
W V [Y
] [ z._ Y

d

[ V [Y [ ] Y ]

[ z'
Y

[ d Y [ V [(Y T]
]
[ '.O
[ d Y
V [(Y ] [
._
cW(WZB

l
l
._cW2V(WZB.(W2V(WZBDV W
T]^Yz'OW9Y
V(WOB
9Y WV(WZBD/W Y V
WT]
[YBD9Y
W
[2V(WXBD W
VN[(YBD W
VN[(Y2
cW
V(W2VN[(YBW2VN
W2VNW [(YNT
]%['.#OcWOWW2V3O
B
B OVNWd
[ VNWT]^YN]
[
* 'cW29Y B9Y WZBD9Y V(WXB [3V(WVN[(WZYNB] cW]
[
[h
Y [
V W dO
W V [(Y Bc
W V [(Y BD
W
V [(Y
BDV W V T[BXY 9] Y [
W
[B
[
[
#XB

d
[ T] Y ] \[ z#B._ Y
B
[ V [(
Y T]
] \[ TBX Y

[ B._V [(Y T] [
*icWXBWZBDV(Wd/Y
]1Y(]
BDVN[Y(]^YN]%[ZB
[N]
][ZBD._]
[
cWVN[Y?BDWVN[(YBV(WVN[(YNT]^YN]
cW29Y
d/Y
(WXB9Y
V(W
9RY c
W2WV([2W2VNV3[[Y?Y&BD
]%[TB.cWd.OW}d.V(WXBD
Y
VNWScWV
[Y
* B W V
[(
Y ] ] [
[3TVNBXW2VN [(W Y3 T]^[ Y(B
W [ l [ V W d W Y V [(Y d Y V W V [(Y
R W
[ V
[Y
]
[ #BD

Y
B
[
B Y

[ V [Y BDV
[(
Y T]
] [[
37
Box 3-3b (Computed Grbner basis for the Grunert distances equations-multivariate terms
continued):
Y W [ V W W [ B W [ B [ V W ] Y ] z#B._ W l. W d W
.
(V WZBD

[ V(WT]^YN]%
['./Y
VNWS
cW [2VN[(Y
BD(W [3VN[(YNT] ]%[L
[
'/Y W [}
d/Y [2V(WXB._cW2
VN[(Y+l.W
VN[Y+dcW

VN[Y3]

4*
[ \
[
[
BD
[ ]^YN] [ z._9Y l [3VN[(YNT] ] [ z'/Y [}m._V3[Y(T] [
' W Y
d Y W BD Y V
W T] Y
TB.
W d.O W d
.
V W BD
V W ]
Y
4
* #BX W

[ BD
W
[ d

[ V W d W Y
V [Y B Y
W V [Y ] [ l
._V [(Y ]
[ Y ]
] [

[ V [Y ] Y ]
[ # BD
Y
B
[ B Y

[ V [Y ]
]
[ B Y
V [(Y ] [
4*`#BXOW(WO=VNW2]1Y9Y V(W3] TBXcW3VN[(Y_PW3V3[Y2][OP [3]^Y(] ]%[cP._]^Y$]
[ 9Y [2VN[Y3] ]
[
VN[YN] [[
BXJ Y V
WNW PTBXJ W W 3W ] Y ]
P#BXJ W %
[ 9
[NW ] Y ] [
TBX W 2[Y W 3[([ Y ]
] [ \ P#BXJ Y
W BXJ Y
NW BXJ W %
[ 3[(Y T]
[ .] Y ]
]
[
4cYW *
'
[ N[Y T]
] [ BXJ Y
] [
4cYYZ*V(WS]
Y dVN[(Y]1Y(][]
[
' ]1Y] #]%[ ) above, We note that the element 4cY in Box (3-3a) is a
From the computed Grbner basis of the Ideal 1M2
B
univariate polynomial in ]%[O0 With the coefficients of 4cY known, the univariate polynomial is then solved for ][o

0 B The obtained values of ]%[o B and
and the admissible values inserted in 4cYY in Box (3-3b) to obtain ]1Yo

B
]^Yo
are now inserted in any
the remaining elements of the Grbner basis 4 1030203> 4cYTW in Box (3-3b) to obtain
0 B ofThe
the remaining variable ] o

correct distances are finally deduced with the help of prior information (e.g.
from an existing map).

(iii)
Multipolynomial resultants solution of the Grunert distance equations for the general three-dimension
resection problem
Besides the use of Grbner bases approach as demonstrated above, the Multipolynomial resultants technique can
also be used to solve the Grunert equations for distances. We illustrate the solution of the problem first using the F.
Macaulay (1902) formulation of the coefficient matrix and then the B. Sturmfels (1998) formulation of the coefficient
matrix. We start with the Grunert equations expressed in the form (3-46) as
~Y >*z]
Y S
]
S
9Y
]1Y]
ScW *
~
>@*]
S]
[
l
[N] ]%[dW *
~ [>@*i]
Y S] [
dVN[(
(Y ]1
Y]%[SV(W *
90
(3-50)
]\
Clearly equation (3-50) is not homogeneous. It is therefore homogenized by introducing the fourth variable
and
treating the variable which is to be solved first, say
as a constant (i.e. hiding it by giving it degree zero). The
resulting homogenized polynomial is
]Y
~ $Y Y >*z]
S Y
] Y ]
] \ z' W ]
Y T] \
*
~ Y >@*z]
S]
[ l [ ] ] [ d W ] \
*
~ [Y
>@*z]
[ d
V [(Y ] Y ] [ ]
\ z
]
Y dV W ] \
*
(3-51)
which we simplify as
~ Y$Y >*i]
d Y ] ] \ S ] \
~
Y >*z]

]
[ d
Y ]
] [ d

] \
~ [Y >@*]
[ dV Y ] [ ] \ SV
] \
with the coefficients given as 9YZ*z9Y ]^Y_9 *Q'cW]

Y N%3YZ* [%
(3-52)
*
* W9VYX*iVN[(YN]^Y%V
*`'V(WS]
Y . We now
formulate the coefficient matrix of (3-52) by first using the F. Macaulay (1902) approach followed by the B. Sturmfels
(1998) approach.
Approach 1 (F. Macaulay 1902):
38
The first step involves the determination of the total degree of (3-52) by (2-57) in page (21) of Chapter 2 giving the
value of . In the second step, one formulates the general set comprising the monomials of degree 4 in three
variables by multiplying the monomials of (3-52) by each other. These monomials form the elements of the set j
(2-58) in page (21) as
Oj*
] \
9 ] [
] \ /]
]
[ /] [
] [ 9]
] \
9]
] [ ] \ /]
] [[
]
] \[ 9]
]
[ ] \ 9]
] [ ] \
/]
[ ] \
9] [ ] \[ 9] \\ 9] \[ 9] [[ ] \
(3-53)
which is now partitioned in step 3 according to (2-59) in page (22) as
{2\ ] l [ L ! K ! c [ b b b )|
{]
9]
];\c/]
]
[ /]
]%[O9[ ]
] \ \
9] [
]%[2];\O|
(3-54)
{]
]
[ ];\c[ 9]
[ ] \
9]
] [ 9] [ [ /] \ [ ];\O|
{]
] \ /]
] [ ] \
/] [ ] \ /] \ |c0
In the fourth step, we form the polynomials <! using the sets above according to equation (2-60) in page (22) giving
! \j *

\ *
[\ *
\ *
rise to
\
+Y > * ]

O Y *
\
<
>
* ]
[ [
*
\ \ _[ *
<[>*
] \
<
Finally we can now form a matrix
{]
OY9]
];\LYL9]
[ Y9]
][OY9] \
YL/]%[2];\LY|
{]
] \
/] \

9]
] [
9] [

9] [ ] \
|
(3-55)
{]
%] \_[_/]
]%[L[9]%[3];\LL[_9] \
L[_|O0
S
Bl,

m
B
,
s of dimension
k
(in our case ,xkd, ) whose rows are the coefficients of the L! in (3-55) above and the columns are the monomials
{V2Y*`] \
, [ V
*`] [
]%\[ , VN[*`] [
[ ];\ , V(\*f]
]
[ , V *`]
]%[2];[ \ , V *`]
] \
\ , V *f]
] [[ , V *Q]
]
[ ];\ , V *`]
]%[2] \
,
V2YW *z]
] \ , V2Y$YX*] [ , V2Y
*i] [ ];\ , VYT[*i]
[ ] \
, V2YT\R*][3] \ and VY *z] \ | elements of the sets formed in (3-54) as
V2Y V
VN[V(\ V V V V VV2YTW V2Y$Y V2Y
VYT[ VY\ V2Y
Y , 9Y

]
[ Y , 9Y

]
] [ Y , Y

] \
Y , Y

] [ ] \ Y , Y

]
] \ Y , Y

s* ] ]
] \
\
, Y , , Y
,
]
[
, Y ,

] [ ] \
, Y ,

] ]%[
, 3Y ,

]
]%[L[
, V2Y V

]%
[3];\LL[
, V2Y V

] \
L[ , V2Y V

]
];\LL[ , V2Y V

The determinant of the matrix s above is a univariate polynomial of degree 8 in the variable ]^Y given in Box (3-3c)
below:
39
Box 3-3c (F. Macaulay multipolynomial resultants solution of the Grunert distance equations-univariate
polynomial):
]O\
[ Hc
[ Y
V [(Y Bl,:RBD.O
[ V
[(Y BD[ .O
[
Y
B
[
Y
V
[(Y BD \Y
BS.
[ Y
V [[(Y B \
[ B.
[ [Y
V [Y B
V [(Y \

Y
HOV
[Y B._
Y
V
[(Y B._ Y

[ V [(Y T]-Y
\
\
+(-a Y V
[(Y W BS._
Y V
[(Y W BO.V W lO.O W BO._ W Yc W
Y V
[(Y B
Y V [(Y W BS.O
[
Y W B.
Y V
[Y V W B
\
\
\
\
\

[
Y V W =.O
[ W
Y B.
[ V
[(Y V W .
[ W V
[(Y B. [ V W B._ Y V W ._ Y W . W V [Y B._V [(Y W BP,2:c W
Y B
\
[
,:O W V
[[ Y ,2:
Y
V W [ ,2:OV
[Y W i[ V
[(Y[ V W i
Y
W B=[ .O
[ W ,2:c
[ V W ,:O
[ W Bc
[ W BO Y

[ [ V [Y V W
B
O/Y

[ OWVN[(YcBR9Y

[ (W2VN[(YOBR Y
V [Y
[W/B/Y
V [Y
[3V(W/B
Y
[ V
[Y
[ V(W/B.
[ [ W3
Y
V
[(Y BRc
[3 Y
[VN[(Y(V(W9B
[W29Y VN[(Y",. [N9Y V3[YNOWr ,. [(9Y VN[(Y(V(W [2W2 Y VN[YB [3 Y VN[(Y(cWB O [39Y V [(Y cWB
[

Y
V
[(Y
OWl
[2W9
Y
V [
[Y O
B
[ V
[(Y cW
T] Y
\
\
+(-3b W

Y V
[Y B,.O
[ W2cWBO
Y
[ V W
Br
[ W
V
[(Y . Y V(WW.WNV [(Y cW cWNV(W+B.Rc W
O(WNOW+B
\
cW3V(WB.RO W
B,.
[ WV(W z.O
[ cW2VNWB\c [ cWV(WBm
[ V W
V
[(Y Bm
Y V
[(Y V W
Bm._O
Y (W2V(WB.ROV W
B
W

Y
V
[Y BD \
[
Y
W
Bd
\ ,.OW3
Y
cWXB
.O\ W2V
[Y cW
i,._V
[(Y cW2
V(W?,.
Y
VNW
cWZB \
[ W
B
\
[ V W
B W
V \[Y
,O W
V
[Y B Y
W
Bi W
V [(Y A._V
[(Y V W
Bi Y
V W
[ p,2O
Y
V W
p[ ,2O W
V
[Y A._
Y[
W
A.
[ W
`,O
[ W

,
Y W
,2O
[ V W
BQ,.W3V
[Y V(WvBz/Y V3[Y2 [ V W
Bz9Y VN[Y2 [ W2V(WvBz/Y [ OWW2V3[YB.
[
Y VNW2cW
.O
[ V
[(Y Y
V W
B
c Y
[
[ W[ V [Y V W BR
[ W
V
[[(Y Y

UO
Y
V W V

[(Y[ W U
c W
Y

[ V
W B[
W

Y

[ B Y
[ V
[[(Y

[ W V W B

Y
V
[(Y
[ [ W V W BA
[ V W V [(Y W Y
B
[ [ V W V
[Y W
Y
f._ Y
V [(Y W
[ V W B Y
V [(Y
[ V W
B Y
V [(Y
[ W
B

[ W
Y
V [Y BS
[ W
Y
V
[(Y W B
[ Y
V [[Y V W W Bh
[ W Y
V [Y W B[c
[ W Y
V [Y V W \ W
Y
V [
[(Y W
.O
[ W
Y
V [(Y z._c
[ [ Y
V [Y V W W Bm Y

[ W
\ V [(Y D
c
[ W V
[(Y W z.O
[ W
V [Y Y
z.O
[ W Y
V [Y W B
.O
[ V
[(Y W V W BD
[ W Y
V [(Y W BDO
[ W
V
[(Y ] Y
[
+(6a
Y V W W
O
Y V W
W :c W
V
[(Y W B.O W V
[(Y V W
W
V
[(Y V W B
[ W V
[(Y B:
Y V W
W B:O W V
[Y W
B. W
Y W

. \ W V
[(Y V W
B,2:c W V W
[ W . W

[ W
B \ W W

[ SV, O W
V W
[ . W[
[ Z

O W[ 7
B W[ 3
B O[ V W[ .O W

[ V W[ B
.O
[ [ W V W
B7 W
[ V W
d,O W
[ V W
B.O
[ [ W
V W ZO W
V
[Y V W Zc W V W W .
Y
W
V W B[ . W V
[(Y [ ._
Y
V W
. W V
[(Y %. c W W
%. c W V W
BR. c W
W B.O W
Y
B%. c W
V W BR. cV W
W B[ R. OV W W
.O
[ V W Bx Y
V W

[ W V [(Y B

[ W
V
[(Y W ".
[ W V
[(Y W
"
[ W
V [Y V W Y
"
[ W V [Y Y
V W
BP
[ W V [[ Y Y
B W

Y

[ V W . W [
Y

[ V W

O [2cW2VN[(Y(9Y V W

B O
Y (W2V(WOW3
B OW3V
[Y OWV(WC [2(W2 W
9Y VN[(YCB
[ V W
Y B
[ V W
V
[(Y cWB [NV W 9Y VN[(YB

[

Y
W
V(WB

[3 W[ 9Y

VN[(Y;3
[2 W
9Y
VN[YNV(WCB":c

[2W29Y

VN[(Y(VNW2cWC
B"9Y
V(
W [
[ O
WW2V3[Y1Bv9Y
V(
W [
[
W
VN[(Y;

[2 W
cW2VN[(Y(9Y
H
O
Y
W
cW]
Y
\ \ \
B
[ W2 W
V(W;"[ .
[ W
V(WO[ WB,. W
[ VNW2cW^B \
[ [ V W
W
i W[ cW^B=:O W
W

-aW B=V W B W ,.WNV W
cW9,.OW3V(W W

[
[
[
W2 W B: W
V W
BO W V(W;.O
[ W V(W;
[ W
V W
BcV W cW;.
[ V W cW9MOV W W/MOV(W W BP:V W
W
Bc
[ WV W
cWO0
The univariate polynomial in Box (3-3c) is then solved by the algebraic algorithm such as roots command in MATLAB
to obtain the roots. Once these roots have been obtained, the admissible solution is substituted in the third equation of
B
B
(3-50) in page (37) to obtain h value of
The obtained value of
is in turn substituted in the second
B
equation of (3-50) to obtain the last variable
The correct values of distances are deduced with the help of
prior information.
][ o 0
]
o 0
][ o
Approach 2 ( B. Sturmfels 1998):

From (3-52) in page (37), we compute the determinant of the Jacobi matrix as
t
t ~ $Y Y
]
t ~ Y
* t]
t ~ [(Y
t
]
~ $Y Y
]%\
t
~t
Y
]\
t
~ [(Y t t ~ [(Y
t
] [
]%\
~ $Y Y
][
t
~ Y
t
][
t
(3-56)
respectively
.L]
* .L]
S Y ] \
_.
] \ d Y ]

Y]
O. ] \
d Y ] [ _._. ]] [[ m
dV Y ]
\ .V
] \
d
VY] [
(3-57)
40
]
9 ][/]%\ as
s
*_] ] [ V ] \ O] V Y ]
[ = Y ]
V ] \ .O Y ]
V Y ] [ BD_] 3.L] \ ] [ BDO] 3.L] \
V Y O Y ] \
]
._ Y ] \
Y ]

V

m._ Y ] \
Y ]
V Y ] [ B[c
Y ]
\
N.L] [ BS._
Y ] \[ N.V Y Y
_]

] \ ] [ \O
]

] \
V Y \O Y ]
c Y ]
[ ] \ l. Y ] [ ] \
V Y l. Y ]
[ Y ]
which gives a cubic polynomial in
whose partial derivatives with respect to

of
as
!b
]
/ ]%[O9]%\
[V

. Y ] \ V Y ] [

. Y ]
V Y ] \
] [ m
can be written in the form (2-64) in page (23) with the coefficients
Coefficients:
YY * ;
Y * OV Y .O Y Y (
YT[ * TBXc
V Y c
V Y
. Y Y V
(
YT\ *Q
c Y V Y \ Y ,
Y `* ]O Y V
c Y V Y (
Y `* V
BO

. Y Y V Y N
Y *
O Y l. Y V Y N

$
* ,

[ * #XB O Y
. Y
V Y O Y V
(
\ * ]V Y c Y Y (
*
] V l._9Y33Y(V2Y+\
B[O
(

*`# O9YNV2YCh
c3Y(
N ([(YZ*`._9Y(V2YZc3YNV
N ([
*f'.9Y(V2YOc3Y
( ([$[*
BX:O9Y(
VY
N[#\ *`TB O
./YN3Y(V2Y[
O h
V , ([ *`#B O
V2Y[O9Y(3YNV
5
V2Y3( ([ *
] Y V
BhO Y
Hc Y
V Y

and those of the original equations ! as

9YYX*Q,$/Y
*99Y#[X*z
$9Y\ *9$/Y b # W

YZ*`,$

*Q,
[*

\R*z3YL
*z;
*z
[YZ*9$[
*`,[$[*iV
[\R*i;[ *9$[ *V2Y
The computation of the resultant of the matrix (2-65) in page (23) with the coefficients as given above leads to a
univariate polynomial in
of degree eight as
]1Y
Box 3-3d (B. Sturmfels multipolynomial resultants solution of the Grunert distances equationsunivariate polynomial):
, :O
[ V W
[ W
Br,2:c
[ W W V W
B\ .Rc W
V W
,2:c W W[ B.R W
W
B \
[ V [ W
W
Bc W
V W [ W Ec[ W W V W
Ec
[ V W[ W
c
[ V W W =c W V W W
BYc W BA,2:c
[ W V W W
Bm.RcV W
W
-,:O W W B,2:OV W W Q,:O W V W c
[ W
V W W
,:
V W[ W i,:O W V W[ BOcV \W BhO \W
\
+(-8b [ V W W
7
:c [ W V W
rc. W[ 7
Y
V W[ 7c
[ W
V W 6:O
[ V W
W 6O
[ W
[ W B.%O
Y
W V W
_:O
[ V W W
,:O
[ W V W
E
V
[(Y W x.V [(Y [ Y V W W
gc [ Y W
V [(Y W BcV [(Y [ Y W BvO.
Y W V W W ,c. W V W W
. O

Y
W
V W BvR. W V
[(Y W
r,:

Y
W

W BO \
[ [ V
W
W Bi
:c W
W Bc W
V
[Y

[
W Bc
[
Y

W
V W B:c
[ W V W W
%

[
Y V W W
BO. W V
[(Y V W W
B c [ Y W V [Y t._OV [Y [ Y W V W
O
[
Y W V W
BOV
[Y
[ W V W
. O
V [(Y
[ Y
W V W W ,:
Y
W V W
Bv,:O

[
W W
[ OV
[Y W[ V W

O:O W
W
[
Y
V W
W
[ B8

[
Y
V W[ B.OV
[
Y
[ W[
%
. c W
V
[Y W
%
B c W V
[(Y V W
p,:O W
V
[Y V W {
c
[ W
B c W V
[(Y BO
Y W {c
[ V W B:O W
V W B
Y W W
,:[ VN
[(Y V W W
BO.V W[ BO[ ._ W[ 6
B :OV W W
6
B :O
W VNW
Z
OV [Y [
[ Y
W W

B7c [
[ Y

V [(Y W VNW
Zc
[ Y
NW

V [(Y V W
[ 9Y VN[(Y(V(W W
O
B
[ 9Y
W2VN[(Y(VNW2cWXh
B OVN[(Y(
[39Y
V W H
OV3[Y3
[29Y
W2V W
H
cW3V
[(Y
[ W
]
Y
B
B
B
B
41
Box 3-3d (B. Sturmfels multipolynomial resultants solution of the Grunert distances equationsunivariate polynomial continued):
W
B c W V
[(Y W BcV
[Y
Y \
V W
Bc W
V
[(Y
[ v,O. W W B,._
Y
V
[(Y W
UOV
[Y W
B(O \Y
V W
v,[,.
[ V W W B
O
Y W V W 6c
Y W
B3c W
V [(Y B7c
[ W V W B3O W V
[Y V W OV
[Y V W
OV [(Y [ Y W
B3O Y V [(Y [ V W W
,:
Y
V
[Y W W B,2:c
\
[ V W W E Y
V [[(\ Y

[ W W ,:O
[
Y
W V W Bic
[
Y
V
[(Y
W V W
E
Y
W
B
cV
[
Y
Y
W
B
O
[ W W
[ OO
Y
V W
",O. W V W B(c Y
W
B8c
[
Y
W
B8O
Y
W W U
Y
V W W BP[ ,._V
[Y
[ W
Oc [
[ W
B
cV [Y [ Y V W
YOV [(Y [ Y V W W ,2:
Y V
[(Y W V W B5O
[
Y V W W \V [(Y [ Y W V W Bhc Y [ V [Y W
B
[
B cVN[Y3 [3
[Y V(
W2cW B[OV
[(Y

[ V(WcW?YOWV \[(
Y cW
B[OV \[Y W
B[c
\ [ W
B[c \ [ V W
\
O
c W
V
[(Y B
cVN[Y3 [3 Y W
[

B
[
Y
W
5
cV
[Y
[ V W
Z
O
c
[ V W
[OV
[(
Y V(W2cW}5O [\Y
W2V(W?Bl,.
[
Y
V W
Bl
[ ,O.V(W3cW?
BOO: W
[ Bh:OV W

c
[ W
{
B c [
[ /Y
V3[YNV W

{
B OVN[(YN
[ [2 Y
(W2cWBO[ 9Y
V [(Y
[3 W

B c
B [
Y

V [
[( Y
Y (W2V c
[WY V(B W2c,2W/:O(B VNc[YN [
[ [299Y Y
VN[(W2Y(c Wx
c [29Y W
VN[(Y
=
B

,
O
:
Y
N
V
(
[
N
Y
(
V

W
c
c
W
W
[
OVN[(Y(
[39Y V W
Bm,2
:OVN[Y3
[39Y W2VNWi
,:O
W3V
[Y
[ c(B WOT9] Y\Y
V [(Y
[2(W2V(W/B(c
[ 9Y
W2V3[YNOW/B
[

[ 9Y W2VN[(Y(VNWH
B O \
[ OWPBQ,. cW
B c
Y
W3V [ \[(Y Q[ %: OV
[Y cWP
B \ V
[(Y
Y
V(WP[ B`,%. VNWR
O(W3V \[(Y {
c
[
Y WB
+(128b W
:O
Y
W^Bx,2:OV
[(Y
[ cW^BV
[Y
Y
cW^Bc Y
(W2V [(Y
[9MO[ Y
(WBO9Y
V [(Y [
[3V(W1BiO
[
Y
V
[(Y W;
:%c

[ VNW9
c._\
Y
V
[Y W \ %:
[ OW
VN[(Y(
[39Y
\ cWR
OVN[(YN
[2 Y
(WY
B cVN[(YN
[N \ Y
cWBA,2:O
[
Y
V(W=Bm: cW3V
[Y B
OV [Y W B5c Y W V
[(Y mc._
Y W B\O Y V W mO.V
[(Y V W m:%O
Y V W BHO [ V W BHOV
[Y
[ V W BHO
[
Y W
[ B c
[

Y V
[Y W (B
c
[
Y V
[(Y V
W B,.O [ [
Y V [
Y W B
c W V
[(Y
[ B c.
[ W B,._ Y V [(Y [ W B,._V [Y [ Y V W
[
[
[
,.
[ Y
V [(Y V W HO Y
V [(Y
[ W Bhc
[ Y
W V [Y H
OV [(Y

[ Y

V W Bm,2:OV [Y

[
Y
W T] Y

\
[
[
\
+(-4c [Y B
[
Y V
[Y BO Y V [Y [ BiV [(Y [ Y O.
Y BO Y BP:%C"c.
[ BOV
[Y
Y "c._V [(Y [ Y B
\
[
[ [
B O
[
Y
[
B V
[(Y
[ T] Y
B O
[ Y
V [Y dO.V
[(Y [
+(-96b 8
M. A. Fischler and R. Bolles (1981, pp. 386-387, Figure 5) have demonstrated that because every term in (3-31)
is either a constant or of degree 2, for every real positive solution, there exist a geometrically isomorphic negative
solution. Thus there are at most four positive solutions to (3-31). This is because (3-31) has eight solutions according
to G. Chrystal (1964, p.415) who states that for independent polynomial equations in unknowns, there can be no
more solution than the product of their respective degree. Since each equation of (3-31) is of degree there can only
be upto eight solutions.
Finally, in comparing the two methods, the Grbner bases approach in most cases is slow and there is always a
risk of the computer breaking down during computations. Besides, the Grbner bases approach computes unwanted
intermediary elements which occupy more space and thus leads to storage problem. The overall speed of computation
is said to be proportional to twice exponential the number of variables (D. Manocha 1994 a, b, c, D. Manocha and
F. Canny 1991). This has led to various studies advocating for the use of the alternate method, the resultant and
specifically multipolynomial resultant approach. The Grbner bases can be made a bit faster by computing the reduced
Grbner bases as explained in Chapter 2. The Multipolynomial resultants on the other hand involve computing with
larger matrices which may require a lot of work. For linear systems and ternary quadrics, B. Sturmfels (1998) offers a
remedy through the application of the Jacobi determinants.
(b)
The position-derivation step (3d-ranging or 3d-Bogenschnitt):
This step is commonly referred to in German literature as "Bogenschnitt" problem and in english literature as the
"ranging problem" or "Arc section" (H. Kahmen and W. Faig 1988, p.215) and is the problem of establishing the
position of a point given the distances from the unknown point
to three other known stations
In general the threedimensional "Bogenschnitt" problem can be formulated as follows: Given distances
as observations or pseudo-observations from an unknown point
to a minimum of three known points
, determine the position
of the unknown point
When only three known stations
are used to determine the position of the unknown station in threedimension, the problem reduces to that of 3d closed
form solution. We present below four approaches that can be used to solve the "3d-Bogenschnitt" problem in a closed
form. This problem is a traditional problem both in Geodesy, Photogrammetry and Robotics. In all the three areas, the
determination of the coordinates of the unknown point given the distances from this point to three other known points
is the key issue.
)* , ./;0
[ ;)*,O$./$
{|

o [
ol [
oS "0 [
!Po [
! o
Starting from three nonlinear 3d Pythagorus distance observation equations (3-58) in Box (3-4) relating to the three
unknowns
, two equations with three unknowns are derived. Equation (3-58) is expanded in the form given
by (3-59) and differenced in (3-60) to eliminate the quadratic terms
. Collecting all the known terms of
{2R|

42
equation (3-60) to the right hand side and those relating to the unknowns on the left hand side leads to equation (3-61)
with the terms
given by (3-62). The solution of the unknown terms
now involves solving equation
(3-61), which has two equations with three unknowns. To circumvent the problem of having more unknowns than the
4
D4
).
equations, two of the unknowns are sought in terms of the third unknown (e.g.
{2$|
(i)
{3$}(R|
8* 1 (f* Û
Grbner bases approach:

We express equation (3-61) in the algebraic form (3-63) in Box (3-5)with the coefficients given as in
(3-64). The Grbner basis is then obtained using the GroebnerBasis command in Mathematica 3.0 as
illustrated by (3-65) giving the computed Grbner basis as in (3-66). The first equation of (3-66) is
+4
solved for
and is as presented in (3-67). This value is substituted in the second equation of
(3-66) to give
4
presented in the first equation of (3-68). The obtained values of and are
substituted in the first equation of (3-58) to give a quadratic equation in . Once this quadratic has been
solved for , The values of and can be obtained from (3-67) and (3-68) respectively. We mention
here that the direct solution of 4
as presented in the second equation of (3-68) could be obtained
by computing the reduced Grbner basis as explained in Chapter 2 rather than solving for
4
and substituting in the second equation of (3-66) to give
presented in first equation of (3-68).
4
Similarly we could obtain
alone by replacing with in the option section of the reduced
4
Grbner basis discussed in Chapter 2.
* Û
* 1

8* 1
* 1
* 1

t* Û
(ii)
Multipolynomial resultant (the Sylvester resultant) approach:

The problem is solved in four steps as illustrated in Box (3-6). In the first step, we solve for the first
variable in (3-63) by hiding it as a constant and homogenizing the equation using a variable W as in (369). In the second step, the Sylvester resultant or the Jacobian determinant is obtained as in (3-70). The
{4
resulting determinant (3-71) is solved for
and presented in (3-72). The procedure is repeated
for steps three and four as in equations (3-73) to (3-76) to solve for 4
. The obtained values of
4
and
are substituted in the first equation of (3-58) to give a quadratic equation in .
Once this quadratic has been solved for , The values of and can be obtained from (3-72) and (3-76)
respectively.
* 1
* 1
(iii)
*
* 1
Solution by elimination approach-1:

In the elimination approach presented in Box (3-7), equations (3-77) is a simultaneous equation version of
written in terms of the unknown . By
equations (3-61) with two equations and two unknowns
first eliminating the value of is obtained in terms of the unknown value and substituted in either of
the two equations of (3-77) to give the value of . The values of
are as depicted by (3-78) which
are expressed in a simplified form (3-79) with the coefficients
given by (3-80). The values of
in (3-79) are substituted in the first equation of (3-58) in Box (3-4) to get the quadratic equation
(3-81) respectively (3-82) in terms of as the unknown. The two solutions of are now given by the
second equation of (3-82) with the coefficients 4
given in (3-84). Once we solve (3-82) for we
substitute in (3-79) to obtain the corresponding pair of solutions for
.
{3"|
(iv)
{2$v|
{2|
{VL $OC|
{ ^$C#)|
X
{3$"|
Solution by elimination approach-2:

The second approach presented in Box (3-8) involves first writing equation (3-61) in the simultaneous
form (3-85), which is expressed in matrix form as (3-86). We now seek the matrix solution of
in
terms of the unknown element as expressed by equation (3-87), which is written in a simpler form (389) whose the elements are given by (3-88). The solution of equation (3-87) for
in terms of is
given by (3-90) respectively (3-91) and (3-92) with the coefficients of (3-92) given by (3-93). Substituting
the obtained values of
in terms of in the first equation of (3-58) we obtain a quadratic equation
(3-94) in terms of as the unknown. The two solutions for are given by the second equation in (3-94)
and substituted back in (3-92) to obtain the values of
. The coefficients
in equation
(3-94) are given by (3-96).
FxY$^Y_YN
{}(R|
F
$

{|

{}(R|
{2}(|

{# #|
A pair of solution
and
are obtained. The correct solution from this pair is obtained with the
help of prior information e.g. from an existing map. Of importance is the problem of bifurcation, that is, to identify
the point where the quadratic equation has only one solution, i.e. bifurcates. S. Bancroft (1985), J. S. Abel and J. W.
43
Chaffee (1991), J. W. Chaffee and J. S. Abel (1994), and E. Grafarend and J. Shan (1996) have already treated this
%4 and
D
problem. In the present study, the bifurcation point for (3-82) and (3-94) will be
respectively.

* )
* )U
In Boxes (3-9) and (3-10) the critical configuration of the two-dimensional and three-dimensional ranging problems are
presented. First, the derivatives of the ranging equations (3-97) and (3-104) respectively are computed as in (3-98) and
(3-105). The determinants of the matrices formed by the derivatives are obtained as in (3-100) for the twodimensional
case, while a triple scalar product is obtained for the three-dimensional case to give (3-107). indicate the critical
and
lie in a line with gradient
and
configuration to be the case when the points
intercept E for the two-dimensional case and when the points
and
lie on a plane for the three-dimensional case.
"F$=(X Y F Y $ Y
B
[Ov[[_(+[2

B
" (Z+_Y YL1Y(}YNN;

(

Box 3-4 (differencing of the nonlinear distance equations):
+Y
[
Y
*`xYB
1YBP
YBS

*`
B

BP

BS
[
*` [ B
[ BP
[ BS
*z Y
d Y
m Y
S
d
m
B.L Y tB._ Y -B.O Y
*z

d=

mX

S
}d
mX
?B.L
tB._
-B.O

*z [
d [
m [
S
d
m
B.L"[3tB._[3-B.O&[2
(3-58)
(3-59)
differencing above
"
"
B
*i Y
B
d Y
B
l
Y
5
B5
[
*i

B
[
d

B
[
l
.Ld
B Y l._
B Y d.cU
.Ld [ B
l._ [ B
d.cU
"
v
* Y
5
B
B Y
S
B Y
d
Z*

B[ [
B

S [
BD

d [

B
d
Y

B
[
l
Y &*
[ BS
BS
&*
BS Y
m
BS

l [
(3-60)
(3-61)
(3-62)
Box 3-5 (Grbner basis approach):
W
l W
dV W
Sl W
*z
Y
l Y
dV Y
Sl Y
*z
cW
*.;FxYB
($(W
*.9'^YBD
($V(W
*z.;}YBd

Y
*.;F
B [ ($ Y
*.9'
BD [ ($V Y
*z.;
Bd [
W *` Y
B Y
B Y
B Y
B]
B
B
B

Y

*`]

B

BD

BS

B] [

B [
BD [

BS [

(0
?LLLL7=/2)'{ W d W SV W dd W $ Y l Y dV Y dl Y |cN{2P| )
4 Y * W Y -B Y W -BD Y V W Sd W V Y dd W Y B Y W

4
*z
Y l
Y
dV Y
Sl
Y

4 [ * W
l W
dV W
Sl W
0
f* {c' Y
V W
BD cW
W V Y3
Y $SBDS/Y Y (
W W
BD W
Y
|
Y
Y
Y
W
Y

B

d
V
S

l

c
{
'

/Y
=8*

BD'cY
W V W3
Y SBDl9Y W
W Y
B Y
W
|
8*
(3-63)
(3-64)
(3-65)
(3-66)
(3-67)
(3-68)
44
Box 3-6 (Multipolynomial resultants approach):

in terms of

Y >@*`'cW
dV(W
SlW
l
W

>@*`'9Y
dV2Y
SlY
l
3Y

Step 2: Obtain the sylvester resultant
t
t
t Y t Y
"
s
dV W
Sl W

* t t * 3WY
9WY

t
dV2Y
SlY
E
t
Step 1: Solve for
* W
Y
l W
V Y
Sl W
Y
B Y
W
B Y
V W
mBD Y
W
from (3-71)
* {c' Y
V W
B W W
V Y
Y SBl Y Y
W W
BD W
Y
|
Step 3: Solve for in terms of

L[>@*`W SVNW SlW d(W
\ >@*` Y
dV Y
Sm Y
d Y

Step 4: Obtain the sylvester resultant
t
t
t [ t [
"

cW
'W
dV(W
ddW

* t
*
t
9Y
'3Y
dV2Y
ddY
E

\ t \
t

s
*z W
Y
d W
V Y
Sd W
Y
B Y
W
-BD Y
V W
mBD Y
W
from (3-75)
f* { Y
V W
BD 'cW
W V 3Y
Y dBd9Y Y
W W
B W
Y
|
(3-69)
(3-70)
(3-71)
(3-72)
(3-73)
(3-74)
(3-75)
(3-76)
45
Box 3-7 (solution of the simultaneous equation by elimination):
.Ld
BxYNl._
B1YN*iBS.cU
BYN
(3-77)
L
.
d

v

?
[
B

l
_
.

Z
[
B

z
*

D
B
c
.

U

+

?
[
S
B

^ [ B
+BDL'
B Y
8*
.{cF
B Y N' [ BD
BF [ B
N'
B Y N| BY
(3-78)
f*
^F"[ZB
BD_F
BY3
.{
BD Y N [ B
+Bm' [ BD
3F
B Y | B
"
*zVB
(3-79)
f*zXB
^YZ* {c
BYNN'%[B
+BmU+[?BS
3
BD^Y3(|
{O
B Y N [ BD
BF [ B
N'
B Y (|

* {c
BYNN"[ZB
BmU+[ZBS
N
BxYN(|
{
BD Y NF [ B
B [ BD
N
B Y |
^[ZBD
BL
B1Y2
VX*
.{OF
B Y N' [ B
+Bm [ B
3
BD Y |
(3-80)
{cU BS Y N [ BD B [ B N' B Y N|
* {c
BxYN3[XBD
Bm"[ZB
N
BD^Y2|
* .{' B1^YN3F "[ [ZBB

BBL [Z
BDB NY BxYN(|
r* {{c
BB 1Y NY2N F"[ [ZBB
BmBU [ [ZBSBD
NNF
BB xY YN(N||
Y "^ Y BP Y B=V B=L1 and substituting (3-79) in (3-58)i, one gets a quadratic equation
z,$
l%. `d Y
d Y
m Y
BD._ Y VBD. Y X5
(3-81)
B Y
dV
S
*
B
h
B %4)
*49
lSS)*9(}Y *
(3-82)
.R4
*f
l
i,NF Y
S Y
m Y
B.LxYV?B._1Y(XB5 Y
dV
d

(3-83)
"
*`

d
considering `
in as
where

l
i,
4v*
x*.`
)* Y
S Y
l Y

B .LxY(VB._1Y(XB5 Y
d
V
d

(3-84)
46
Box 3-8 (solution of the simultaneous equation by the matrix approach):
.
BD^Y3Cd.cU
BS}Y3&*z=BD._dF
BY3
.[ZBD
Cd.cU+[?BS
&*B.Ldv[?B

BD Y
B Y " * , " BD. "
BD Y
[ BD
[ B
.
h
[ BD
d
"
"
YN BD. "
BxY
* ., B+'%[?[?BSBD
B'U
BBS1}Y3

h
v[?BY

with *Q{O' B1YN3+[XB B[ZBD N BS}YNN|
"
"
"
Y Y *
B Y X Y
*
B Y X
Y *z [ B
Y ?
$
* Y [ BS
V Y *`BF
B Y (XV
*`BF [ B
^Q Y *

*

"
*-{9Y$Y( B9Y Y| Y "
$
BX9Y
" 2Y " VY

BX
Y Y$Y

V

"
Y
`*-{ Y$Y
$
B Y

Y | Y {
$
Y dV Y BD Y
'
SV
|
i*-{/Y$Y3
$
BD/Y

Y| {/Y$YL
dV
B
YL'2Y&SVY(|
"
f*z_ '{
$
3YBD9Y

|i{
$
V2YBD9Y
V
| )
z*z_'{2 Y$Y
" BD
Y Y |i{ Y$Y V
BD
Y V Y | )
`*z'v5
4c
i*z=)U
"
R*f9YYN

BD9Y

Y3 Y *

3YB9Y

A 4P*z
$
V2Y?B9Y
V
x*z9Y$YN
B
YN3YL)+*i9YY3V
BD
YNV2YLX"*z Y
S Y
l Y
with *z
-4d
9)BxYBD 41YB)#}Y and substituting (3-92) in (3-58) i
B OF
*F

S*z9 * B ._
h

*`
)
d
4
i,N'5
B R
B._ Y Ld

BS.O Y L=d

(3-85)
(3-86)
(3-87)
"
(3-88)
(3-89)
(3-90)
(3-91)
(3-92)
(3-93)
(3-94)
(3-95)
where
C *i
)
S
4
z,
*.9
*B5
B._ Y Ld

B.O Y L=S

(3-96)
47
Box 3-9 (critical configuration of the twodimensional ranging problem):
YLg#xY$^Y_ CY3&*`xYB
1YBP
B5 Y

g # $ &*`
B

BP
B5

t
t Y *`B.9YBN t
*`B.9F B

t
t
Y
t
t
*`B.9'
`
*
B
9
.
'

^

Y
D
B

N

BD
t

* t L! * xYB
B
b Y BD
BD
B
B * x Y ^ Y ,, *

x1YYBD

BD ,
Y\ *fF Y BN BDBF B

N
' Y B P
*i Y
B Y -B
S
AB
Y
S
S Y B
*id Y BD
CSF
B Y CS Y
B
Y
"
(3-97)
(3-98)
(3-99)
(3-100)
,
Y Y , *z.PkLP_

,
*
thus
)Ucm4/X"( Y F Y $ Y Tc F $

,
Y 1Y , *

, V
(3-101)
(3-102)
results in a system of homogeneous equations
cd(dVX*z
c Y dN Y dVX*z
c
dN
dVX*z
(3-103)
48
Box 3-10 (critical configuration of the three-dimensional ranging):
YL#xY$^YLY YN}*`FxYB
1YBD
zU}YBS
B[

#
$

}*`F
B

BD
zU
BS
B[ Y
(3-104)
L[ #"[L$%[&[_ [2}*`F"[ZB
[ZBD
zU+[ZBS
B[
[
t Y
t
t
t *QB.;F Y B( t
*QB.;F
B( t _[ *`B.9 [ B

t
t
t
Y
L[
(3-105)
t
*QB.; Y BP( t
*QB.;
BP( t *QB.; [ BP
t
t
tY
[ *QB.9 [ B
t
t
t
Y
Q
*
B
;
.

B

(

*
B
;
.

S
B

(

X

t
Y
Y
Y

B

D
B

S
B

t
Y

B

D
B

S
B

*
*
B
[
[
Y
b

B

D
B

S
B

(3-106)

,
xYB 1YBD }YBS xY ^Y Y ,

BD }YBS * , *z
+ "
[ZB
B
[ZBD }YBS "[
%
[ &[
B *A{cB Y [ d Y [ BD
[ d [
B Y
S
Y |
S{B Y
B Y [ l Y [ S Y
B [

[ |Cv
(3-107)
S{2 Y
BD Y [ S [ Y B
[ B Y
d
[ |R
Y
[ B Y [
B [
Y
[ Y B
Y [ [ Y
,
Y ^Y Y ,
(3-108)

,
[ [ [ ,
describes six times volume of the tetrahedron formed by the points "F$}(N# Y Y Y ( Y N

(
N and [ F [ $ [ [ (0 Therefore

,
Y 1Y }Y , *z9
* x
(3-109)

, V
"[ [ +[,
results in a system of homogeneous equations
cl(dV2S *z
c Y l( Y dVN Y *z
(3-110)
c
l(
dV2
*
c [ l( [ dV2 [ *
(c)
The orientation step

The fourth procedure presented in Section (3-13) in page (29) of Chapter 2 can then be applied following
the determination of position. This step concludes the solution of three-dimensional resection problem in
a closed form.
3-22 The minimum distance mapping problem

In the minimum distance mapping problem introduced in Example A-3 in Appendix A.1 in page (97), the Cartesian
coordinates
of a point on the topographical surface are given. The task now is to project this point to
a (reference) ellipsoid of revolution under the constrain that the projection distance be minimum. Desired are the
Cartesian ellipsoidal coordinates of this projected point so as to be able to derive the Jacobi ellipsoidal coordinates
(geodetic coordinates) (i.e. ellipsoidal longitude , ellipsoidal latitute and height ) as will be seen in Chapter
5. The solution to the problem involves solving a system of nonlinear equations of Example A-3 in Appendix A.1 in
{3R|
49
page (97) either in closed form or iteratively (see E. Grafarend and P. Lohse 1991). In this section, we present the
application of B. Buchberger algorithm to solving the same problem by computing the Grbner basis of the Ideal
formed by these equations (see Chapter 2). We begin by the equations as presented in E. Grafarend and P. Lohse
(1991) (Example A-3 in Appendix A.1) as follows
BB] Y Cl
] Y ] \ *z
B'-B] Cl
] ] \ *z
BUmB] [
Cd
]
[ ] \ *i

]
Y l
]
d
]
[ B

*z
(3-111)
being the partial derivative of the constrained minimization problem
._ F] Y #]
] [ ] \ &*fBO}
S] \U
F]
Y ]
S
]
[ BD

q *
(3-112)
*`B]1Y3
z'-B]

mB]%[
S];\
]
Y S]
Cd
]
[ BD

[ to a point {2| on the topographical surface along the
e is the displacement vector from the origin of
[ to a
orthonomal triad of base vectors {2 Y# # [_| while denotes the displacement vector from the origin of
point {] Y #] #] [ | on the ellipsoid of revolution along the orthonomal triad of base vectors { Y [ | generated by
(]
o [ >@* ]o [ ]
Y S
]
]

[ *-, d 0
(3-113)
The problem can now be formulated as follows: We are given the Cartesian coordinates {2$}(R| of a point on the
topographic surface, semimajor axis {;| and semiminor axis {| . If for simplicity the origins and are assumed
to coincide i.e. =o, { Y [ |* {3 Y [ | and that the minimum of the distance V O
B is constrained by
the condition of equation (3-113). The task at hand is to find the ellipsoidal Cartesian coordinates {2] Y ] ] [ | of
the topographic point and the Lagrange factor {3];\O| . We have obtained (3-111) by taking the partial derivatives of
(3-112) with respect to the unknowns {2]1YL] ]%[O#]%\_| . In order to solve the nonlinear system of equation for the
unknowns {3]^Y_#] ]%[#];\O| using B. Buchberger algorithm, we write down the generators of the Ideal 1 formed by the
polynomials (3-111) following the lexicographic order {]^Y em] el]%[em];\| as
] Y d
] Y ] \ B] l
] ] \ BD}] [ d
] [ ] \ BX
(3-114)
1 R1*
(
N]
Y dN
N]%
S
N]
[ BD

2(
e
0
The Grbner basis can now be computed by using either Mathematica software or Maple software. For this problem,
Mathematica 2.2 for DOS 387 was used. The executable command is GroebnerBasis[Polynomials,Variables in a
specified ordering]. In this case, the Grbner basis of the Ideal (3-114) is computed following lexicographic ordering
as
{2] Y l
e ]
ed] [ em] \ |
"
{]^Y+l
]1Y(]%\XB#] d
] ];\BD#]%[}d
][N]%\XBX 0
?LLLL7=/2)

]
Y l
]
d
]
[ B

|O
3{2] Y #]
] [ #] \ |
The results of the executable command above are the computed Grbner basis in Box (3-11) .
(3-115)
50
3-11 (Computed Grbner basis for the Minimum Distance Mapping problem):
Box
\
\ ] \ z._ \ \ d. \ T] \[ \ '

H\ O \ \ d
B \

B
,O0

B

] \
'.
l._
B._

B._

B
.

T] \ z'

B

BD

BD

(0
\
' mB._
\
\ Sl\ \ ][ZB \ ] \[ Bm._ \ \ S \ \ ] \
B
2. '
l._ B

BD

BD

T];\XBD

d

l._

B \
0
.\
Sd \ ] \ \ mBD
T] [ S \ ] \[ z\ ._ \ \ d
] \

3. '
l._
B

BD

B
T] \

BD

B

BD.O

0
4. T,d
] \ ] [ BS
"
\ \ B._

d \ ]
[ '._

\ mB. \ T] [ BD \ ] \
B._ \ \ ] \ B
5.

d

d

d
(0
.
\ BD
d \ ] \ \ ]
[ BD
Z] [ d \ ] \[ \ '._ \ \ d.
] \

6. P' (
}H
O
2 BD
2(
3
?BDc
(
3
ZB
3] \ l._
3(
XBD.O(
(
B.O

B.

0
"
F
[ S
]
d
\ =] \
dx'

l
]
[ B
X] [ ] \ S=]
[
7.
BX BDPX] [ BD
0
8.
T,l
] \ T]
BD
9. (

] [ B
] [ l
]
B
] [
10.
] Y B]
11.
] Y S
\ ] \
z'

l
]
[ B
X] [ ] \ S]
[ BSZ] [ d]

B
BD
0
12.
#,d
];\]^YZB
13.
'
]%[ZB
]%[l
]1YBD
]%[
14.
]
Y S
\ ] \
z._

d
]
[ B
Z][2]%\?l.L]
[ BD.cZ][}]
B
B
0
From the computed Grbner basis above, it is clearly seen that the first equation is a univariate polynomial of order
four in
which is identical to that obtained by E. Grafarend and P. Lohse (1991, p.94) and can easily be solved for
the four roots using the available software such as Matlab or Maple. Once the four roots of have been obtained, they
are substituted in the polynomial equations (4, 8 and 12) to obtain the unknown variables
respectively
thus concluding the solution of the Minimum Distance Mapping problem. In Chapter 5, we apply these computed
Grbner basis to solve the Minimum Distance Mapping problem for a real case study in order to obtain the Jacobi
ellipsoidal coordinates given geocentric cartesian coordinates. In this section, we test these computed Grbner basis
to solve the Minimum Distance Mapping problem presented by E. Grafarend and P. Lohse (1991) below:
]\
]\
{2] Y ]
#] [ |
Example 3-1 (E. Grafarend and P. Lohse 1991, p.108):

Given are the geometric parameters of the ellipsoid of revolution; semi-major axis
= 6378137.000m and first
numerical eccentricity
0.00669437999013 from which the semi-minor axis
can be computed. The Input are
Cartesian coordinates of 8 points on the surface of the earth presented in Table (3.1). From the computed Grbner
{|
{%|
51
Table 3.1: Cartesian coordinates of topographic points

Point
1
2
3
4
5
6
7
8
d
3980192.960
0
0
4423689.486
4157619.145
-2125699.324
5069470.828
213750.930
x
=
0
0
0
529842.355
664852.698
6012793.226
3878707.846
5641092.098
4967325.285
6356852.314
-6357252.314
4555616.169
4775310.888
-91773.648
-55331.828
2977743.624
basis in Box (3-11), we have the first polynomial equation as a univariate polynomial equation given as
V \ ] \\ S V [ ] \[ SV
] \
SV Y ] \ dV *
V \ *z \
V [ *`._ d.\ \ \ \
V
*`' \
HO \ S
B

D
B \

BD
\

V2YZ*`._
d.
B._

BD.

B._

V *Q'

B

BD

BD

(0
(3-116)
We then proceed to compute the coefficients of the given univariate polynomial (3-116) using the input data of
Table (3.1). The computed coefficients are as given in Table (3.2) . With the computed coefficients, the polyTable 3.2: Computed polynomial coefficients
Point
1
2
3
4
5
6
7
8
VW
-7.7479e+22
-5.1720e+22
-2.5861e+23
-2.5311e+24
-1.8076e+23
-5.1549e+21
-2.6815e+24
-4.5942e+24
] \ *LL V(\ VN[ V

V2Y V(W
VY
1.3339e+41
1.3374e+41
1.3372e+41
1.3310e+41
1.3334e+41
1.3285e+41
1.3263e+41
1.3267e+41
V[
1.3515e+55
1.3529e+55
1.3529e+55
1.3507e+55
1.3513e+55
1.3493e+55
1.3488e+55
1.3493e+55
4.3824e+68
4.3824e+68
4.3824e+68
4.3824e+68
4.3824e+68
4.3824e+68
4.3824e+68
4.3824e+68
V\
4.4420e+81
4.4420e+81
4.4420e+81
4.4420e+81
4.4420e+81
4.4420e+81
4.4420e+81
4.4420e+81
nomial roots can be computed in Matlab by the roots command (D. Hanselman and B. Littlefield 1997, p.146) as
. The obtained roots are then substituted in the polynomials (4, 8, and 12) of the
computed Grbner basis
{2]%[O#]
# ]1Y|
T,d
] \ ] [ S
B
T,l
] \ T]
B
T,l
] \ T] Y BD
(3-117)
respectively. The computed results presented in Table (3.3) are identical to those
to give the values of
obtained by E. Grafarend and P. Lohse (1991, Table 4, p.108). Once the ellipsoidal Cartesian coordinates have
been derived, the Jacobi ellipsoidal coordinates (ellipsoidal longitude , ellipsoidal latitute and height ) can be
computed as in Box (5-1) of Chapter 5.
3-23 GPS positioning with observations of type pseudo-ranges

3-231
The pseudo-ranging four-point problem in GPS positioning
E. Grafarend and J. Shan (1996) have defined the GPS pseudo-ranging four-point problem ("pseudo 4P") as the problem of determining the four unknowns comprising the three components of the receiver position and the stationary
receiver range bias from four observed pseudo-ranges to four satellite transmitter of given geocentric position. Geometrically, the four unknowns are obtained from the intersection of four spherical cones given by the pseudo-ranging
52
Table 3.3: Computed ellipsoidal cartesian coordinates and the Lagrange factor
]Y
Point
1
2
3
4
5
6
7
8
][
]

0.000
3980099.549
0.000
0.000
4420299.446
4157391.441
-2125695.991
5065341.132
213453.298
0.000
0.000
529436.317
664816.285
6012783.798
3875548.170
5633237.315
4967207.921
6356752.314
-6356752.314
4552101.519
4775047.592
-91773.503
-55286.450
2973569.442
]\

5.808116e-019
3.867016e-019
1.933512e-018
1.897940e-017
1.355437e-018
3.880221e-020
2.017617e-017
3.450687e-017
equations. Several procedures have been put forward for obtaining closed form solution of the problem. Amongst the
procedures include the vectorial approach evidenced in the works of S. Bancroft (1985), P. Singer et al. (1993), H.
Lichtenegger (1995) and A. Kleusberg (1994,1999). E. Grafarend and J. Shan (1996) propose two approaches.
"k
One approach is based on the inversion of a

coefficient matrix of a linear system formed by differencing of the
nonlinear pseudo-ranging equations in geocentric coordinates, while the other approach uses the coefficient matrix
from the linear system to solve the same equations in barycentric coordinates. In this section we present both the
approaches of Grbner bases and Multipolynomial resultants (B. Sturmfel 1998 approach) to solve the same problem.
We demonstrate our algorithms by solving the Pseudo-ranging four-point problem already solved by A. Kleusberg
(1994) and E. Grafarend and J. Shan (1996). It will be illustrated that both the Grbner bases and the Multipolynomial
resultant solve the same linear equations as those of E. Grafarend and J. Shan (1996) and lead to identical results (see
also J. L. Awange and E. Grafarend 2002). We start with the pseudo-ranging equations written algebraically as
]
]
]
]
Y BD
Y BD
Y BD
Y BD
W
Y
[
W Y $
$ [
{ W Y
[|
z]
B W
F] [ BV W
Bm] \
z]
B Y
F] [ BV Y
Bm] \
z]
B

F] [ BV

Bm] \
z]
B [
F] [ BV [
Bm] \
%2L ] Y ] ] [ #] \ o
W W $V W &*fF] WY
a WY b WY z WY
Y Y $V Y &*fF] a b z

$V
&*fF]
[ a
[ bc
3[ z
[
[_([_$VN[2&*fF] a b z
B W
B Y
B

B [
*z
*z
*z
*
z
(3-118)

{ W $ W V W |}3{2 Y $ Y V Y |&N{
$
V
|&N{ [ $ [ V [ |}N{ W Y
[ |

{3] Y #]
] [ |
with
being the position of the four GPS satellites, their ranges to the stationary receiver at
given by
. The parameters
are
elements of the spherical cone that intersect at to give the coordinates
of the receiver and the stationary receiver range bias
. The equations above can be expanded and arranged in the lexicographic order
as follows
]%\
{3]^YRem]
m
e ]%[ed];\O|
]
Y BD. W ] Y S]
]
BD. Y ] Y S]
]
YY BD.
] Y S]
]
Y BD.[(]^Y&S]
{3] Y #]
] [ #] \ |
]
Y ]
S]
[
]
Y ]
S]
[
]
Y ]
S]
[
]
Y ]
S]
[
BD.O W ]
BD.O Y ]
BD.O
]
BD.O([(]
[ B._V W ] [ B]
[ B._V Y ] [ B]
[ B._V
] [ B]
[ B._VN[3]%[ZB]
. W ] \ l
\
l
\
l. Y ] \ l
\
l.
] \ l
\
l. [N]%\l
W
d
Y d
Y
[ d
[
dV W
dV
Y
dV
dV
[
B W
B
Y
B
B
[
*i
*i
*i
*
i
(3-119)
with the variables

, the other terms being known constants. We re-write (3-119) above with the linear
terms on one side and the nonlinear terms on the other as
B]
B]
B]
B ]
* ._
\

\
*._
\
*._
\
*._
W]
Y]
[ ]]
Y l. W ]
Y l. Y ]
Y l.
]
Y l. [ ]
._V W ]
l
._V Y ]
l
._V ]
l._V
[ ]
[
B . W ]
[ B. Y ]
[ B.
]
[ B. [ ]
\ W
\
Y
\
\
[
BD
BD
BD
B
D
B W
W
D
Y BD
Y
BD
[ BD
[
BV W
BV
Y
BV
BV
[ 0
(3-120)
On subtracting (3-120iv) from (3-120i), (3-120ii), and (3-120iii) we obtain
W[ @> *z W[ ] Y d W[ ]

dV W$[ ] [
Y#[ @> *z Y#[ ] Y d Y#[ ]
dV YT[ ] [

[ @> *i
[ ] Y l
[ ]
dV [ ] [
#[ W ] \ S
W[ i
*
[Y ] \ S
#
Y
[
*
[ ] \ d [ *i
(3-121)
53
cW[*z.;cWXBDc[(%(W$[*.9WZBDN[2(%V(W$[*.9'V(WZBDVN[2N [WR*.9 [?B WL(

Y#[ *z.; Y BD [ (% YT[ *.9 Y BD [ (%V YT[ *.9'V Y BDV [ N [(Y *.9 [ B Y (

[ *z.;
BD [ (%
[ *.9
BD [ (%V
[ * .9'V
BDV [ N [
*.9 [ B
(
W$[ *` W
BD W
BD W
BDV W
Bm
[ B
[ B
[ BDV
[ N
YT[ *`
Y BD
Y BD
Y BDV
Y Bm
[ B
[ B
[ BDV
[ N

[ *`
BD
BD
BDV
Bm
[ B
[ B
[ BDV
[ N0
We note immediately
that (3-121)
comprises three equations which are linear with four unknowns. Treating the
unknown variable ] \ as a constant, we apply the Grbner bases and the Multipolynomial resultant techniques to solve
the linear system of equation for ] Y =
* 4^] \ N#]
D
* 41F] \ N#] [
* 41F] \ , where 4^] \ is a linear function.
with:
Approach 1 (B. Sturmfels 1998 Multipolynomial resultants approach):

Depending on which variable we want to solve, we can re-write equation (3-121) such that this particular variable is
D4
for instant, we write (3-121) by hiding
hidden (i.e. is treated as a constant). If we are interested in solving
as
]^Y
with
]%
OYX>@*fcW$[2]^Y}

>@*f9YT[2]^Y}
[ >@*f
[ ] Y
1] YX* 1F]%\
[#W2];\ZSW[2] lW$[3]
dV(W[N][
[Y(];\ZS_Y#[2] l3YT[3]
dV2Y#[N][
[
] \ S Y#[ ]l
[ ]
dV
[ ] [
being a homogenizing factor. The Jacobian determinant of (3-122) then becomes
y
t
t Y
]
t
* t ]
t
[
t
]
t
t
t
t
][
]
[
][
t
t
t
Y
]%
]%
t
t
t
[
]%
t
W$[ NV W$[ c W[3]1Y&

* 3 YT[ V YT[ 9 Y#[3]1Y&

[ V
[
[3]1Y&
The determinant obtained in (3-123) gives the variable
]^Y*41F];\L
[W3];\?SW[
[(Y];\?S_Y#[ 0
[ ];\?S [
(3-123)
as
] Y *QB' W$[ YT[ V

[ [
] \ W$[ V YT[ [ W ] \ Y#[ V
[ B [# W ] \ V Y#[
[ B [ Y ] \ W$[ V
[ B W$[ V YT[
[ B
Y#[ W[ V
[ Y#[ V W$[
[
[ W$[ V Y#[ [Y ] \ V W[
[ B [
] \ V W[ Y#[ B
[ V W[ YT[ 9
[ V Y#[ W[
Y#[
[ V W$[ B YT[ V
[ W[ B
[ YT[ V W$[ BD W$[ V Y#[
[ d W[ V
[ Y#[ (0
For
(3-122)
]

* 41F] \
we have
L\R>@*f'W[N]
>@*f'3Y#[N]
>@*f'
[N]
W[2T] d
cW$[3]^Y&dV(W[N][
[([ W3Y]%]%\\S
S
#
Y
2
[
T

]
9YT[3]^Y&dV2Y#[N][
[ ]%\S [2T] d

[3]^Y&dV
[N][
whose Jacobian determinant is given by
t
t \
] Y
t L
* t ]^Y
t
t
]Y
t
t
][
L
] [
][
t
t
\
]
L
O/ WY#[[ (V2V WY#[[ 3 WY#[N[N]]

y
t
t
*
]

[ V
[
[N]
t
t
t
t
]
The determinant obtained in (3-125) gives the variable ]
* 41F];\L as
W[
[([ W3Y];];\?\?S
_Y#[ 0
[ ];\?S
S
[

]
Q
* B '
[2V2YT[ [W];\%" cW[3V
[ [(Y(]%\%"cW$[2V
[3_YT[^BP
[2V(W$[ [(YN];\1BcW[3V2Y#[ [
]%\1BcW$[2V2Y#[3
[9
Y#[ V W[ [
] \ Bx Y#[ V
[ [W ] \ Bx Y#[ V
[ W[ B
[ V W[ Y#[ r
[ V YT[ W[ r YT[ V W$[
[ /'
[ V YT[ W[
B YT[ V
[ W[ B
[ YT[ V W$[ BD W$[ V Y#[
[ d W[ V
[ Y#[ (0
Y#[
[ V W$[

(3-124)
(3-125)
54
For
][ *=4^];\
we have
R>@*`'V W$[ ] [ [#W ] \

>@*`'V YT[ ] [ [ Y ] \
>@*`'V
[ ] [ [
] \
whose Jacobian determinant is given by
t
t
t
t t t
] Y ]
]%
tt tt tt *
s
y *
] Y ]
]%
t t t
t
]1Y t ]
t ]
The determinant obtained in (3-123) gives the variable
d W$[ ]d W[ ] Y l

W$[ ]
d YT[ ]d Y#[ ] Y l

YT[ ]
d
[ ]d
[ ] Y l

[]
W#Y [[ W$TY [[ VV W$YT[[ ]]

[
[ V
[]
]%[*=4^];\L
(3-126)
[ [ W ]
[ [(Y ]
[ [
]
\ d W$[
\ d YT[ 0
\ d
[
(3-127)
as
] [ *`B
[ W[ [(Y ] \ x W$[ Y#[ [
] \ x W[ YT[
[ BP
[ Y#[ [W ] \ Bv W$[
[ [Y ] \ Bv W[
[ Y#[
YT[
[ [W ] \ Bx YT[ W$[ [
] \ Bx YT[ W$[
[ Bx
[ Y#[ W$[ r
[ W[ Y#[ r Y#[
[ W[ /
[ W[ V YT[
YT[
[ V W[ B Y#[ W[ V
[ BD
[ Y#[ V W[ BD W[
[ V YT[ S W[ Y#[ V
[ N0
On substituting the obtained values of
function in
given in Box (3-12).
]\
] Y *41F] \ (]

* 41F] \
and
] [ *4^] \
in (3-119i) we a get a quadratic
Box 3-12 (Univariate (quadratic) polynomial obtained from Multipolymomial resultants solution of
GPS pseudo-range equations):
TBX
[ W[ [Y
[ Y#[ [#W W[
[ [Y Bf W[ Y#[ [
Bf Y#[
[ [W Y#[ W$[ [

9
[ V Y#[ W[
Y#[
[ V W[ Bf YT[ V
[ W[ Bf
[ Y#[ V W[ Bf W$[ V Y#[
[ W$[ V
[ Y#[
TBX
[ V Y#[ [#W
[ V W[ [(Y
W[ V Y#[ [
BP W$[ V
[ [Y B Y#[ V W[ [
" YT[ V
[ [#W
/'
[ V YT[ W$ [ Y#[
[ V W$[ BP Y#[ V
[ W[ BP
[ YT[ V W[ B
W[ V Y#[
[ W[ V
[ Y#[
8TB [
W[ V YT[ B [W YT[ V
[ [Y W[ V
[ B [(Y V W[
[ [
V W$[ YT[
[#W V Y#[ [
9 [ V Y#[ W[ Y#[
[ V W$[ B Y#[ V
[ W[ B
[ YT[ V W[ B W[ V Y#[
[ D W$[ V
[ Y#[
BS,] \

'.9
Y#[ V
[ W
[ Q W[ V Y#[
[ B
W$[ V [ YT[ BA YT[ V W$[
[ `
[ V W$[ YT[ B-
[ V YT[ W[ /'
[ V YT[ W[
9Y#[2
[3V( W[PB9YT[2V
[W [B
[3YT[2V(W[ B
OW[3VYT[
[cW[3V [22YT[BAW3TBX [2V2Y#[ [#W=- [3V(W[ [(YR
cW[3V2Y#[ [
B cW$[2V
[ [YB /Y#[3VNW$[ [
9Y#[2V
[ [W9

[2V2Y#[2(W$[z 9Y#[

[2V(W$[AB /Y#
[V
[W[B

[23Y#[3V(W[BfcW$[V2Y#[2
[ cW$[2V
[23Y#[2.;#TBXW[23Y#[3V
[ W[3V2Y#[2
[ _YT[W[3V
[Bf_YT[2V(W$[
[xB

[2W[3 V2Y#[1r
[NVNW$[3Y# [2 9
[3V2Y#[2W [1r9YT[3
[2V(W$[ Bx9Y#[3V
[2W[Bx
[32YT[2V(W$[Bx cW[3V2Y#[2
[OW[3V
[3YT[9B
cW3TB [
W[3V2Y#[B [#W3YT[2V
[" [(YNW[3V
[B [Y3V(W[2
[" [
V(W$[3Y#[v [WV2Y#[2
[/'
[V2Y#[2W[v
9Y#[2
[3V(W[?B9Y#[3V
[W$[B
[3YT[2V(W[BcW$[V2Y#[2
[&ScW[3V
[23Y#[2d. W}d.;#9Y#[2W[3
[&cW$[
[3_Y#[B
W[ Y#[
[ B YT[
[ W[
[ Y#[ W$[ B
[ W[ Y#[ /
[ V Y#[ W$[ Y#[
[ V W[ B Y#[ V
[ W[ B
[ YT[ V W[ B
W[ V Y#[
[ A W[ V
[ YT[ ZBzV W NTBX
[ W[ [(Y
[ Y#[ [#W A W[
[ [(Y Bz W[ Y#[ [
Bz YT[
[ [W
Y#[ W[ [
9
[ V Y#[ W[ d YT[
[ V W[ B Y#[ V
[ W$[ BD
[ Y#[ V W[ BD W[ V YT[
[ S W[ V
[ Y#[ ] \
#TBX W[ Y#[ V [ W[ V YT[ [ YT[ W$[ V [ B YT[ V W$[ [ Bp [ W[ V Y#[ [ V W[ Y#[ /' [ V YT[ W[
Y#[
[ V W[ Bv Y#
[ V
[ W$[ Bv
[
YT[ V W[ BP W[ V Y#
[
[ x W[ V
[
YT[ cBP
W
B W

Y#[ W[
[
W[
[ Y#[ B
W[ Y#[
[ B YT[
[ W[
[ Y#[ W$[ B
[ W[ Y#[ /
[ V Y#[ W$[ Y#[
[ V W[ B Y#[ V
[ W[ B
[ YT[ V W[ B
W[ V Y#[
[ d W$[ V
[ Y#[ BDV W
i Y#[ V
[ W[ d W[ V YT[
[ BD W[ V
[ YT[ BD YT[ V W$[
[ S
[ V W[ YT[ B

[3V2Y#[2W[2 /
[3V2Y#[2(W$[d9YT[
[3V(W[?B/Y#[3V
[W$[?BD
[3Y#[3V(W[?BDOW[V2YT[
[}ScW[2V
[22YT[B(W2
Approach 2 (Grbner bases technique):

Using the Grbner bases approach, we compute the Grbner basis of the linear system of equation (3-121) as
follows:
{ W[ ] Y l W$[ ]
dV W[ ] [ [W ] \ S W[
Y#[ ] Y d YT[ ]
d V Y#[ ] [ [Y ] \ S Y#[
?LLL7=/2)T

[ ] Y l
[ ]
SV [ ] [ [
] \ S
[ |O3{2] Y #]
] [ ] \ |
giving the computed Groebner basis as in Box (3-13).
(3-128)
55
Box 3-13 (computed Grbner basis):
Y @> *tTBX [ # Y#[ W[ YT[ [ W$[ [ W[ Y#[ Bl W[ [ Y#[ Bd YT[ W[ [ W[ Y#[ [ Bl [ YT[ V W$[ ] [
Y#[
[ V W$[ ]
[
[ W$[ V YT[ ] [ B
W$[
[ V Y#
[ ] [ B YT[ W$[ V
[ ]
[ W$[ YT[ V
[ ] [
B
[ Y#[ [#W
] \
Y#[
[ [W ] \

[ W$[ [(Y ] \ B W[
[ [(Y ] \ B Y#[ W$[ [
] \ S W[ YT[ [
] \
4
TBX
[ T Y#[ S Y#[
[ B
[ Y#[ ]
d Y#[
[ ]
BD
[ V Y#[ ] [ S Y#[ V
[ ] [ BD
[ [(Y ] \ S Y#[ [
] \
4 [ >>*`
*`TBX
[ T W[ S W[
[ B
[ W[ ]
d W[
[ ]
BD
[ V W[ ] [ S W[ V
[ ] [ BD
[ [W ] \ S W[ [
] \
4_\>*`TBX9YT[TW[ScW[3_Y#[ZB9Y#[2W[N]
d[ O]%W\ [22YT[3]
BD9YT[V(W[N]%[ScW[3V2Y#[3][ZBD/Y#[ [W2]%\ScW[ [Y(];\
4 >*z [d [3]^Y&l [N]
S
V
N
[

SV2
Y#[N][ [(
Y]%\
4 >*zLY#[d9YT[3]^Y&l3Y#[N]
SV(W[N][ [W3]%\O0
4 >*z2W[dcW$[3]^Y&lW[N]
][ ]\
]
* 1F] \ (
] [ * 1F] \
] Y * ^] \ (#]
* ^] \ ] [ * 1F] \
We notice from the computed Groebner basis in Box (3-2) that 4 is a polynomial with only
and
as variables.
With 4 expressed as
, it is substituted in 4 to obtain
which together with
are
4
4
4
substituted in 4 to give
. On substituting the obtained values of
and
4
4
=4
4
in (3-119i) we get a quadratic equation in
given in Box (3-14) as
] [ * 1F] \
] Y * ^] \
]%\
Box 3-14 (Univariate (quadratic) polynomial obtained from Grbner basis solution of GPS
pseudo-range equations ):
TBX
[ Y#[ [#W Y#[
[ [W
[ W$[ [(Y B W[
[ [Y B Y#[ W$[ [
W$[ Y#[ [

/
[ Y#[ V W$[ B
Y#[
[ V W$[ BD
[ W[ V Y#[ d W[
[ V YT[ d YT[ W[ V
[ BD W[ YT[ V
[
TBX
[ V Y#[ #BX
[ YT[ [W d YT[
[ [W

[ W$[ [(Y Bl W[
[ [(Y BS Y#[ W$[ [
W$[ YT[ [
/'
[ YT[ V W[ Bd Y#[
[ V W$[ Bd
[ W[ V Y#[ i W[
[ V YT[
/Y#[2(W$[2V
[PBcW[2YT[2V
[Z-9YT[V
[TBX
[23Y#[ [W`9YT[
[ [W=Q
[2W[ [Y=BAOW[2
[ [(Y=B9YT[W[ [

OW[22YT[ [
/
[2YT[2V(W$[vBA9YT[
[2V(W[vBA
[2W[3V2Y#[=fOW[2
[2V2YT[=`9YT[ W$[2V
[vBAcW[23 Y#[3V
[2RB-
[ [(Y
/Y#[ [

/'
[3YT[BA 9Y#[2
[2
'
[#BX
[3VYT[OTBX
[3Y#[ [#Wv`9Y#[2
[ [#W"`
[2W[ [(YvB-OW[2
[ [(YPB
/Y#[2(W$[ [
lcW$[ 3Y#[ [
9
[3 Y#[3V(W[XB9YT[
[3V(W[XB
[2W [3V2Y#[lcW[2
[3 V2Y#[m9Y#[2W[3V
[BDcW[23Y#[3V
[C
/Y#[3V
[OTBX
[3Y#[ [#W 9YT[
[ [W
[2(W$[ [(Y"BQcW[
[ [(YvB`9YT[ W[ [
cW$[3YT[ [
/'
[3YT[2V(W[B
/Y#[2
[2V(W$[B
[2(W$[2V2YT[&S OW[2
[2V2YT[} /Y#[W$[2V
[B cW$[3YT[2V
[2 B
[ [YS /Y#[ [
/
[23Y#[B9Y#[2
[21
V
[ TBX
[ Y#[ [#W Y#[
[ [W
[ W[ [Y B W[
[ [(Y B Y#[ W$[ [
W$[ Y#[ [
9
[ Y#[ V W[ B YT[
[ V W[ B

[ W$[ V YT[ S W[
[ V Y#[ d Y#[ W$[ V
[ B W[ Y#[ V
[ C [

L
[ Bl,] \

+(2((-a [ YT[ Y#[ [ B [ V YT[ TBX [ Y#[ W[ YT[ [ W$[ [ W$[ Y#[ Bp W[ [ Y#[ B YT[ W$[ [
W[ YT[

[ /'
[ YT[ V W$
[ B
YT[
[ V W[
B
[ W[ V Y#[
W$[
[
V Y#[ Y#[ W[ V
[ B W[ Y#[ V
[
p
Y#[ V
[ TBX
[ Y#[ W$[ YT[
[ W$[
[ W$[ YT[ B W[
[ Y#[ B Y#[ W[
[ W[ Y#[
[ 9
[ Y#[ V W[ B
Y#[
[ V W$[ B
[ W[ V YT[ W[
[ V Y#[ YT[ W[ V
[ B W$[ YT[ V
[ # /'
[ YT[ B Y#[
[ B
W N#BX [ V Y#[ TBX [ Y#[ [W YT[ [ [W [ W[ [Y B W$[ [ [(Y B Y#[ W[ [ W[ YT[ [ / [ Y#[ V W$[ B
Y#[
[ V
W$[ BA
[
W[ V Y#[ f W$[
[
V Y#[ ` Y#
[ W[ V
[ B- W$[
Y#[ V
[ Xf YT[ V
[ #
BX
[ YT[ [W
f Y#
[
[ [W

[ W$[ [(Y Bl W[
[ [(Y BS Y#[ W$[ [
W$[ YT[ [
/'
[ YT[ V W[ Bd Y#[
[ V W$[ Bd
[ W[ V Y#[ i W[
[ V YT[
Y#[ W$[ V
[ BS W$[ Y#[ V
[ }BS
[ [(Y m Y#[ [
9
[ Y#[ BS Y#[
[ }Bd.9#B
[
[ #BX
[ Y#[ m Y#[
[ B

[3VYT[OTBX
[3Y#[3W$[x9YT[
[2W$[x
[W$[2_YT[B OW[2
[2LY#[B 9Y#[2W[3
[xcW[23Y#[3
[2 9
[3Y#[NV(W[rB
/Y#[2
[2V(W$[B-
[2W[3VYT[vfcW[2
[3VYT[p9Y#[2W[3V
[BQcW[23Y#[3V
[2R`9Y#[2V
[OTBX
[3Y#[3W$[Pf9Y#[2
[3W[

[2(W$[2LY#[RBlcW[2
[3_Y#[Bl9YT[W[3
[zcW$[3YT[2
[ /'
[3YT[2V(W[Bl9YT[
[3V(W[Bl
[2W[3V2Y#[icW[2
[3VYT[X
/Y#[2(W$[2V
[tB OW[22YT[2V
[# /'
[22YT[B 9Y#[
[- V
[TBX
[23Y#[3W[ 9Y#[2
[3W[
[W$[2_YT[tB
OW[2
[2LY#[RBm9Y#[2W[3
[icW[23Y#[3
[2 9
[23Y#[3V( W[Bl9Y#[
[2V(W$[RB
[W$[2V2Y#[icW$[
[3V2Y#[i9Y#[W[3V
[RB
W[ YT[ V
[ # L
[ B W 3'
[ TBX
[ V Y#[ TBX
[ Y#[ [W S Y#[
[ [W l
[ W$[ [(Y BD W[
[ [(Y BD Y#[ W$[ [

W[ YT[ [
/
[ YT[ V W$[ B Y# [
[ V W$[ B
[ W$[ V YT[ W[
[ V YT[ Y#[ W$[ V
[ B W[ Y#[ V
[ f
Y#[ V
[ TBX
[ Y#[ [#W YT[
[ [W
[ W$[ [(Y BQ W[
[ [(Y B` YT[ W[ [
W$[ YT[ [
/'
[ YT[ V W[ B
Y#[
[ V W$[ B
[ W$[ V YT[ S W [
[ V YT[ Y#[ W$[ V
[ B W$[ YT [ V
[ B
[ [ Y S Y#[ [
/
[ Y#[ B Y#[
[ 1
V
[ TBX
[ Y#[ [#W YT[
[ [#W
[ W[ [Y B W$[
[ [Y B YT[ W$[ [
W$[ YT[ [
/'
[ YT[ V W[ B
Y#[
[ V W$[ Bi
[ W$[ V YT[ W[
[ V YT[ Y#[ W$[ V
[ Bi W[ YT[ V
[ [
_
[ -. W A.;#TBX
[ Y#[ W[
Y#[
[ W[ i
[ W[ Y#[ Bm W$[
[ YT[ Bl Y#[ W$[
[ i W[ Y#[
[ 9
[ Y#[ V W[ Bl Y#[
[ V W[ Bl
[ W[ V YT[
W[
[ V YT[ Y#[ W [ V
[ B W$[ Y#[ V
[ BV W N#BX
[ YT[ [W YT[
[ [W
[ W[ [Y B W$[
[ [Y B
/Y#[2(W$[ [
cW$[23YT[ [
/'
[3YT[2V(W[B9YT[3
[2V(W$[B
[3W[3VYT[cW[2
[3VYT[19Y#[2W[3V
[BcW[33Y#[3V
[2];\_
56
Box 3-14 (Univariate (quadratic) polynomial obtained from Grbner basis solution of GPS
pseudo-range equations continued):
[ [ TBX [ Y#[ Y#[ [ Bt [ V Y#[ #BX [ YT[ W$[ 8 YT[ [ W$[ 8 [ W[ YT[ B W$[ [ Y#[ B
YT[ W
[
[ m
W$[ Y#[

[ 9
[ Y#[ V
W[ BS Y#
[
[ V W$[ BS

[ W$[ V Y#[ m W$[

[ V Y#[ m
Y#[ W[ V
[ BS W$[ YT
[ V
[
YT[ V
[ #BX
[ YT[ W[ Y#[
[ W[
[ W[ Y#[ Bp W$[
[ YT[ B YT[ W[
[ W$[ YT[
[ /'
[ YT[ V W[ B
YT[
[ V W[ BD
[ W$[ V Y#[ l W[
[ V YT[ S Y#[ W[ V
[ BD W$[ Y#[ V
[ /'
[ Y#[ B Y#[
[ SV
[ TBX
[ Y#[ W$[
9YT[
[3W$[i
[W$[2_YT[RBlcW[2
[3_Y#[RBl9YT[W[3
[XzOW[22YT[2
[2 /'
[3YT[2V(W[ Bm9YT[
[2V(W[Bd
[W[3V2Y#[X
cW$[
[3V2Y#[=Q9Y#[2W[3V
[PBcW[23Y#[3V
[#_
[PBcW
B W
TBX
[3Y#[3W$[=Q/Y#[
[2W$[Q
[2W[3_Y#[PB
cW$[
[3_YT[Bm9YT[W[N
[zOW[3YT[2
[/'
[3YT[2V(W[Bl9YT[
[3V(W[Bl
[W[NV2Y#[OW[2
[2V2YT[z9YT[W$[2V
[B
cW$[3Y#[3V
[CBV(W
TBX
[2_YT[&D9Y#[3
[B
[3V2Y#[O#BX
[3YT[2W$[+D9YT[
[3W$[+D
[2W[3_Y#[BOW[2
[2LY#[B
9YT[W[3
[ZmcW$[3Y#[3
[ 9
[3Y#[3V(W[BS9Y#[2
[3VNW$[ BS
[W$[2V2Y#[mcW$[
[3V2Y#[m9Y#[2W[3V
[ BScW$[3YT[2V
[2
YT[ V
[ #BX
[ YT[ W[ Y#[
[ W[
[ W[ Y#[ Bp W$[
[ YT[ B YT[ W[
[ W$[ YT[
[ /'
[ YT[ V W[ B
YT[
[ V W[ BD
[ W$[ V YT[ S W[
[ V Y#[ d YT[ W$[ V
[ B W[ Y#[ V
[ # 9
[ Y#[ BD Y#[
[ BD W
The algorithms for solving the unknown value ] \ (i.e. the range bias) from Boxes (3-12) or (3-14) and the respective
+(-(e
stationary receiver coordinates are Awange-Grafarend Groebner basis algorithm and Awange-Grafarend Multipolynomial Resultant algorithm and can be accessed in the GPS toolbox (http://www.ngs.noaa.gov/gps-toolbox).
The distinction between the Multipolynomial resultant approach and the approach of E. Grafarend and J. Shan (1996)
is that; for Multipolynomial resultant approach, we do not have to invert the coefficient matrix but instead use the
necessary and sufficient conditions that the determinant has to vanish if the three equations have a nontrivial solution.
We next consider an example that has been considered already by A. Kleusberg (1994) and E. Grafarend and J. Shan
(1996).
Example 3-2: From the coordinates of four GPS satellites given in Table (3.4) below, we apply the Awange-Grafarend
Groebner basis algorithm and Awange-Grafarend Multipolynomial Resultant algorithm to compute the
coordinates of the stationary GPS receiver and the receiver range bias term. The Awange-Grafarend Groebner basis algorithm and Awange-Grafarend Multipolynomial Resultant algorithm compute the coefficients
of the quadratic equations in Boxes (3-12) and (3-14) respectively as
-9.104704113943708e-1,
5.233385578536521e+7 and
-5.233405293375e+9.
VY * Y *
V
*
*
VW * W *
] ! * !
a ! * !
b ! *V !
Table 3.4: Geocentric coordinates of four GPS satellites and the pseudo-ranging observations
0
1
2
3
1.483230866e+7
-1.579985405e+7
1.98481891e+6
-1.248027319e+7
-2.046671589e+7
-1.330112917e+7
-1.186767296e+7
-2.338256053e+7
-7.42863475e+6
1.713383824e+7
2.371692013e+7
3.27847268e+6
2.4310764064e+7
2.2914600784e+7
2.0628809405e+7
2.3422377972e+7
{3];\O|
] Y * ^] \ (#]
*
Once these coefficients have been computed, the algorithms proceed to solve (using these coefficients) the roots
of the quadratic equations in Boxes (3-12) and (3-14) respectively giving the stationary receiver range bias term. The
admissible value of the stationary receiver range bias term is then substituted in the expressions
4
to give the values of stationary receiver coordinates
respectively. The complete pair
4
4
of solutions are
^] \ N#] [ * ^] \
{3] Y ]
#] [ |
]Y `
* B.%O./,._;0 %,. #]
*4_:c4R;0 % #] [ * B 4_._90 ,O.
];\ `
* BO4R4R9,90,2:% ] \
1] YZ*Q,O,,O,%O90 OB :O #]
*-BXOc.%;0 :O9, ]%[*=.c4L/,0 O.
] \ *`B ,2O90 O: ] \
From the results above, we note that the solution space is non unique. In order to decide between the correct solution
B
from the pair above, we compute the norm (length) of the positional vector
and
If
the receiver coordinates are in the Global Reference Frame, the norm of the positional vector of the receiver station
will approximate the value of the Earths radius while the norm for the other pair of solution will be in space. The
computed norms are
"
{3] [ # ]
] Y | ] \
{3]%[O#]
]^Y| ]
{3]%[O#]
]^Y| ]
B
thus clearly giving the second solution {]%[_] #]1Y| ] \
{] [ ]
#] Y | ] \ 0
* ,2%/_:O:90 O.:
\B z
*
\ :4O;0 .9,V
as the admissible solution of the receiver position.

3-232
57
Overdetermined GPS positioning with observations of type pseudo-ranges
In this section, we consider the case where more than four satellites have been observed. We will apply the GaussJacobi combinatorial algorithm to the example provided by G. Strang and K. Borre (1997) and compare the results
obtained to those obtained by the linear Gauss-Markov Model after linearization of the observations by Taylor series
expansion. Pseudo-ranges are measured to six satellites whose coordinates are also given as in Table (3.5).
] ! *z!
a ! *zN!
b ! *zVN!
Table 3.5: Geocentric coordinates of six GPS satellites and the pseudo-range observations
~
23
9
5
1
21
17
14177553.47
15097199.81
23460342.33
-8206488.95
1399988.07
6995655.48
-18814768.09
-4636088.67
-9433518.58
-18217989.14
-17563734.90
-23537808.26
12243866.38
21326706.55
8174941.25
17605231.99
19705591.18
-9927906.48
21119278.32
22527064.18
23674159.88
20951647.38
20155401.42
24222110.91
From the data above and using (2-26) in page (13), we obtain 15 possible combinations listed in Table (3.6) whose
PDOP are computed as in B. Hofmann et al (1994, pp. 249-253) below.
Table 3.6: Possible combinations and the computed PDOP
Combination Number Combination Computed PDOP
1
23-9-5-1
4.8
2
23-9-5-21
8.6
3
23-9-5-17
4.0
4
23-9-1-21
6.5
5
23-9-1-17
3.3
6
23-9-21-17
3.6
7
23-5-1-21
6.6
8
23-5-1-17
6.6
9
23-5-21-17
4.8
10
23-1-21-17
137.8
11
9-5-1-21
5.6
12
9-5-1-17
14.0
13
9-5-21-17
6.6
14
9-1-21-17
5.2
15
5-1-21-17
6.6
It is clearly seen from the computed PDOP that the 10th combination had a poor geometry. We plot the PDOP versus
the combination number to have a clear picture in Figure (3.2).
Figure (3.2) indicates clearly that the 10th combination had a weaker geometry. The Gauss-Jacobi combinatorial
algorithm takes care of this weaker geometry during the adjustment process by the use of the variance-covariance
matrix computed through nonlinear error propagation for that respective set. We next use the derived quadratic
formulae by Grbner basis in Box (3-12) or by Multipolynomial resultants in Box (3-14) for the minimal case to
compute the coefficients presented in Table (3.7).
From the computed coefficients in Table (3.7), the 10th combination is once again identified as having significantly
different values from the rest. Using the coefficients of Table (3.7), we compute the solution of the minimal combinatorial sets (each combination ) being the receiver position
and the range bias
and present them in Table
(3.8).
{|
The final adjustment is performed using the linear Gauss-Markov model with the random values of Table (3.8) as
pseudo-observations and the dispersion matrix obtained from the nonlinear error propagation (Chapter 2).
Figure (3.3) gives the plot of the scatter of the 15 Gauss-Jacobi combinatorial solutions (shown by points) around
the adjusted value (indicated by a star), while Figure (3.4) gives the magnification of the scatter of 14 Gauss-Jacobi
combinatorial solutions (shown by points) that are very close to the adjusted value (indicated by a star) ignoring the
outlying point in Figure (3.3).
58
Plot of Position Dilution of Precision against combinatorial No.

140
120
100
PDOP
80
60
40
20
10
15
Combinatorial No.
Figure 3.2: Computed PDOP for respective combinations
Table 3.7: Computed coefficients of the combinations

C/No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
-0.914220949236445
-0.934176403102736
-0.921130625833683
-0.865060899130107
-0.922335616484969
-0.919296962706157
-0.894980063579044
-0.917233949644576
-0.925853049262193
3369.83293928593
-0.877892756651551
-0.942581538318523
-0.908215141659006
-0.883364070549387
-0.866750765656126
VY
52374122.9848733
50396827.4998945
51741826.0147786
54950460.2842167
51877166.0451888
51562232.9601199
53302005.6927825
52194946.1124139
51140847.6331213
-1792713339.80277
54023883.5656926
50793361.5303674
52246642.0794924
53566554.3869961
54380648.2092251
VW
49022682.3125
7915541824.84375
343282824.25
-10201105114.5
280298481.625
1354267366.4375
-3642644147.5625
132408747.46875
3726719112.1875
6251615074927.06
-6514735288.13762
784684294.241371
-2499054749.05572
-5481411035.37882
-7320871488.80859
V
Table 3.8: Computed combinatorial solution points in a polyhedron

C/No.
1
596925.3485 -4847817.3618 4088206.7822
-0.9360
2
596790.3124 -4847765.7637 4088115.7092 -157.0638
3
596920.4198 -4847815.4785 4088203.4581
-6.6345
4
596972.8261 -4847933.4365 4088412.0909 185.6424
5
596924.2118 -4847814.5827 4088201.8667
-5.4031
6
596859.9715 -4847829.7585 4088228.8277
-26.2647
7
596973.5779 -4847762.4719 4088399.8670
68.3398
8
596924.2341 -4847818.6302 4088202.3205
-2.5368
9
596858.7650 -4847764.5341 4088221.8468
-72.8716
10
596951.5275 -4852779.5675 4088758.6420 3510.4002
11
597004.7562 -4847965.2225 4088300.6135 120.5901
12
596915.8657 -4847799.7045 4088195.5770
-15.4486
13
596948.5619 -4847912.9549 4088252.1599
47.8319
597013.7194 -4847974.1452 4088269.3206 102.3292
14
15
597013.1300 -4848019.6766 4088273.9565 134.6230
d

=
59
3dplot of the scatter of the 15combinatorial solutions around the adjusted value
1000
(Z+4088000)m
500
500
1000
1500
0.7
0.8
7050
0.9
7000
1
x 10
6950
6900
1.1
6850
1.2
6800
1.3
(Y4840000)m
6750
(X+590000)m
Figure 3.3: Scatter of the 15 Gauss-Jacobi combinatorial solutions ( ) around the adjusted value ( ).
3dplot of the scatter of the 14combinatorial solutions around the adjusted value
500
(Z+4088000)m
400
300
200
100
0
7750
7800
7050
7850
7000
7900
6950
6900
7950
6850
8000
(Y4840000)m
6800
8050
6750
(X+590000)m
Figure 3.4: Magnification of the scatter of 14 Gauss-Jacobi combinatorial solutions ( ) around the adjusted value ( ).
3-233
Linearized least squares versus Gauss-Jacobi combinatorial algorithm
We conclude in this section by comparing between the linearized least squares approach and the Gauss-Jacobi combinatorial approach. Using the Gauss-Jacobi combinatorial approach, the stationary receiver position and the stationary
receiver range bias are computed as discussed in Section (3-232). For the linearized least squares approach, the nonlinear observation equations (3-118) are linearized using Taylor series expansion for the satellites in Table (3.5) to
generate the Jacobi matrix required for the linearized least squares approach. As approximate starting values, we
first initiated the stationary receiver position and the stationary receiver range bias as zero and set the convergence
as the difference of values between two successive iterations. With approximate values taken as
limit to
In the second case, we considered the values of the
zero, 6 iterations were required for convergence limit
,"km,2
,Pk,2;0
60
Gauss-Jacobi combinatorial algorithm as the starting (approximate) values for the linearized least squares solution.
This time round, only two iterations were required to achieve convergence.
From the nonlinear equations (3-118) and the results of both linearized least squares approach and Gauss-Jacobi
combinatorial algorithm, we computed the residuals and obtained the squares of these residuals and finally the error
norm from

! BD' ] Y BD!
z]
B(!
] [ BDVN!'
B] \ )
0
(3-129)
!
Y
where {] Y ] ] [ ] \ | are the computed values of the stationary receiver position and the stationary receiver range
bias, {!T$N!#$VN!|^)*A{c,203030:9| are the coordinates of the six satellites in Table (3.5) and { !|^)*-{c,203030:9| are
*

the measured pseudo-ranges.
Table (3.9) gives the obtained results from the Gauss-Jacobi combinatorial algorithm and those obtained from linearized Least squares approach by first linearizing using Taylor series expansion. Table (3.10) presents the rootmean-square-errors and the differences between the solutions of the two procedures. In Table (3.11), we present the
computed residuals, sum of squares of these residuals and the computed error norm from (3-129). The computed error
norm are identical for both procedures.
Further comparison of the two procedures is to be found in Chapter (5) where the procedures are used to compute
the 7-datum parameter transformation. Once the parameters have been computed, they are used to transform the
Cartesian coordinates from the Local Reference System (Table 5.5) to the Global Reference System (WGS 84, Table
5.6) as shown in Tables (5.10) and Table (5.11). The residuals from both Gauss-Jacobi combinatorial algorithm and
Linearized Least Squares Solution are in the same range in magnitude. We also compute the residual norm (square
root of the sum of squares of residuals) and present them in Table (5.12) in page (93). In this case, the error norm from
the Gauss-Jacobi combinatorial algorithm is somewhat better than those of the linearized least squares solution.
V
Table 3.9: Computed receiver position and stationary receiver range bias
GJ- Approach
(Linearized)LS
Difference
d
596929.6542
596929.6535
0.0007
x
-4847851.5021
-4847851.5526
0.0505
=
4088226.7858
4088226.7957
-0.0098
-15.5098
-15.5181
0.0083
Table 3.10: Computed root-mean-square errors and the difference in solution between the two approaches
GJ-Approach
GM solution
g
6.4968
34.3769
g
11.0141
58.2787
g
5.4789
28.9909

8.8071
46.6018
Table 3.11: Computed residuals, squares of residuals and error norm

PRN
Gauss-Jacobi (m) Linearized Least Squares (m)
23
-16.6260
-16.6545
9
-1.3122
-1.3106
5
2.2215
2.2189
1
-16.4369
-16.4675
21
26.8623
26.8311
17
5.4074
5.3825
Sum of Squares
1304.0713
1304.0680
Error norm
36.1119
36.1119
3-234 Outlier diagnosis

In this section, we demonstrate the capability of the Gauss-Jacobi combinatorial algorithm to diagnose outliers. Consider that a satellite signal meant to travel straight from the satellite to the receiver is reflected by the reflecting surfaces
(multipath effect) in built up areas for instance. The measured pseudorange reaching the receiver ends up being longer
than the actual would be pseudorange. The scanned Figure from B. Hoffman et al. (1994, p.124, Figure 6.5) illustrates
the case of multipath.
61
Let us now suppose that satellite number 23 had its pseudorange measurement longer by 500m owing to multipath
effect. This particular satellite is chosen as it is the first in the combinatorial formation and as such the left most
element of the combination and thus easy to identify in the combinatorial list of Table (3.12). For each combination,
we compute the position by using either the Grbner basis or Multipolynomial resultant algebraic tools. Once the
positions have been obtained, the positional norm are then computed for each combination and written besides the
combinations as in Table (3.12).
Table 3.12: Computed positional norm
Combination Number Combination Positional norm (km)
1
23-9-5-1
6368.126
2
23-9-5-21
6367.147
3
23-9-5-17
6368.387
4
23-9-1-21
6370.117
5
23-9-1-17
6368.474
6
23-9-21-17
6368.638
7
23-5-1-21
6368.894
8
23-5-1-17
6368.256
9
23-5-21-17
6368.005
10
23-1-21-17
6398.053
11
9-5-1-21
6369.723
12
9-5-1-17
6369.522
13
9-5-21-17
6369.647
14
9-1-21-17
6369.711
15
5-1-21-17
6369.749
It is clearly seen that the computed positional norms of the first 10 combinatorials were varying while the variation of
the computed positional norm of the last 5 combinatorial was to a lesser degree. The possible explanation could be that
satellite number 23 whose pseudorange is contaminated appeared in the first 10 combinations. The last 5 combinations
are not affected with satellite number 23. In order to view the variation of the computed positional norm, we have
computed the mean of these positional norms and subtracted it from the other norms. In Figure (3.5), we have plotted
these deviations of the computed positional norm from the mean positional norm. Whereas the deviations of the
first 10 combinations are seen to fluctuate, those of the last 5 combinations are seen to be almost constant and with
minimum deviation from the mean thus clearly indicating the presence of outlier in the first 10 combinations. Since
62
the satellite 23 is the only common satellite in the first 10 combinations, this outlier could be attributed to it.
4
x 10
Computed position norm for respective combinatorials
Computed positional norm mean value(m)
2.5
1.5
0.5
0.5
10
Combinatorial No.
Figure 3.5: Deviation of positional norm from the mean value
15
Chapter 4
Test network Stuttgart Central

As an extension of Chapter 3 we consider in this Chapter the closed form solution and the overdetermined solution of
the three-dimensional resection problem using the Grbner bases, Multipolynomial resultants and the Gauss-Jacobi
combinatorial algorithms discussed in Chapter 2. The test network Stuttgart Central in Figure (4.1) below is
selected for study. In Section (4-1) we consider the observations of the test network Stuttgart Central of types GPS
coordinates, horizontal directions and vertical directions that will be used in subsequent sections. Section (4-2)
considers the closed form three-dimensional resection solution of the test network Stuttgart Central while Section
(4-3) considers the overdetermined three-dimensional resection solution of the test network Stuttgart Central.
Haumannstr. (1324.238 m)
Schloplatz (566.864 m)
Dach FH (269.231 m)
K1
Dach LVM (400.584 m)

Eduardpfeiffer
(542.261 m)
Liederhalle (430.529 m)
Lindenmuseum (364.980 m)
Figure 4.1: Graph of the Test network Stuttgart Central
63
CHAPTER 4. TEST NETWORK STUTTGART CENTRAL
64
4-1
Observations
The following experiment was performed at the centre of Stuttgart on one of the pillars of the University buildings
along Kepler Strasse 11 as depicted by Figure (4.1). The test-network "Stuttgart Central" consisted of 8 GPS points
listed in Table (4.1). A theodolite has been stationed at pillar K1 whose astronomical longitude ! as well as astronomic latitude "! were known from previous astrogeodetic observations made by the Department of Geodesy
and GeoInformatics, Stuttgart University. Since theodolite observations of type horizontal directions as well as
vertical directions from the pillar K1 to the target points #%$&#('*)+$-,.$ /0/ /0$213$%4.$ were only partially available we
decided to simulate the horizontal and vertical directions from the given values of 56 $7" 98 as well as the Cartesian
coordinates of the station point :<;=$?>@$%AB and target points :C;EDF$G> D $%A D B using (3-18) and (3-19). The relationship between the observations of type horizontal directions , vertical directions , values of 5 $%" 98 and the Cartesian
coordinates of the station point :<;($G>@$%A9B and target points :C;EDF$G> D $7A D B that enabled generation of the observation
data sets 1 to 11 is presented in Chapter 3, Section (3-14). Such a procedure had also the advantage that we had
full control of the algorithms that we discussed in Chapter 2. In detail, the directional parameters 5 $%" 98 of the
local gravity vector were adopted from the astrogeodetic observations reported by S. Kurz (1996 p.46) with a rootmean-square error HJIK'LH3MN'O)PRQ . Table (4.1) contains the :<;=$?>@$%AB coordinates obtained from a GPS survey of
the test-network "Stuttgart Central", in particular with root-mean-square errors :CHJS&$-H.TU$2HWV.B neglecting the covariances :XHJSTU$2HJTV9$-H3V6SYB . The spherical coordinates of the relative position vector, namely of the coordinate differences
:Z D9[ Z9$2\ D [ \$?] D [ ]6B , are called horizontal directions , vertical directions and distances ^ and are given by
Table (4.2). The standard deviations/root-mean-square errors were fixed to HW_`'N1+QJ$2HWab'K1+Q . Such root mean square
errors can be obtained on the basis of a proper refraction model. Since the horizontal and vertical directions of Table
(4.2) were simulated data, with zero noise level, we used a random generator randn in MATLAB version 5.3 (e.g. D.
Hanselman and B. Littlefield 1997, pp. 84, 144) to produce additional observational data sets within the framework of
the given root-mean-square errors. For each observable of type and c , 30 randomly simulated data were obtained
and the mean taken. Let us refer to the observational data sets 50?$-c 8 $?#d'e)+$-,.$/ /0/ $21J$74$ of Table (4.3) to Table
(4.13) which were enriched by the root-mean-square errors of the individual randomly generated observations as well
as by the differences f(@gh'i [ 2:CjRk lJk0m%npo-k qJB7$frcsgt'* [ -:<jRk lJk0m%npo-k qJB . Such differences :Xf(G$%f(uB
indicate the difference between the ideal values of Table (4.2) and those randomly generated.
ervations are thus designed such that by observing the other seven GPS stations, the orientation of the Local Level
Reference Frame vdw whose origin is station K1, to the Global Reference Frame v
x is obtained. The relationship
between the vyw Reference Frame and the vyx Reference Frame is presented in Chapter 3, Section (3-13). The direction of Schlossplatz is chosen as the zero direction of the theodolite and this leads to the determination of the
third component z! of the three-dimensional orientation parameters. To each of the GPS target points # , the observations of the type horizontal directions and the vertical directions are measured. The spatial distances
^| { :X}`$2}~FB'}s [ } are readily obtained from the observation of type horizontal directions and vertical directions using the algebraic computational techniques discussed in Chapter 2. Once we have Euclidean distances
^ computed from the observations of type horizontal directions and vertical directions c , a foward computation
using any of the three-dimensional ranging ( "Bogenschnitt") procedures discussed in Section (3-21) is used to compute the coordinates 5 $-$7 86 from which the direction parameters ( $7" ) of the local gravity vector at K1
and the "orientation unknown" element z are finally computed as discussed in J. L. Awange (1999) and E. Grafarend
and J. L. Awange (2000). The obtained values are then compared to the starting values. The following symbols have
been used: H S $-H T $-H V are the standard errors of the GPS Cartesian coordinates. Covariances H ST $2H TV $2H V6S were
neglected. H3_!$2H3a are the standard deviation of horizontal and vertical directions respectively after an adjustment,
f(_$-fra the magnitude of the noise on the horizontal and vertical directions, respectively.
Table 4.1: GPS Coordinates in the Global Reference Frame vyxR:C$2!$%cB ,
Station
b:CB
s:CB
d:CB
Dach K1
1
2
3
4
5
6
7
4157066.1116
4157246.5346
4156749.5977
4156748.6829
4157066.8851
4157266.6181
4157307.5147
4157244.9515
671429.6655
671877.0281
672711.4554
671171.9385
671064.9381
671099.1577
671171.7006
671338.5915
4774879.3704
4774581.6314
4774981.5459
4775235.5483
4774865.8238
4774689.8536
4774690.5691
4774699.9070
WH
~
:<?$-W2$%YXB$#|')+$-,.$/ /0/ $%4

H3
H3
1.07
0.76
1.77
1.93
1.38
1.29
0.20
2.80
1.06
0.76
1.59
1.84
1.29
1.28
0.10
1.50
1.09
0.76
1.61
1.87
1.38
1.34
0.30
3.10
4-1. OBSERVATIONS
65
Table 4.2: Ideal spherical coordinates of the relative position vector in the Local Horizontal Reference Frame
Spatial distances, horizontal directions, vertical directions
Station Observed
Distances
Horizontal
Vertical
from K1
(m)
directions(gon) directions(gon)
Schlossplatz (1)
566.8635
52.320062
-6.705164
Haussmanstr. (2)
1324.2380
107.160333
0.271038
Eduardpfeiffer (3)
542.2609
224.582723
4.036011
Lindenmuseum (4) 364.9797
293.965493
-8.398004
Liederhalle (5)
430.5286
336.851237
-6.941728
Dach LVM (6)
400.5837
347.702846
-1.921509
Dach FH (7)
269.2309
370.832476
-6.686951
vyw :
Table 4.3: Randomly generated spherical coordinates of the relative position vector: horizontal directions and
vertical directions c?$?#')+$-,J$/ /0/ $21J$74$ root-mean-square errors of individual observations, differences f(gh'
[ ?:CjRk0l3k0m%npo2kqJB7$frcYgt'c [ c-:<j+k0lJk m-no2k qJB with respect to :<G$-cuB ideal data of Table 4.2, first data set: set 1
St. H/Dir.(gon) V/Dir.(gon)
H _ :<6B H a :<6B f _ :<6B f a :<6B
1
0.000000
-6.705138
0.0025794 0.0024898 -0.000228 -0.000039
2
54.840342
0.271005
0.0028756 0.0027171 -0.000298 0.000033
3 172.262141
4.035491
0.0023303 0.0022050 0.000293
0.000520
4 241.644854
-8.398175
0.0025255 0.0024874 0.000350
0.000171
5 284.531189
-6.942558
0.0020781 0.0022399 -0.000024 0.000830
6 295.382909
-1.921008
0.0029555 0.0024234 0.000278 -0.000275
7 318.512158
-6.687226
0.0026747 0.0024193 -0.000352 0.000500
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.0000000
54.841828
172.262016
241.645929
284.531106
295.382535
318.512615
Table 4.4: Second data set: set 2

V/Dir.(gon)
H _ :<6B H a :<6B
-6.705636
0.0029467 0.0022479
0.270494
0.0023740 0.0018085
4.035712
0.0025738 0.0025891
-8.397520
0.0025012 0.0027585
-6.940833
0.0025388 0.0021120
-1.920744
0.0024122 0.0022379
-6.686485
0.0024235 0.0027708
f _ :<6B
f a :<6B
0.000655
-0.000902
0.001300
0.000156
0.000723
0.000904
0.000453
0.000459
0.000544
0.000300
-0.000484
-0.000895
-0.000765
-0.000466
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.839743
172.261715
241.645032
284.530697
295.382921
318.511249
Table 4.5: Third data set: set 3

V/Dir.(gon)
H _ :<6B
H a :<6B
-6.704849
0.0025794 0.0023074
0.272062
0.0027316 0.0036383
4.036063
0.0022680 0.0025318
-8.398128
0.0031452 0.0030835
-6.941783
0.0025214 0.0024290
-1.922053
0.0024296 0.0028454
-6.686536
0.0024345 0.00227063
f _ :<6B
f a :<6B
-0.001470
0.000524
0.000942
0.000395
0.000473
-0.000141
0.001174
-0.000329
-0.001024
-0.000052
0.000124
0.000055
0.000544
-0.000415
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.841145
172.264284
241.645972
284.532505
295.384465
318.513839
Table 4.6: Fourth data set: set 4

V/Dir.(gon)
H _ :<6B H a :<6B
-6.704682
0.0023308 0.0027599
0.271960
0.0026907 0.0021463
4.036170
0.0028699 0.0024486
-8.397035
0.0035089 0.0024921
-6.941248
0.0026110 0.0033665
-1.921296
0.0027294 0.0026283
-6.686125
0.0020477 0.0030185
f _ :<6B
f a :<6B
0.000862
-0.000011
-0.000760
0.000322
-0.000468
-0.000818
-0.000562
-0.000496
-0.000922
-0.000159
-0.000969
-0.000480
-0.000213
-0.000826
66
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.839952
172.262789
241.645827
284.530609
295.383197
318.513393
Table 4.7: Fifth data set: set 5

H3_!:.6B H3ac:.6B
V/Dir.(gon)
-6.705407
0.0023550 0.0026607
0.271814
0.0027139 0.0024570
4.036099
0.0028628 0.0020811
-8.398001
0.0027261 0.0027714
-6.940954
0.0029166 0.0024115
-1.921506
0.0032741 0.0025684
-6.686562
0.0031545 0.0028330
fr_!:.6B
f(ac:.6B
0.000275
0.000594
0.000148
-0.000121
0.000840
-0.000138
-0.000705
0.000229
-0.000775
-0.000088
-0.000003
-0.000774
-0.000003
-0.000993
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.841100
172.262254
241.645033
284.531250
295.383176
318.512147
Table 4.8: Sixth data set: set 6

V/Dir.(gon)
H _ :.6B H a :.6B
-6.705699
0.0032227 0.0026362
0.272198
0.0028716 0.0032300
4.035556
0.0027485 0.0022965
-8.398092
0.0028093 0.0030335
-6.941579
0.0022418 0.0023971
-1.921632
0.0028193 0.0031391
-6.687006
0.0026446 0.0018992
f _ :.6B
f a :.6B
-0.000230
-0.001059
0.000177
0.000167
-0.000306
-0.000622
0.000037
0.000522
-0.001160
0.000455
0.000088
-0.000149
0.000123
0.000055
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.840622
172.262586
241.645766
284.533069
295.383199
318.512078
Table 4.9: seventh data set: set 7

V/Dir.(gon)
H _ :.6B H a :.6B
-6.704710
0.0032501 0.0021664
0.271205
0.0025500 0.0026468
4.035479
0.0028646 0.0030243
-8.397192
0.0020303 0.0026158
-6.940859
0.0026240 0.0022506
-1.920591
0.0029904 0.0026217
-6.686979
0.0024550 0.0023116
f _ :.6B
f a :.6B
0.000796
0.000446
0.000872
0.000461
-0.001098
0.000381
0.001132
-0.000467
-0.000167
0.000532
-0.000811
-0.000869
-0.000918
0.000028
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.841233
172.263880
241.645783
284.532564
295.383289
318.514380
Table 4.10: Eigth data set: set 8

H3_!:.6B H3ac:.6B
V/Dir.(gon)
-6.705117
0.0019401 0.0025817
0.271160
0.0020984 0.0028927
4.036182
0.0026151 0.0022965
-8.397963
0.0029220 0.0022676
-6.941989
0.0024886 0.0021962
-1.920585
0.0025328 0.0025163
-6.686908
0.0028717 0.0028983
fr_!:.6B
f(ac:.6B
0.001199
0.000238
-0.000019
0.000847
-0.000188
0.000694
-0.000767
-0.000060
-0.000122
-0.000170
-0.000041
0.000261
-0.000924
-0.000043
4-2. CLOSED FORM SOLUTION
4-2
67
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.838686
172.261443
241.645374
284.530552
295.381778
318.511475
Table 4.11: Ninth data set: set 9

HW_!:<6B HWac:<6B
V/Dir.(gon)
-6.705021
0.0023099 0.0029693
0.271061
0.0025460 0.0024923
4.035575
0.0027183 0.0026865
-8.397707
0.0024564 0.0024467
-6.941628
0.0034024 0.0027446
-1.921467
0.0022630 0.0027665
-6.686698
0.0022266 0.0025736
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.841489
172.263665
241.645336
284.531567
295.383055
318.512017
Table 4.12: Tenth data set: set 10

HW_!:<6B HWac:<6B
V/Dir.(gon)
-6.705515
0.0024938 0.0032987
0.270932
0.0029717 0.0022950
4.036147
0.0032672 0.0024499
-8.397823
0.0028515 0.0025473
-6.941534
0.0022931 0.0021688
-1.922041
0.0033986 0.0028467
-6.686773
0.0024359 0.0021356
St.
1
2
3
4
5
6
7
H/Dir.(gon)
0.000000
54.840818
172.263416
241.645322
284.532013
295.383571
318.513029
Table 4.13: Eleventh data set: set 11

V/Dir.(gon)
H _ :<6B H a :<6B
-6.704889
0.0024962 0.0031604
0.271340
0.0027559 0.0028895
4.035779
0.0023929 0.0032068
-8.398136
0.0031984 0.0019623
-6.942079
0.0027289 0.0032386
-1.921888
0.0026898 0.0023682
-6.686424
0.0027481 0.0026191
fr_!:<6B
frac:<6B
-0.000872
0.000714
0.000347
-0.000815
-0.000249
0.000134
0.000068
-0.000156
-0.000023
0.000436
-0.000296
-0.000100
-0.000042
-0.000253
fr_!:<6B
frac:<6B
0.000299
-0.000918
-0.000704
0.000395
-0.000093
0.000028
0.000696
0.000338
0.000106
-0.000136
-0.000181
-0.000194
0.000532
-0.000178
f _ :<6B
f a :<6B
0.000459
-0.000088
-0.000296
0.000568
-0.000379
-0.000328
-0.000157
-0.000288
0.001659
0.000232
0.000132
0.000351
0.000379
-0.000527
Closed form solution
In order to position point K1 in the GPS network of Stuttgart Central using LPS observable of types horizontal
directions and vertical directions c , three known stations (Haussmanstr., Eduardpfeiffer and Liederhalle) of the
test network Stuttgart Central in Figure (4.1) are used. We proceed in three steps: The first step considers the
computation of the spatial distances, the second step is the computation of the coordinates of the unknown station,
and the final step is the computation of the threedimensional orientation parameters. In this section, the threedimensional resection method is considered with the aim of providing the threedimensional geocentric GPS coordinates in
the Global Reference Frame. The threedimensional orientation parameters of type astronomical longitude , astronomical latitude " , and the orientation unknown z in the horizontal plane and the deflection of the vertical can
be obtained as in J. L. Awange (1999) and E. Grafarend and J. L. Awange (2000).
The solution of Grunert equations is achieved using the algebraic computational techniques i.e. Grbner bases or
Multipolynomial resultant, the position-derivation step involves computing the desired threedimensional GPS Cartesian coordinates 5 $2!$% 8 of the unknown point ed in the Global Reference Frame. This is achieved by
analytically solving the three-dimensional ranging problem (also known in German literature as "dreidimensionales
Bogenschnitt") as discussed in Section (3-21).
4-21 Experiment
Known GPS stations (Haussmanstr., Eduardpfeiffer and Liederhalle) of the test network Stuttgart Central in Figure
(4.1) together with K1 form the tetrahedron 5y|% 8 in the Appendix A.2 in page (104). Algebraic computational
{
tools Grbner bases or Multipolynomial resultants discussed in Chapter 2 are used to determine the distances of the
tetrahedron.
68
Using the computed univariate polynomial (element of Grbner basis of the Ideal subset R[x p$? $? ) in Box (3-3a)
{
in Section (3-21, page 36), we determine the distances ^ ' $ b#'5+)R$%,.$- 8 etwen the unknown
station d and the known stations d expressed in (3-31) for the test network "Stuttgart Central" in
Figure (4.1). The unknown point in this case is the pillar K1 on top of the University building at Kepler Strasse 11.
Points , ?3Gp.36+6656y 8 in Figure (3.1) correspond to the chosen known GPS stations Haussmannstr.,
{
Eduardpfeiffer, and Liederhalle. The distance from K1 to Haussmannstr. is designated ^ ' $ K1 to
Eduardpfeiffer ^
{ ' { $ while that of K1 to Liederhalle is designated ^ ' / The distances
between the known stations 5^ $^ $W^ 8 are computed from their respective GPS coordinates as indicated
{ {2
in Box (4-1) below. Their corresponding space angles
{0 {- are computed from (3-29). In order to control
the computations, the Cartesian GPS coordinates of point K1 are also known. Box (4-1) below gives the complete
solution of the unknowns 5 $2 $? 8 from the computed Grbner basis of Boxes (3-3a) and (3-3b) in Section
{
(3-21, pages 36 and 37 respectively). The univariate polynomial in has eight roots, four of which are complex and
four real. Of the four real roots two are positives and two are negative. The desired distance is thus chosen
from the two positive roots with the help of prior information and substituted in - in Box (3-3b) in page (37) to give
two solutions of 9 , one of which is positive. Finally the obtained values of 596$2 8 are substituted in R Box
(3-3b) in page (37) to obtain the remaining indeterminate . Using this procedure, we have in Box (4-1) below that
{
^ '5+RP3/ +,+1J$0)63/t4.)R) , 8 . Since ^ ' $ from apriori information we choose ^ '++PJ/tR,RR1 , leading to
^')+,RW/ ,RR3) , and ^ 'R,J/ ,R1RP+ . These values compare well with their real values depicted in Figure (4.1).
Box 4-1 (computation of distances for test network "Stuttgart Central"):

Using
the
entries
of
Table (4.1) in page (64), we computed inter-station distances by pythagorus
:<( [ FB {! :CJ [ XB { :u [ YXB { and spatial angles from (3-29) are given as
^h'
^ { ')R1RP3/ +RP+,

')R/ %R1+,RP
^ {-
' 4RR./ R1+J) {2{ ')R/h4p1+RR+
^ ')64.) 3/)PRRPR
',./ 1R14
and substituted in (3-47) to compute the terms 56 $% $2 $2R$%%R$-7 8 which are needed to compute the
{ {-
coefficients of the Grbner basis element in Box (3-3a) in Section (3-21, page 36). Expressing the univariate
polynomial in Box (3-3a) in Section (3-21, page 36) as c0 c0 U W{ 'NPJ$ the computed
{
coefficients are

' J RR++,+,R,1+14PR1+,J)0k P+,R

d
,
' [ , +PR1+%4.)641+.)P+R+46k PJ)
'))%P ,,RR+,1,+4.) pk PJ)
(
1 =' [ . P+R+PJ)64RR, ,R++Pk PRP+
='J ,+R1R3)64R,R+1+PR3) pk [ PRP.
The solution to the univariate polynomial equation is then obtained from the Matlab command "roots" (e.g. D.
Hanselman and B. Littlefield 1997, p.146) as
' ccEcE {
9
N
' 6p6?:X0B
the other coefficients being zero. The obtained values of
[ , P4R+4Rt,R+P4%+4,
[ ,P4R+4Rt,R+P4%+4,
,,PP4R4R++4R4Rtt,R,R++PP44%% ++44,, [
' +RPtR,R+4pJ)PR+1%
[ +RPtR,R+4pJ)PR +%1

)h4)+),+,R,+4P++,+
[ )h4)+),+,R,+4P++,+
where the chosen value 430.5286 of
to give
are
R 1,1 +,1+,4RR+R
[ R1,1 +,1+,4RR+R
R1,1 +,1+,4RR+
R1,1 +,1+,4RR+
using prior information is substituted in - in Box (3-3b) in page (37)

'5 [ ,PR+)6R,+4pP4+$0) ,R.t,+RPRJ) +J) 8
and finally the values of 5 6$? 8 are substituted in R in Box (3-3b) in page (37) to give 'R,t,1RP4p14R4pP+R,
4-2. CLOSED FORM SOLUTION
69
The computation procedure using the B. Buchberger algorithm (Grbner bases) is summarized as follows:
Arrange the given polynomial equations using a chosen monomial order as in (3-46)
Determine the polynomial Ideal as in (3-48)
Compute the Grbner basis of this Ideal using either Mathematica or Maple softwares.
From the computed Grbner basis of the Ideal, solve the univariate polynomial for the desired roots using the
roots command of MATLAB.
Substitute the admissible value of the univariate polynomial solution in the other Grbner basis elements to
obtain the remaining variables.
Alternatively, the approach below based on Multipolynomial resultants technique can be used to solve the Grunert
equations.
(a)
The
F. Macaulay (1902) approach discussed in Section (2-322) solves for the determinant of the matrix
leading
to a univariate polynomial in . The solution of the obtained univariate polynomial equation
expressed in Box (3-3c) in page (39) leads to similar results as those of Grbner basis i.e.
:XcB'
'
E'
9'
(b)
'
{ '
{ W{ E
[ 3 4.)RR4pR+PR1R,+,p{ $J { 'D3h4J)6RR.4)64PR{
Uc' [ )R)J) P+4RR1+P+RP.)64
[ PJ PRP+%P +,+R1R3)64p%, 4R4pR.$cc'D+R,RR3 .4,P+4R1R
[ ,R%, 13 R3) PR4 , )p4p+J ,R4P+,R4 RPR1p#
[ ,R%, 13 R3) PR4 , [ )p4p+J ,R4P+,R4 RPR1p#
,R%, 13 R3) PR4 %, + )p4p+J ,R4P+,R4 +P1R#
,R%, 13 R3) PR4 %, + [ )p4p+J ,R4P+,R4 +P1R#
)RRPJ) P++%, +4R44
[ )RRPJ) P++%, +4R+4+4
) R, 3t,+RP+% J) %
[ ) R, 3t,+RP+% J) R
++PJtR,+4J) P++R1J$ [ ,46RJ +R
P ,R41R1RR1
,.t,1RP4p14R4pP+R,pJ$ [ 4)+)R RP+PR% .4)PRR++4
R
The B. Sturmfels (1998) approach discussed in Section (2-322) using the Jacobian determinant solves the
determinant of the 11 matrix and leads to a univariate polynomial in . The solution of the obtained
univariate polynomial equation expressed in Box (3-3d) in page (41) leads to similar results as those of
Grbner basis i.e.
:CB'
E'
'
'
'
{ '
{ W{ U
[ )+ ++1J) ++J) +,R,RR+{ $W { ')R RR,1,.) +R1+%+RR{
' [ R%, +R+P+,+R, ,RPPR14
[ P3 P+PJ)p4,+% .4,1RR+R4 +1.$ ')p4W)R) +J R++,R,R4+4
[ ,+R, 1J ++J)P4PRR1 )64p.t,4pP+,+++R #
[ ,+R, 1J ++J)P4PRR1 [ )64p.t,4pP+,+++R #
,+R, 1J ++J)P4RP%1 )64+.t,4pP,R++ #
,+R, 1J ++J)P4RP%1 [ )64+.t,4pP,R++ #
)6+PJ) PRR, 4p++4R4
[ )6+PJ) PRR, 4pR44
)+%, 3t,R+PR .)R
[ )+%, 3t,R+PR .) % %
RP3 +,+4p3) PRp+.$ [ ,R4J %
P ,+46+11R+1
,J ,R1RP414+4pPR+RR, 3$ [ 4)R)+ +PRP+% 4.) P+R+
%
70
The computed distances from F. Macaulay (1902) and B. Sturmfels (1998) above tallies. The required solutions
5 9p$J { $J 8 obtained from Grbner basis computation and those of Multipolynomial resultants are the same {i.e
) +,%3t,+J)R$R,.t,1RP+J$+RP3 +,+1 } respectively.
For each observational data sets 0-11 in Tables (4.2)-(4.13), distances are obtained using either Grbner basis or
Multipolynomial resultant techniques as illustrated above. The results of the computed distances for the closed form
solutions are presented in Table (4.14) of Section (4-41). The computed distances are used with the help of Grbner
basis approach to determine the position of K1 for each observational data set 0-11in Tables (4.2)-(4.13) as discussed
in Section (3-21, page 43). The results for the computed position of K1 for the closed form approach are presented
in Section (4-42). For the orientation elements we refer to J. L. Awange (1999) and E. Grafarend and J. L. Awange
(2000).
4-3
Overdetermined solution
In the preceding sections, only three known points were required to solve the closed form three-dimension resection
problem for position and orientation of the unknown point K1. If superfluous observations are available made possible by the availability of several known points as in the case of the test network Stuttgart Central, the closed
form three-dimensional resection procedure gives way to the overdetermined three-dimension resection case. In this
case therefore, all the known GPS network stations (Haussmanstr., Eduardpfeiffer, Lindenmuseum, Liederhalle, Dach
LVM, Dach FH, and Schlossplatz) of the test network Stuttgart Central in Figure (4.1) are used.
4-31 Experiment
Using the observation data in Tables (4.2)-(4.13), we proceed in six steps as follows:
Step 1 (construction of minimal combinatorial subsets for determination of distances):
From (2-26) in page (13 ), 35 minimal combinatorials are formed for the test network Stuttgart Central and
are as presented by the 35 combinatorial simplexes in Appendix A.2 (page 104). For each minimal combinatorial
simplex, the distances are computed from the univariate polynomials obtained using either Grbner basis or
Multipolynomial resultants algorithms and presented in Boxes (3-3a, page 36), (3-3c, page 39) or (3-3d, page
41) in Chapter 3. The admissible solution from the univariate polynomials are substituted into other polynomials
(as explained in Section 3-21 in page 31 and Section 4-21 in page 67) to get the remaining two distances. Each
combinatorial minimal subset results in distances thus giving rise to a total of :XsRB)P+ distances which
we consider in the subsequent steps as pseudo-observations. The computed distances ^ link the known points
- #')R$ 0$74 to the unknown point (K1) in Figure (4.1).
Step 2 (nonlinear error propagation to determine the dispersion matrix ):

In this step, the dispersion matrix is sought. This is achieved via the Error propagation law/variancecovariance propagation for each of the combinatorial set 'L)R$ 0 $2+ above. The closed form observational
equations for the first combinatorial subset b'*) (tetrahedron y ) in Appendix A.2 are written alge{
braically with (3-30) as
gh'^ { ^ { { [ ,R^ ^ { :C { :C { [ B #u #u { B [ ^ { {

(4-1)
{ gh'^ { { ^ { [ ,R^ { ^ :C { :C [ { B #u { #u B [ ^ {-{
gh'^ { ^ { [ ,R^^ :Cy7 :C [ B #u
7 #u B [ ^ {
where ^h! #%$F@`5)R$%,.$2 8 $2#!'b
are the distances between known GPS stations of the test network Stuttgart
Central, ^& ~5)R$%,.$2 8 are the unknown distances measured from the unknown GPS point ebd to the
known GPS stations d #5+)+$-,.$- 8 and 9?$23 #5+)R$%,.$- 8 are the LPS observable of types horizontal
and vertical directions from the unknown point to the known GPS stations @ #5+)+$-,.$- 8
respectively. With (2-77) and (2-68) in pages (25 and 24 respectively) we have the Jacobi matrices as
R

^
U ' {
^
^ R
^ {
^ R
{
^{
{
^
{
(4-2)
4-3. OVERDETERMINED SOLUTION

and

^ {

' {
^ {
^
{
^
{-
^ {
{-
^
{-
71
^ {
y
{
{
{
{
{
{
(4-3)
$ are obtained from the closed form solution of the first

The values 56^$-^ $-^ 8 appearing in the Jacobi matrices
of the vector
{
of observations and with (4-2) and (4-3) forming
combinatorial set in step 1. From the dispersion
'
U , the variance-covariance matrix
'
HR{ H pH
H R H { H
H R H H{
is finally obtained from (2-68) as
H {
P
P
P
P
P
H -pFHW - P
P
P
P
P
P
HWpF- H p{ X HWpF- P
W
H
p

H
p

H
{
P
P
P
P
P
P
P
P
Ha{ P
P
P
P
P U
P
P
P
P Ha{ P
P
P
P
' P
(4-4)
P
P
P
P
P Ha{ P
P
P
P
P
P
P
P
P H _{ P
P
P
P
P
P
P
P
P H_{ P
P
P
P
P
P
P
P
P H_{
with the elements of

on the right hand side of (4-4) given by
H{ HW -X HW %p
HWX% H{ F HWX%p '
H 2R H ? X H{
(4-5)
56R#+6 5 H { H { H { H { H { H { H { H { H { 8
Y

and
in (4-5) being the Jacobi matrix of the partial derivatives of the distance equations
{
^ { ':C { [ B { :C { [ B { :F { [ B {
^ {-{ ':C [ { B { :C [ { B { :F [ { B {
(4-6)
^ { ':C [ B { :C [ B { :F [ B {
connecting the known GPS points d #5+)R$%,.$- 8 with respect to the known coordinates 5 $2 $7 68 $
{ 5 { $2 { $% { 8 and 5 $2 $7 8 of the GPS stations involved in the minimal combinatorial set. The variancecovariance matrix computed above is obtained for every combinatorial subset. Finally we obtained the variancecovariance matrix from (2-70) in Chapter 2 page (24).
Step 3 (rigorous adjustment of the combinatorial solution points in a polyhedron):
Once the ) P combinatorial solution points in a polyhedron have been obtained in step 1, they are finally adjusted using the linear Gauss-Markov model (2-1), page 7) with the dispersion matrix obtained via the error
propagation law or variance-covariance propagation in step 2. Expressing each of the 105 pseudo-observation
distances as
^ N
' ^
#Y5)R$-,J$23$3$%.$-1J$%4 8 $uy5+)R$%,.$-J$W$-J$21J$74$J 0 $2+ 8 $
and placing the pseudo-observation distances ^ in the vector of observation , the coefficients
of the unknown
seven distances ^ of the test network Stuttgart Central forming the coefficient matrix and comprising the
vector of unknowns ^ , the adjusted solution is obtained via (2-4) and the dispersion of the estimated parameters
through (2-5) in Chapter 2 page (8).
72
Step 4 (construction of minimal combinatorial subsets for position determination):

Once the adjusted distances and their dispersion matrix have been estimated using (2-4) and (2-5) respectively
in step 3, the position of the unknown point is then determined using either Grbner basis approach (Box 3-5
in page 43) or Multipolynomial resultant approach (Box 3-6 in page 44). Similar to the distances, we have 35
combinatorial subsets giving 35 different positions `$2!$%( of the same point (for each simplex of Appendix
A.2, we get the position value 5$2!$% 8 ). In total we have 105 ( ` ) values of $-$ and which will be
treated as pseudo-observations.
Step 5 (nonlinear error propagation to determine the dispersion matrix ):
The variance-covariance matrix are computed for each of the combinatorial set (')R$ 0 $2+ using error propagation. The closed form observational equations for the first combinatorial subset ' ) (i.e. tetrahedron
y ) Appendix A.2 are written algebraically as
B { C: [ =
B { u: [ c B { [ ^ {
R cgt':<s [
{ gt':< { [
B { C: { [ =
B { u: { [ c B { [ ^ { {
gt':< [
B { :C [ =
B { :u [ c B { [ ^ {
(4-7)
where ^ #5+)+$-,J$2 8 r') are the distances between known GPS stations |d #5+)+$-,.$- 8 of the test
network Stuttgart Central and the unknown GPS point d for first combination set ' ) . With (2-77)
and (2-68) in pages (25 and 24 respectively) we have the Jacobi matrices as
U ' {

(4-8)
and

^
' {
^
{
{
{
{
{
{
+
{
{ {
{
(4-9)
U $ are obtained from the closed form solution of the first

of the vector of observations and with (4-8) and (4-9)
The values 5$2!$% 8 appearing in the Jacobi matrices

combinatorial set in step 4. From the dispersion matrix
y , the variance-covariance matrix
forming '
'
H{
H
H
H
H{
H
H
H
H{
4-4. RESULTS
73
is finally obtained from (2-68) as
H {
P
P
P
P
P
P
P
P
H R R H H R{ G HH R G PP
P
P
P
P
P
P
P
P
R
H

H
{
P
P
P
P
P
P
P
P
P
p
{
P
P
P
H
P
P
P
P
P
P
P
P

{
P
P
P
P
H
P
P
R
P
P
P
P
P
P
P
P
P
P
P

H
{
P
P
P
P
P
P
(4-10)
' P
P
P
P
P
P

H
{
P
P
P
P
P

P
P
P
P
P
P H{ P
P
P
P
P

P
P
P
P
P
P
P H{ P
P
P
P
P
P
P
P
P
P
P
P
P H{ P
P
P
P
P
P
P
P
P
P
P
P H{ P
P
P
P
P
P
P
P
P
P
P
P H{
on the right hand side of (4-10) given by (4-4) in page (71). The variance-covariance
with the @ elements of
matrix computed as explained above is obtained for every combinatorial set ')R$0 0$-+ . Finally we obtained the
dispersion matrix from (2-70) in Chapter 2 page (24).
Step 6 (rigorous adjustment of the combinatorial solution points in a polyhedron):
For each of the 35 computed coordinates of point K1 in Figure (4.1) in step 4, we write the observation equations
as
'
h$u5+)+$-,.$-J$3$-J$213$%4$300 $2 8

(4-11)
'K
5+)+$-,.$-J$3$-J$213$%4$300 $2 8

8
'
F
$
+
5
+
)
$
J
,
2
$
J

$
3

%
$
.
2
$
3
1
%
$
.
4
J
$
0
2
$

With the values 5 $- $% 8 treated as pseudo-observation and placed in the vector

of observation , the
coefficients of the unknown position 5 $-$7 8 being placed in the coefficient matrix and comprising the
vector of unknowns 5$2!$% 8 $ The solution is obtained via (2-4) and the dispersion of the estimated parameters
through (2-5) in Chapter 2 page (8).
4-4
Results
4-41 Distances
Presented in Table (4.14) are the results of the computed closed form three-dimensional resection distances from K1
to Haussmanstr. :Xy gh'L^ B , K1-Eduardpfeiffer :Xy
K1-Liederhalle :Xy gh' ^ B (e.g. tetrahedron
{ gh' ^ f{ B ^ and
obtained by subtracting the computed distance ^

5y { +8 in Appendix A.2 page 104) and their deviation
from its ideal value ^ in Table (4.2). Observational set number 0* comprised the ideal values of Table (4.2) which are
used to control the Experiment.
In Tables (4.15), (4.16) and (4.17) are presented the results of the adjusted distances, root-mean-square-errors and
the deviations in distances computed using the Gauss-Jacobi combinatorial algorithm. The root-mean-square error
are computed from (2-5) in Chapter 2 page (8). The deviation in distances are obtained by subtracting the computed distance ^ from its ideal value ^ (in Table 4.2, page 65). The adjusted distances in Tables (4.15) were: K1Haussmanstr. :F^ B , K1-Eduardpfeiffer :X^ B , K1-Lindenmuseum :X^ B , K1-Liederhalle :X^ B , K1-Dach LVM :X^ B ,
{
K1-Dach FH :X^ B and K1-Haussmanstr. :X^ B .
4-42
Position
The obtained position of station K1 from the 11 sets under study are presented in Tables (4.18) and (4.19). Set 0* indicate the results of the theoretical set. Since the value of K1 is known (e.g. Table (4.1), the deviations 5pf($-fr!$-f 8
of the computed positions from their real values are computed for both the closed form three-dimensional resection
and the overdetermined three-dimensional resection for each observational data set and are plotted as in Figure (4.2).
Figure (4.2) indicates the results of the overdetermined three-dimension resection for the test network Stuttgart Central computed from the Gauss-Jacobi algorithm to be better than those computed from closed form procedures. The
74
root-mean-square errors are computed from (2-5) in Chapter 2 page (8). Figures (4.3)-(4.8) illustrates the plotted
three-dimensional positional scatter of the 35 minimal combinatorial subsets (indicated by doted points ( )) around
for the adjusted value of position indicated by a star ( ). The plot is done for each observational data set in Tables
(4.3) to (4.13).
Table 4.14: Distances computed by Grbner basis or Multipolynomial resultants
set no.
^ :CB
^ { :<B
^ :<B f^ :CB f^ { :CB f^ :CB
0*
1324.2380 542.2609 430.5286
0.0000
0.0000
0.0000
1
1324.2420 542.2613 430.5247 -0.0040
-0.0004
0.0039
2
1324.2479 542.2707 430.5138 -0.0099
-0.0098
0.0148
3
1324.2412 542.2615 430.5247 -0.0032
-0.0006
0.0039
4
1324.2342 542.2588 430.5330
0.0038
0.0021
-0.0043
5
1324.2340 542.2626 430.5313
0.0040
-0.0017
-0.0027
6
1324.2442 542.2664 430.5193 -0.0063
-0.0055
0.0093
7
1324.2453 542.2540 430.5241 -0.0073
0.0069
0.0045
8
1324.2376 542.2585 430.5301
0.0003
0.0024
-0.0015
9
1324.2382 542.2560 430.5307 -0.0002
0.0049
-0.0021
10
1324.2369 542.2652 430.5276
0.0011
-0.0043
0.0010
11
1324.2375 542.2594 430.5299
0.0005
0.0015
-0.0013
Table 4.15: Distances computed by Gauss-Jacobi combinatorial algorithm

No.
1
2
3
4
5
6
7
8
9
10
11
^ :<B
1324.2394
1324.2387
1324.2381
1324.2363
1324.2396
1324.2378
1324.2368
1324.2388
1324.2393
1324.2337
1324.2375
^ { C: B
542.2598
542.2606
542.2604
542.2545
542.2611
542.2584
542.2558
542.2575
542.2646
542.2598
542.2573
^ C: B
364.9782
364.9801
364.9791
364.9782
364.9779
364.9791
364.9790
364.9779
364.9794
364.9832
364.9787
^ :<B
^ :<B
^ :<B
^ :CB
430.5281
430.5274
430.5267
430.5355
430.5259
430.5300
430.5328
430.5324
430.5232
430.5350
430.5326
400.5834
400.5818
400.5847
400.5931
400.5834
400.5868
400.5857
400.5845
400.5770
400.5904
400.5884
269.2303
269.2292
269.2296
269.2385
269.2306
269.2320
269.2345
269.2342
269.2265
269.2346
269.2344
566.8641
566.8635
566.8632
566.8664
566.8658
566.8637
566.8644
566.8664
566.8623
566.8608
566.8650
Table 4.16: Root-mean-square-errors of distances in Table (4.15)

set no. ^ :CB
^ { :<B ^ :CB ^ :<B ^ :<B ^ :CB ^ :<B
1
0.0004 0.0004 0.0004 0.0005 0.0004 0.0006 0.0003
2
0.0011 0.0012 0.0013 0.0015 0.0014 0.0019 0.0009
3
0.0008 0.0008 0.0009 0.0010 0.0009 0.0013 0.0006
4
0.0007 0.0007 0.0008 0.0010 0.0009 0.0013 0.0006
5
0.0008 0.0008 0.0009 0.0010 0.0010 0.0013 0.0006
6
0.0008 0.0008 0.0009 0.0010 0.0010 0.0013 0.0006
7
0.0012 0.0012 0.0013 0.0016 0.0014 0.0020 0.0010
8
0.0010 0.0010 0.0011 0.0013 0.0012 0.0016 0.0008
9
0.0009 0.0009 0.0010 0.0012 0.0011 0.0015 0.0007
10
0.0006 0.0006 0.0006 0.0008 0.0007 0.0010 0.0005
11
0.0005 0.0005 0.0005 0.0006 0.0006 0.0008 0.0004
4-4. RESULTS
75
Table 4.17: Deviations of distances in Table (4.15) from values in Table (4.2)
set no. f^ :<B
f^ { :<B f^ :<B f^ :CB f^ :CB fr^ :CB fr^ :<B
1
-0.0014
0.0011
0.0015
0.0005
0.0002
0.0006
-0.0006
2
-0.0007
0.0003
-0.0004
0.0012
0.0019
0.0017
0.0000
3
-0.0002
0.0005
0.0006
0.0019
- 0.0010
0.0013
0.0004
4
0.0017
0.0064
0.0015
-0.0069
-0.0094
-0.0075
-0.0028
5
-0.0016
-0.0002
0.0018
0.0027
0.0003
0.0003
-0.0023
6
0.0002
0.0025
0.0006
-0.0014
-0.0031
-0.0011
-0.0002
7
0.0012
0.0051
0.0007
-0.0042
-0.0020
-0.0035
-0.0009
8
-0.0009
0.0034
0.0018
-0.0037
-0.0008
-0.0033
-0.0028
9
-0.0013
-0.0037
0.0003
0.0054
0.0067
0.0044
0.0013
10
0.0042
0.0011
-0.0035
-0.0063
-0.0067
-0.0037
0.0027
11
0.0004
0.0036
0.0010
-0.0040
-0.0047
-0.0034
-0.0015
Table 4.18: K1 computed by Grbner basis or Multipolynomial resultants
Set No.
0
1
2
3
4
5
6
7
8
9
10
11
b:<B
s:CB
d:CB
f(b:CB
f(s:<B
f=:<B
4157066.1116
4157066.1166
4157066.1220
4157066.1105
4157066.1045
4157066.1068
4157066.1149
4157066.1074
4157066.1099
4157066.1084
4157066.1138
4157066.1107
671429.6655
671429.6625
671429.6586
671429.6622
671429.6678
671429.6688
671429.6606
671429.6569
671429.6653
671429.6642
671429.6674
671429.6657
4774879.3704
4774879.3720
4774879.3599
4774879.3661
4774879.3688
4774879.3658
4774879.3614
4774879.3708
4774879.3724
4774879.3740
4774879.3673
4774879.3721
0
0.0050
-0.0104
0.0011
0.0071
0.0048
-0.0033
0.0042
0.0017
0.0032
-0.0022
0.0009
0
0.0030
0.0069
0.0033
-0.0023
-0.0033
0.0049
0.0086
0.0002
0.0013
-0.0019
-0.0002
0
-0.0016
0.0105
0.0043
0.0016
0.0046
0.0090
-0.0004
-0.0020
-0.0036
0.0031
-0.0017
Table 4.19: Position of K1 computed by Gauss-Jacobi combinatorial algorithm

Set No.
b:CB
~:<B
=:<B
H :<B H :CB H :<B
1
4157066.1142 671429.6642 4774879.3705 0.00007 0.00002 0.00007
2
4157066.1150 671429.6656 4774879.3695 0.00009 0.00001 0.00008
3
4157066.1100 671429.6650 4774879.3676 0.00010 0.00002 0.00010
4
4157066.1040 671429.6648 4774879.3688 0.00008 0.00002 0.00008
5
4157066.1089 671429.6635 4774879.3699 0.00016 0.00003 0.00015
6
4157066.1127 671429.6651 4774879.3684 0.00017 0.00003 0.00016
7
4157066.1089 671429.6655 4774879.3699 0.00009 0.00002 0.00009
8
4157066.1102 671429.6643 4774879.3720 0.00009 0.00002 0.00008
9
4157066.1106 671429.6649 4774879.3699 0.00004 0.00001 0.00003
10
4157066.1121 671429.6694 4774879.3697 0.00005 0.00001 0.00005
11
4157066.1100 671429.6654 4774879.3705 0.00005 0.00001 0.00005
Table 4.20: Deviation of K1 in Table (4.19) from the real value in Table (4.1)
Set No. f(b:CB
f(~:<B f=:<B
1
-0.0026
0.0013
-0.0001
2
-0.0034 -0.0001
0.0009
3
0.0016
0.0005
0.0028
4
0.0076
0.0007
0.0016
5
0.0027
0.0020
0.0005
6
-0.0011
0.0004
0.0020
7
0.0027
-0.0000
0.0005
8
0.0014
0.0012
-0.0016
9
0.0010
0.0006
0.0005
10
-0.0005 -0.0039
0.0007
11
0.0016
0.0001
-0.0001
76
Deviations in X coordinates from the theoretical value of point K1
x 10
Closed form 3dresection

GaussJacobi algorithm
Deviation in X(m)
10
12
10
6
7
Experiment No.
10
11
10
11
Deviations in Y coordinates from the theoretical value of point K1
x 10

8
Deviation in Y(m)
12
6
7
Experiment No.
Deviations in Z coordinates from the theoretical value of point K1
x 10

10
Deviation in Z(m)
6
7
Experiment No.
10
11
Figure 4.2: Deviation of computed position of station K1 in Tables (4.18) and (4.20) from the real value in Table (4.1)
4-4. RESULTS
77
3dplot of the scatter of the combinatorial solutions around the adjusted value(set1)
0.08
0.078
Z(m)+4774879.3(m)
0.076
0.074
0.072
0.07
0.068
0.066
0.064
5
4.5
0.02
0.018
4
x 10
0.016
0.014
3.5
0.012
0.01
3
Y(m)+671429.66(m)
0.008
X(m)+4157066.1(m)
0.085
Z(m)+4774879.3(m)
0.08
0.075
0.07
0.065
0.06
8
7
3
x 10
0.03
0.025
0.02
0.015
5
0.01
Y(m)+671429.66(m)
0.005
X(m)+4157066.1(m)
Figure 4.3: Scatter of combinatorial solutions for data sets 1 and 2 in Tables (4.3) and (4.4)
78
0.09
Z(m)+4774879.3(m)
0.085
0.08
0.075
0.07
0.065
0.06
0.055
8
7
0.025
6
0.02
x 10
0.015
5
0.01
0.005
3
Y(m)+671429.66(m)
X(m)+4157066.1(m)
0.085
Z(m)+4774879.3(m)
0.08
0.075
0.07
0.065
0.06
7
6.5
20
6
3
x 10
15
5.5
10
5
5
4.5
Y(m)+671429.66(m)
0
4
x 10
X(m)+4157066.1(m)
4-4. RESULTS
79
0.085
0.08
Z(m)+4774879.3(m)
0.075
0.07
0.065
0.06
0.055
0.05
0.045
6
5
0.03
4
0.02
x 10
0.01
3
0
0.01
1
Y(m)+671429.66(m)
0.02
X(m)+4157066.1(m)
0.095
0.09
Z(m)+4774879.3(m)
0.085
0.08
0.075
0.07
0.065
0.06
0.055
10
8
3
x 10
0.04
0.03
6
0.02
4
Y(m)+671429.66(m)
0.01
2
X(m)+4157066.1(m)
80
0.085
Z(m)+4774879.3(m)
0.08
0.075
0.07
0.065
0.06
7.5
7
0.02
6.5
x 10
0.015
6
0.01
5.5
5
Y(m)+671429.66(m)
0.005
X(m)+4157066.1(m)
0.085
Z(m)+4774879.3(m)
0.08
0.075
0.07
0.065
0.06
6
5
3
x 10
0.02
0.015
4
0.01
3
Y(m)+671429.66(m)
0.005
2
X(m)+4157066.1(m)
4-4. RESULTS
81
0.075
0.074
Z(m)+4774879.3(m)
0.073
0.072
0.071
0.07
0.069
0.068
0.067
5.4
5.2
0.014
0.013
5
3
x 10
0.012
4.8
0.011
0.01
4.6
0.009
4.4
Y(m)+671429.66(m)
0.008
X(m)+4157066.1(m)
0.09
Z(m)+4774879.3(m)
0.085
0.08
0.075
0.07
0.065
0.06
0.055
0.012
0.011
0.025
0.02
0.01
0.015
0.009
Y(m)+671429.66(m)
0.01
0.008
0.005
X(m)+4157066.1(m)
82
0.082
0.08
Z(m)+4774879.3(m)
0.078
0.076
0.074
0.072
0.07
0.068
0.066
0.064
7
6.5
0.025
6
3
x 10
0.02
5.5
0.015
5
0.01
4.5
Y(m)+671429.66(m)
0.005
X(m)+4157066.1(m)
Figure 4.8: Scatter of combinatorial solutions for data set 11 Table (4.13)
Chapter 5
Case studies
5-1
Geocentric Cartesian to Gauss ellipsoidal coordinates
In order to relate a point on the Earths topographic surface to a point on the International Reference Ellipsoid
d{ we work with a bundle of half straight lines so called projection lines which depart from and intersect d{
either not at all or in two points. There is one projection line which is at minimum distance relating to . Figure
(5.1) is an illustration of such a Minimum Distance Mapping. Such an optimization problem is formulated by means
of the Lagrangean y:<p$2 $2 $?WB as in (3-112) respectively (3-113).
Figure 5.1: Minimum distance mapping of a point on the Earths topographic surface to a point on the International
Reference Ellipsoid d{
In the first case we represent the Euclidean distance between the points and in terms of Cartesian coordinates of
:<`$2!$%cB and of :< $2 { $? B . The Cartesian coordinates :< $2 { $? B of the projection point are unknown. The
constraint that the n is an element of the ellipsoid-of-revolution d{ gh' 5K @R {R:<{ W{ B { [ {7{'
PJ$2 c~ 8 is substituted into the Lagrangean by means of the Lagrange multiplier W{ which is{ unknown,
too. :< $2 $2 $? B'5y:<p$2 $2 $?W6Bs'i#u 8 is the argument of the minimum of the constrained La{
{
grangean y:<6$? $2 $?WB . The result of the minimization procedure is presented by Lemma (5-1). (LC1) provides
{
the necessary conditions to constitute an extremum: The normal equations are of bilinear type. Products of the unknowns for instance 93$? 3$? W and squares of the unknowns, for instance { $? { $2 { appear. Finally the matrix
{
{
of second derivatives in (LC2) to be positive definite constitutes the necessary condition to obtain a minimum. For
tunately the matrix of second derivatives
is diagonal. The eigenvalues of are ' { '6 $- ' 6
and must be positive.
83
CHAPTER 5. CASE STUDIES
84
Lemma 5-1 (constrained minimum distance mapping):

The functional y:< $2 $? $? B ist minimal, if the conditions (LC1) and (LC2) hold.
<: $? { $2 $? BY'NP"! i=1,2,3,4
#: B ' [ :< [ B { 'KP

$: B ' [ :C [ { B { { 'KP

#: B ' [ :u [ B { 'NP

)
'
B , { :< { { B { { [ { { 'P
(LC1)
C: #B
:C#F#GB
(5-1)
:<#u#u#GB
:<#&%B :#
{'
(LC2)
:C $2 { $? $2 YB 'P(! i,j {1,2,3}

)
{' :#+* B-,
gh'
) {
P
/..
P
) {
P
'
P
P
) {
(5-2)
(5-3)
"eigenvalues"
) [ 10 + 'KP 243

'N { hg ') { ' '
gh') { '
(5-4)
(5-5)
(5-6)
Box 5-1 (Conversion from Cartesian coordinates to Gauss ellipsoidal coordinates):

Closed form solution:
5$2!$% 8 65r{ 5 6$? { 2$ 8 87d {

to :5 9U$2s$; 8
*gt'
G'
Pythagoras in three dimension
:C [ B {! :X [ { B {! :F [ B {
convert 5 $2 $? 8 and 5 $-$7 8
{
to <5 9E$2 8
>
G+=9' [[ { ' [
[
[
[
'
C: [ B { X: [ { B {
:< [ B { X: [> B {
(5-7)
(5-8)
(5-9)
The nonlinear (algebraic: bilinear) normal equations (LC1) of Minimum distance Mapping 5({ [ ? d{ was solved

in Section (3-22) in a closed form by means of Grbner basis. The computed elements of the Grbner basis are
presented in Box (3-11) in page (50). Using (3-116) and (3-117) in page (51), the Lagrangean multiplier 3 and
ellipsoidal Cartesian coordinates 5 6$? $2 8 are computed. Finally, by means of Box (5-1) we convert the Cartesian
{
coordinates :C$2!$%cB85r{ and :<p$2 $? B&d{ to Gauss ellipsoidal coordinates 9U$2s$; .
5-1. GEOCENTRIC CARTESIAN TO GAUSS ELLIPSOIDAL COORDINATES
85
Case study:
Let us adopt the World Geodetic Datum 2000 with the data r'N1R4pJ)R13 1+P+,p and &'N1++14J)R R1RPR of type semimajor axis and semi-minor axis respectively, of the International Reference Ellipsoid (E. Grafarend and A. Ardalan
1999). Here we take advantage of given Cartesian coordinates of 21 points of the topographic surface of the Earth
presented in Table (5.1). From the algorithm of Box (3-11) in page (50), the first polynomial equation of fourth order
of the Grbner basis (3-116) in page (51) is solved. We compute the coefficients from the input data Table (5.1) and
solve for 3 according to Table (5.2). With the admissible values 3 substituted in the computed Grbner basis (3-117)
in page (51) we finally produce the values :< $2 $? B' :<$ > $;@B in Table (5.3). At this end Table (5.4) converts by
{
means of (5-1) the Cartesian coordinates :<`$2!$%cB and :<$ > $;@B to :#9U$-@$A`B .
Table 5.1: Cartesian coordinates of topographic point (Baltic Sea Level Project)
Station
b:<B
s:CB
d:CB
Borkum (Ger)
3770667.9989 446076.4896 5107686.2085
Degerby (Fin)
2994064.9360 1112559.0570 5502241.3760
Furugrund (Swe) 2527022.8721 981957.2890 5753940.9920
Hamina (Fin)
2795471.2067 1435427.7930 5531682.2031
Hanko (Fin)
2959210.9709 1254679.1202 5490594.4410
Helgoland (Ger)
3706044.9443 513713.2151 5148193.4472
Helsinki (Fin)
2885137.3909 1342710.2301 5509039.1190
Kemi (Fin)
2397071.5771 1093330.3129 5789108.4470
Klagshamn (Swe)
3527585.7675 807513.8946 5234549.7020
Klaipeda (Lit)
3353590.2428 1302063.0141 5249159.4123
List/Sylt (Ger)
3625339.9221 537853.8704 5202539.0255
Molas (Lit)
3358793.3811 1294907.4149 5247584.4010
Mntyluoto (Fin)
2831096.7193 1113102.7637 5587165.0458
Raahe (Fin)
2494035.0244 1131370.9936 5740955.4096
Ratan (Swe)
2620087.6160 1000008.2649 5709322.5771
Spikarna (Swe)
2828573.4638 893623.7288 5627447.0693
Stockholm (Swe)
3101008.8620 1013021.0372 5462373.3830
Ustka (Pol)
3545014.3300 1073939.7720 5174949.9470
Vaasa (Fin)
2691307.2541 1063691.5238 5664806.3799
Visby (Swe)
3249304.4375 1073624.8912 5364363.0732
lands N. U. (Swe) 3295551.5710 1012564.9063 5348113.6687
86
Table 5.2: Polynomial coefficients

Point
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
-2.3309099e+22
-1.142213e+22
-1.720998e+22
-8.871288e+21
-1.308070e+22
-2.275210e+22
-1.272935e+22
-1.373946e+22
-1.981047e+22
-2.755981e+22
-2.330047e+22
-1.538357e+22
-1.117760e+22
-1.124559e+22
-1.200556e+22
-1.427443e+22
-1.836471e+22
-1.772332e+22
-1.012020e+22
-1.427711e+22
-1.644250e+22
$U p$U { $ $U7

1.334253e+41
1.3351890e+41
1.335813e+41
1.335264e+41
1.335160e+41
1.334345e+41
1.335205e+41
1.335906e+41
1.334546e+41
1.334574e+41
1.334469e+41
1.334580e+41
1.335399e+41
1.335785e+41
1.335704e+41
1.335496e+41
1.335087e+41
1.334410e+41
1.335593e+41
1.334856e+41
1.334815e+41
of the univariate polynomial of order four in W
1.351627e+55
1.352005e+55
1.352259e+55
1.352035e+55
1.351993e+55
1.351665e+55
1.352012e+55
1.352296e+55
1.351746e+55
1.351758e+55
1.351715e+55
1.351759e+55
1.352090e+55
1.352246e+55
1.352214e+55
1.352130e+55
1.351965e+55
1.351690e+55
1.352168e+55
1.351870e+55
1.351854e+55
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
4.38e+68
7
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.441e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
4.44e+81
Table 5.3: Computed Cartesian coordinates :<p$2 $? BY':<$ > $;@B and Lagrange multiplier 3
Station
:CB
{ :<B { :<B
:< { B
Borkum
3770641.3815 446073.3407 5107649.9100 1.746947e-019
Degerby
2994054.5862 1112555.2111 5502222.2279 8.554612e-020
Furugrund 2527009.7166 981952.1770 5753910.8356 1.288336e-019
Hamina
2795463.7019 1435423.9394 5531667.2524 6.643801e-020
Hanko
2959199.2560 1254674.1532 5490572.5584 9.797001e-020
Helgoland
3706019.4100 513709.6757 5148157.7376 1.705084e-019
Helsinki
2885126.2764 1342705.0575 5509017.7534 9.533532e-020
Kemi
2397061.6153 1093325.7692 5789084.2263 1.028464e-019
Klagshamn
3527564.6083 807509.0510 5234518.0924 1.484413e-019
Klaipeda
3353562.2593 1302052.1493 5249115.3164 2.065021e-019
List/Sylt
3625314.3442 537850.0757 5202502.0726 1.746017e-019
Molas
3358777.7367 1294901.3835 5247559.7944 1.152676e-019
Mntyluoto
2831087.1439 1113098.9988 5587146.0214 8.370165e-020
Raahe
2494026.5401 1131367.1449 5740935.7483 8.418639e-020
Ratan
2620078.1000 1000004.6329 5709301.7015 8.988111e-020
Spikarna
2828561.2473 893619.8693 5627422.6007 1.068837e-019
Stockholm
3100991.6259 1013015.4066 5462342.8173 1.375524e-019
Ustka
3544995.3045 1073934.0083 5174921.9867 1.328158e-019
Vaasa
2691299.0138 1063688.2670 5664788.9183 7.577249e-020
Visby
3249290.3945 1073620.2512 5364339.7330 1.069551e-019
lands N. U. 3295535.1675 1012559.8663 5348086.8692 1.231803e-019
5-2. 7-PARAMETER DATUM TRANSFORMATION
87
Table 5.4: Geodetic Coordinates computed from ellipsoidal Cartesian coordinates in closed form (Baltic Sea Level
Project)
Station
Longitude 9
Latitude
height
Borkum (Ger)
Degerby (Fin)
Furugrund (Swe)
Hamina (Fin)
Hanko (Fin)
Helgoland (Ger)
Helsinki (Fin)
Kemi (Fin)
Klagshamn (Swe)
Klaipeda (Lit)
List/Sylt (Ger)
Molas (Lit)
Mntyluoto (Fin)
Raahe (Fin)
Ratan (Swe)
Spikarna (Swe)
Stockholm (Swe)
Ustka (Pol)
Vaasa (Fin)
Visby (Swe)
lands N. U. (Swe)
5-2
6 44 48.5914
20 23 4.0906
21 14 6.9490
27 10 47.0690
22 58 35.4445
7 53 30.3480
24 57 24.2446
24 31 5.6737
12 53 37.1597
21 13 9.0156
8 26 19.7594
21 4 58.8931
21 27 47.7777
24 24 1.8197
20 53 25.2392
17 31 57.9060
18 5 27.2528
16 51 13.8751
21 33 55.9146
18 17 3.9292
17 4 46.8542
53 33 27.4808
60 1 52.8558
64 55 10.2131
60 33 52.9819
59 49 21.6459
54 10 29.3979
60 9 13.2416
65 40 27.7029
55 31 20.3311
55 45 16.5952
55 1 3.0992
55 43 47.2453
61 35 39.3552
64 38 46.8352
63 59 29.5936
62 21 48.7645
59 19 20.4054
54 35 15.6866
63 5 42.8394
57 38 21.3487
57 22 3.4508
45.122
22.103
33.296
17.167
25.313
44.042
24.633
26.581
38.345
53.344
45.101
29.776
21.628
21.757
23.228
27.620
35.539
34.307
19.581
27.632
31.823
7-parameter datum transformation
The seven parameters datum transformation C
:CB

7 N
' E D { p#')+$-,J$2J$9 0$?

(5-10)
subject to
_ D 'G0 F
(5-11)
':I0 [ HUB #: 0 HUB
(5-12)
5?$-7?$2 8 5G$2?$7| 8 being coordinates of the same points in both systems and L , { L ,
@/.. has always been solved using least squares solution. We illustrate in this section how the seven parameters datum transformation C :CB of a non-overdetermined and overdetermined cases can be solved by the use
of Grbner bases and Gauss-Jacobi combinatorial algorithms respectively. First, we solve analytically for ,
D `@/ .. ad then use them to transform the coordinates of one system in order to obtain the corresponding transformed values in the other system. By making use of the skew-symmetric matrix H , the rotation matrix is expressed as
with
D
where 0
is the identity matrix and the skew-symmetric matrix H given by
HK'
The rotation matrix D
@/..
[
P [
[ P
P
(5-13)
is parameterized using Euler or Cardan angles. With Cardan angles, we have:
88
Box 5-2 (Parametrization of the rotation matrix by Cardan angles):
:IKB'
leading to
P
0K
[ #F= K
'GJ :IKB;J { :$LB;J :NMB
P
#u=K A$ J { :#LB`'
<K
with
0EL
P
#uO L
(5-14)
[ #uOL
)
P P$ J :NMB`'
P L
P
0+M
[ #F M
P
Jp:IKB;J { :$LB;J :NMBY'
LEM
L #u M
[ 0#uOL
#u=K #uOLEM [ <K #u M #u=K #uOL #u M 0KEM #u=K0EL
<K #FOLE M #u= K0#u M <K #FOL0#u M [ #u=KEM 0KL
The Cardan angles can be obtained from the rotation matrix D
~@/ .. through:
K'KGY: {2 B7$QM'NGY: { B
-[
L' GY:
B.6G+Y:
B
{2 { {
{2{ 2{
#u M
EM
P
P
)
(5-15)
(5-16)
For parameterization using Euler angles we refer to E. Grafarend and J. L. Awange (2000). The properties of the rotation matrix D
@/.. expressed as in (5-12) have been examined by S. Zhang (1994) and shown to fulfill (5-11).
Equation (5-10) is now written for #')+$-,J$2 using (5-12) as
[

)

) [

K
' ) [

)

[
[ )
(5-17)
[
with 5 $- $% 8 6 For three corresponding points in both systems, the minimum observation equations required
{
to solve the seven datum transformation parameter are expressed from (5-17) as:
gh'N [ 0 [ [ 0 7 'NP
{ gt'N 7 [ J 3 [ [ 'KP
gt' [ 90%s 0! [ . .0 [ 'P
(5-18)
6ygh'N9 { [ 90 { 9 0 { [ { [ 0 { 7 { 'NP $
p
gt'N907 { 90 { [ 90J { { [ { [ { 'KP
gt' [ 90% { { 0 { [ { . { [ { 'P
gt' [ 90% 0 [ . [ 'P

where 5W?$ > 2$;@6 8 gh'56G$-?$- 8 9#(5)R$%,.$2 8 , equations 5p $% $% 8 being equations formed from the first point
{
with coordinates in both systems, 5p $% $% p8 being equations formed from the second point with coordinates in both
systems and 56 68 being the third equation extracted from the three equations formed from the third point with coordinates in both systems. In order to eliminate the translation parameters in (5-18), the following differencing is
performed:
R Gygt'R [

gh'N [
{
gt' { [
- gt' [
6c'N9%s [ 909
p E'K907s{ { 7 { { [
pE' [ 07@ 90
' [ 7 {- {-
900 [
90J {{
! [
{2 [
with
hE'N [ ( Whc'N [ .|hc'| [

h'N [ R
7tc'N [ -
J { [ 00 {
0. { [ 0 { [
7J .0
7 {- . {-
7
{ {
[
[ {2
(5-19)
6#%$u5+)R$%,.$- 8 $2#'b
+
hc'N [ -
The reduced Grbner basis of (5-19) is then obtained using Mathematica 3.0 Software by using the command GroebnerBasis[{ +$% $- p$%p- },{6$2$%6$2 }, {$-p$2 }] as explained in Chapter 2, Section (2-321). This gives only the
{
element of Grbner basis in which the variables $-6$- have been eliminated and only the scale factor left. The
scale parameter is then given by the following univariate polynomial of order four:
89
Box 5-3 (quartic polynomial for computing scale parameter ):
W { { J%9 'P
a':<s d {
` { s [ ` { 9 [ s { 9 ! { {- [ {2 = { { s 9 { ! { {- [
{- 9 { ! { { {- s { { {2 { !{- B { {2
a ':C s
[ 0 s [ @ s d { [ ` { @ !
{ s { { {- ! { J{- d { { {- [{ 0 { { {2{- { ! { [ {- {- ={ { { 9{- 0 { { s {- { { {2 [ {2{ s { { {- {
` { { {- [ {- { { [ {2 { { {- { { [ {- = { { [ ` { { {-
{2 { { { { {- {- ` { { { {2 { [ { { {2 [ { {- { B
9 0 ! [ s 9 0 s ! [ 0 !
a ':CJ { {
{ { { {- {-[ 7{ { { { {2 {2[ { { { {- [ {- {2 { ` { { { {2 {2 { d { { [ {-{- { ( { { { {-{ {- { { {
{- { { { { {2 [ { { {- [ {- { { [ { { {2 [ {- { {
{ {2 { { { {- [ { { {2 [ {- { { [ { { {- { {2 {
{ {- { [ { { {- {2 { { { { {2 [ { {- { { { {- [
{ {- { B
a ': [
7{ [ { [ { [
{ {- { { { { {- {- { { { [ { { {- {2 { [ {- {- { { { {- { {2 { {-{ [ {-{- { { { { [
7{ { {- {- { { [ { { {- { { {2 { { {2 [ { {- { [ { {- { [
{ { {- { {- { { { {2 {- 7{ { { { {- B
'N. [ . 7{ { 0 [ { J [ J 7{ 0 0 [
{ { { {- { {- { {2 { {- { {2 { {- { 2{ { -{ { {
Once the admissible value of scale parameter has been chosen amongst the four roots above as ~ in Box (5-3),
the elements of the skew symmetric matrix H can then be obtained via the linear functions in Box (5-4).
Box 5-4 (Linear functions for computing the parameters of the skew-symmetric matrix H ):
f(a)=(-x s 9 ! [ 0 { s ! [ J 0 [ J W{ 9 ! 0 { s { s
. { 909 { { . {{ 0 { { J { 0{ ! { { [ { J 0 { 9 { { [ { { { 9s{ { J { { W{ 9 {! {
W { { { { [ { W{ { { { B? : [ { { [ { {
W{ [ [ d { [ { ` {
W { { { { [ { { { { [ { { { [ { 7{ { { { W{ ` { { { W{ d { { { { W{{ { {B
f(b)=(-b [ [ [
{ W{ { {{ [ W{ { { { { { { { { W { { { [ { { { { { { W{ { { {
{ { [ { W{ { [ { W{ { { { { B? : [ { { [ { {
W{ = { { { [ { [ 7{
{ { { { W{ ` {{ { [ W{ { { { { { { { ={ { { [ 7{ { { { [ { { { B {
f(c)=(-b [ [ W{ [ W{ BG
: [ { { [ { { { { { { { [ { { { { { { { [ { {
W{ { { [ { [ { B
Substituting the skew symmetric matrix H in (5-10) gives the rotation matrix D
from which the Cardan rotation
angles are deduced from (5-16) in Box (5-2). The translation elements can then be computed by substituting the
{
scale parameter and the rotation matrix D
in (5-10). Three sets of translation parameters are then obtained for the
three points under consideration and the mean taken.
Gauss-Jacobi combinatorial algorithm

When more than three points are given and the transformation parameters are to be solved, the Gauss-Jacobi combinatorial algorithm is applied in which case the dispersion matrix has to be obtained via the nonlinear error propagation
law/variance-covariance propagation law. From the algebraic system of equations (5-18), we have with (2-77) and
(2-68) in pages (25 and 24 respectively) the Jacobi matrices as
90

{

'

p

p
3 {
| {

p
|

p
|
(5-20)
and

0{ {
{

0
(5-21)

p p

0

represents the partial derivative of (5-18) with respect 56 $2 $? $- $% $? $2 $% $2 $2 8 .
where the doted points in
of the vector of observations and with (5-20) and (5-21) forming
' { , the
{ disper{
From the dispersion
sion matrix
is then obtained from (2-68). Finally we obtained the dispersion matrix from (2-70) in Chapter 2

{
J

'
J

p

{
{
{
p
{
{
{
{
page (24).
Case Study:
We consider Cartesian coordinates of seven stations given in the Local and Global Reference Systems (WGS 84) as in
Tables (5.5) and (5.6). Desired are the seven parameters of datum transformation. Using the Grbner bases algorithm,
we obtain the 7-transformation parameters given in Table (5.7) which are used to transform the three points involved
in the computations from the Local Reference System (Table 5.5) to the Global Reference System (WGS 84, Table 5.6)
in Table (5.8).
91
Table 5.5: Coordinates for system A (Local system)

Station Name
:<B
~:<B
d:CB
Solitude
4157222.543 664789.307 4774952.099
Buoch Zeil
4149043.336 688836.443 4778632.188
Hohenneuffen 4172803.511 690340.078 4758129.701
Kuehlenberg
4177148.376 642997.635 4760764.800
Ex Mergelaec
4137012.190 671808.029 4791128.215
Ex Hof Asperg 4146292.729 666952.887 4783859.856
Ex Kaisersbach 4138759.902 702670.738 4785552.196
Table 5.6:
Station Name
Solitude
Buoch Zeil
Hohenneuffen
Kuehlenberg
Ex Mergelaec
Ex Hof Asperg
Ex Kaisersbach
Coordinates for system B (WGS 84)
:<B
~:<B
d:CB
4157870.237
4149691.049
4173451.354
4177796.064
4137659.549
4146940.228
4139407.506
664818.678
688865.785
690369.375
643026.700
671837.337
666982.151
702700.227
4775416.524
4779096.588
4758594.075
4761228.899
4791592.531
4784324.099
4786016.645
Table 5.7: Computed 7-parameter datum transformation by Grbner basis

Transformation parameter
Value
unit
Scale [ )
-1.4343
[ppm]
Rotation }:XB
0.32575149
[]
Rotation } :X0B
-0.46037399
[]
{
Rotation } :C B
-0.00810606
[]
Translation f=
643.0953
[m]
Translation fr
22.6163
[m]
Translation f
481.6023
[m]
Table 5.8: Transformed Cartesian coordinates of System A (Table 5.5) into System B (Table 5.6) using the 7-datum
transformation parameters of Table (5.7) computed by Grbner basis
Site
:<B
~:<B
=:<B
System A: Solitude
4157222.5430 664789.3070 4774952.0990
System B
4157870.2370 664818.6780 4775416.5240
Transformed value
4157870.3070 664818.6742 4775416.5240
Residual
- 0.0700
0.0038
0.0000
System A: Buoch Zeil
4149043.3360 688836.4430 4778632.1880
System B
4149691.0490 688865.7850 4779096.5880
Transformed value
4149691.1190 688865.7812 4779096.5880
Residual
- 0.0700
0.0038
0.0000
System A: Hohenneuffen 4172803.5110 690340.0780 4758129.7010
System B
4173451.3540 690369.3750 4758594.0750
Transformed value
4173451.2141 690369.3826 4758594.0750
Residual
0.1399
-0.0076
0.0000
Transformation parameters obtained in the overdetermined case by the Gauss-Jacobi combinatorial algorithm are
presented in Table (5.9) and used to transform the Cartesian coordinates from the Local Reference System (Table 5.5)
to the Global Reference System (WGS 84, Table 5.6) as shown in Table (5.10). Table (5.11) gives for comparison
purposes the transformed values from the 7-datum transformation parameters obtained via least squares solution. The
residuals from both Gauss-Jacobi combinatorial algorithm and least squares solution are of the same magnitude. We
also compute the residual norm (square root of the sum of squares of residuals) and present them in Table (5.12). The
computed norm from the Gauss-Jacobi combinatorial solution is somewhat better than those of the linearized least
squares solution. Figures (5.2) and (5.3) indicate the scatter of the computed 36 minimal combinatorial solutions of
92
scale (indicated by doted points ( )) around for the adjusted value indicated by a line ( [ ). Figures (5.3) indicate the
scatter of the computed 36 minimal combinatorial solutions of translation and rotation parameters (indicated by doted
points ( )) around the adjusted values indicated by a star ( ). The Figures clearly identifies the outlying combinations
from which the respective (suspected outlying points) points can be deduced.
Table 5.9: Computed 7-parameter datum transformation using Gauss-Jacobi combinatorial algorithm
Transformation parameter
Value
Root-mean-square
unit
Scale [ )
4.92377597
0.350619414
[ppm]
Rotation } :C.B
-0.98105498
0.040968549
[]
Rotation } :FB
0.68869774
0.047458707
[]
{
Rotation } :X0B
0.96671738
0.044697434
[]
Translation f(
639.9785
2.4280
[m]
Translation fr
68.1548
3.0123
[m]
Translation f
423.7320
2.7923
[m]
transformation parameters of Table (5.9) computed by the Gauss-Jacobi combinatorial algorithm
Site
b:CB
~:<B
=:<B
System A: Solitude
4157222.5430 664789.3070 4774952.0990
System B
4157870.2370 664818.6780 4775416.5240
Transformed value
4157870.1631 664818.5399 4775416.3843
Residual
0.0739
0.1381
0.1397
4149043.3360 688836.4430 4778632.1880
System B
4149691.0490 688865.7850 4779096.5880
Transformed value
4149691.0162 688865.8151 4779096.5785
Residual
0.0328
-0.0301
0.0095
System B
4173451.3540 690369.3750 4758594.0750
Transformed value
4173451.3837 690369.4437 4758594.0770
Residual
-0.0297
-0.0687
-0.0020
System A: Kuelenberg
4177148.3760 642997.6350 4760764.8000
System B
4177796.0640 643026.7000 4761228.8990
Transformed value
4177796.0394 643026.7347 4761228.9783
Residual
0.0246
-0.0347
-0.0793
System A: Ex Mergelaec
4137012.1900 671808.0290 4791128.2150
System B
4137659.5490 671837.3370 4791592.5310
Transformed value
4137659.6895 671837.3142 4791592.5458
Residual
-0.1405
0.0228
-0.0148
System A: Ex Hof Asperg 4146292.7290 666952.8870 4783859.8560
System B
4146940.2280 666982.1510 4784324.0990
Transformed value
4146940.2757 666982.1394 4784324.1589
Residual
-0.0477
0.0116
-0.0599
System A: Ex Keisersbach 4138759.9020 702670.7380 4785552.1960
System B
4139407.5060 702700.2270 4786016.6450
Transformed value
4139407.5733 702700.1935 4786016.6520
Residual
-0.0673
0.0335
-0.0070
93
transformation parameters computed by Least Squares Solution
Site
b:<B
s:CB
d:CB
System A: Solitude
4157222.5430 664789.3070 4774952.0990
System B
4157870.2370 664818.6780 4775416.5240
Transformed value
4157870.1430 664818.5429 4775416.3838
Residual
0.0940
0.1351
0.1402
4149043.3360 688836.4430 4778632.1880
System B
4149691.0490 688865.7850 4779096.5880
Transformed value
4149690.9902 688865.8347 4779096.5743
Residual
0.0588
-0.0497
0.0137
System B
4173451.3540 690369.3750 4758594.0750
Transformed value
4173451.3939 690369.4629 4758594.0831
Residual
-0.0399
-0.0879
-0.0081
System A: Kuelenberg
4177148.3760 642997.6350 4760764.8000
System B
4177796.0640 643026.7000 4761228.8990
Transformed value
4177796.0438 643026.7220 4761228.9864
Residual
0.0202
-0.0220
-0.0874
System A: Ex Mergelaec
4137012.1900 671808.0290 4791128.2150
System B
4137659.5490 671837.3370 4791592.5310
Transformed value
4137659.6409 671837.3231 4791592.5365
Residual
-0.0919
0.0139
-0.0055
System A: Ex Hof Asperg 4146292.7290 666952.8870 4783859.8560
System B
4146940.2280 666982.1510 4784324.0990
Transformed value
4146940.2398 666982.1445 4784324.1536
Residual
-0.0118
0.0065
-0.0546
System A: Ex Keisersbach 4138759.9020 702670.7380 4785552.1960
System B
4139407.5060 702700.2270 4786016.6450
Transformed value
4139407.5354 702700.2229 4786016.6433
Residual
-0.0294
0.0041
-0.0017
Table 5.12: Computed residual norms
Method
:<B @:CB
Linearized Least Squares Solution
0.1541 0.1708
Nonlinear Gauss-Jacobi Combinatorial 0.1859 0.1664
6
14
=:<B
0.1748
0.1725
Scale combinatorial scatter around the adjusted value
x 10
12
10
Scale (ppm)
scatter in scale
Adjusted mean value
0
10
15
20
Combinatorial No.
25
30
35
Figure 5.2: Scatter of the computed 36 minimal combinatorial values of scale around the adjusted value
94
3dplot of translation combinatorial scatter around the adjusted values
300
Z(m)+400(m)
200
100
0
100
600
200
150
500
400
100
300
50
200
0
100
50
0
100
100
150
200
200
Y(m)+60(m)
X(m)+600(m)
3dplot of rotation combinatorial scatter around the adjusted values
20
c(")
10
0
10
8
6
4
2
0
20
15
4
b(")
10
5
6
0
8
5
a(")
Figure 5.3: Scatter of the 36 computed translations and rotations around the adjusted values
Chapter 6
Summary and Conclusions

In summary, Chapter 3 has demonstrated the power of the algebraic computational tools of Grbner bases and the Multipolynomial resultants in solving selected geodetic problems. In the case of the minimal combinatorial set, we have
demonstrated how the problems of three-dimensional resection, minimum distance mapping and the pseudo-ranging
four-point problem could be solved in a closed form using either Grbner bases approach or the Multipolynomial resultants approach. We have succeeded in demonstrating that by converting the nonlinear observation equations of the
selected geodetic problems above into algebraic (polynomials), the multivariate system of polynomial equations relating the unknown variables (indeterminate) to the known variables can be reduced into polynomial equations consisting
of a univariate polynomial once lexicographic monomial ordering is specified. We have therefore managed to provide
symbolic solutions to the problems of three-dimensional resection, minimum distance mapping and the GPS pseudoranging four-point P4P by obtaining in each case a univariate polynomial that can readily be solved numerically once
the observations are available. Although the algebraic techniques of Grbner bases and Multipolynomial resultants
were tested for these selected geodetic problems, the tools can be used to solve explicitly any closed form problem
in Geodesy. The only limitation may be the storage and computational speed of the computers when compounded
with problems involving many variables. The ability of the algebraic tools (Grbner bases and the Multipolynomial
resultants) to solve closed form solutions gave the Gauss-Jacobi combinatorial algorithm the required operational
engine as evidenced in the solution of the overdetermined GPS pseudo-ranging problem which was achieved without
reverting to iterative or linearization procedures. The results compared well with the solutions obtained using the
linearized Gauss-Markov model giving legitimacy to the Gauss-Jacobi combinatorial procedure.
By solving the test network Stuttgart Central both explicitly (using the Grbner bases and the Multipolynomial
resultants algorithms) and in overdetermined form (using the Gauss-Jacobi combinatorial algorithm) in Chapter 4,
the study has highlighted the capability of the Grbner basis and the Multipolynomial resultant algorithms to solve
explicitly both the Grunert distance equations and the three-dimension ranging (Bogenschnitt) problem as well as
the capability of the Gauss-Jacobi combinatorial algorithm to solve the overdetermined three-dimensional resection.
The deviation of the computed distances fr^6#')R$ 0 $%4 and position 5pf=`$-fr$%f 8 from the real values (Tables
4.1 and 4.2 respectively) obtained using closed form procedures (Grbner basis and the Multipolynomial resultant)
were in millimeter range for distances and position as depicted in Tables (4.14) and (4.18) respectively.
The application of the Gauss-Jacobi combinatorial algorithm to solve the overdetermined three-dimensional resection
problem based on the test network Stuttgart Central improves the results as seen from the deviation of the computed
distances fr^@#r' )+$00 $%4 and position 56f($%fr$%f 8 from the real values (Tables 4.1 and 4.2 respectively)
as depicted in Tables (4.17) and (4.20). In comparing the deviations of the computed values of position of station
K1 using both closed form and Gauss-Jacobi combinatorial algorithms in Tables (4.18) and (4.20), It is evident that
the deviations from the Gauss-Jacobi combinatorial algorithm were smaller. This becomes clearer once we plot
these deviations as in Figure (4.2). One therefore concludes that better accuracies would be achieved if the threedimensional resection problem were to be solved in an overdetermined form as opposed to closed form procedures
which are traditionally used. The Gauss-Jacobi combinatorial algorithm offers the solution to the problem of lacking
the approximate values normally required by the linearized Gauss-Markov model and iterative procedures. Besides,
the Gauss-Jacobi combinatorial algorithm makes use of the full information of the observational data set through the
error propagation/variance-covariance propagation.
For the case studies, the Grbner bases algorithm successfully determines explicitly the 7-parameters of the datum
transformation problem and shows the scale parameter to be represented by a univariate polynomial of fourth order
while the rotation parameters are represented by linear functions comprising the coordinates of the two systems and
95
96
CHAPTER 6. SUMMARY AND CONCLUSIONS
the scale parameter. The admissible value of the scale is readily obtained by solving the univariate polynomial and
restricting the scale to a positive real value s This eliminates the negative components of the roots and the
complex values. The admissible value of the scale parameter is chosen and substituted in the linear functions
that characterize the three elements of the skew-symmetric matrix ^ leading to the solution of the elements of the
rotation matrix D . The translation elements are then deduced from the transformation equation. The advantage of
using the Grbner bases algorithm is the fact that there exists no requirement for prior knowledge of the approximate
values of the 7-transformation parameters as is usually the case.
The Grbner bases algorithm managed to solve the Minimum Distance Mapping problem and in so doing, enabling
the mapping of points from the topographical surface to the International Reference Ellipsoid of Revolution. The
univariate polynomial obtained was identical to that obtained by E. Grafarend and P. Lohse (1991). This implies
that the algebraic tools of Grbner bases and the Multipolynomial resultants can also be used to check the validity of
existing closed form procedures in Geodesy .
The Gauss-Jacobi combinatorial algorithm highlighted one important fact while solving the overdetermined 7 parameter transformation problem; that the stochasticity of both systems involved can be taken into account. This has been
the bottleneck of the problem of 7-datum transformation parameters.
In conclusion, the present study has contributed towards the solution of nonlinear GPS/LPS observations in Geodesy.
By testing the Grbner bases and the Multipolynomial resultants techniques on the selected geodetic problems and
case studies, the study has established that the methods are suitable tools to be applied in solving closed form problems in Geodesy. The only requirements is that the nonlinear geodetic observation equations have to be converted
into algebraic (polynomial) form. Besides the solution of nonlinear equations in closed form, the Grbner bases and
the Multipolynomial resultants approaches can be used to check the validity of the existing closed form procedures
in Geodesy. The study has further established that the Gauss-Jacobi combinatorial algorithm offers an alternative
procedure to iterative and linearization approaches that normally require approximate starting values. With these advantages, the overdetermined problems of GPS pseudo-ranging, three-dimensional resection and 7 parameter datum
transformation were solved. Besides, the Gauss-Jacobi combinatorial algorithm takes into account the full information of the observations and parameter space via the nonlinear error propagation/variance-covariance propagation.
Appendices
Appendix A-1:
Definitions
To enhance the understanding of the theory of Grbner bases presented in Chapter 2, the following definitions supplemented with examples will help in developing ideas that lead to the definition of the basis of an Ideal, in particular
the Standard Basis known as the Grbner basis of an Ideal in a polynomial ring.
Definition A-1 (monomial):
A monomial is a multivariate product of the form R
R 0 URTS $:#K $ 0$;K U B U

{
in the variables
$ 0$? U
Example A-1 (E. Grafarend, P. Lohse and B. Schaffrin 1989, p.128):

Consider equation (1.30) used for the solution of distances in the three-dimensional resection problem given
below as
{ ,J { 9 {
{{ ,R {- {
V { , %
3p
with the variables
5 $? { ?$ $ 8
W{ - 'KP
{ { 2 'P
{ 2 'NP
$J { ~ $2
the other terms being known constants, then
is said to be a set of monomials in the variables
509$? { 2$ 8
W{
$2 { 2$ W{{ ?$ { ?$ { 2$ X
Example A-2 (E. Grafarend and J. Shan 1996, p.138):

Consider the pseudo-ranging four-point problem for determining the unknown coordinates of the stationary receiver and the stationary receiver range bias given by equation (1.6) as follows
:C
:C
:C
:C9
[ B?{
[ B?{
[ {B{
[ B?{
<: { [ %6BG{ :< [ 7B?{ [ :<

:< { [ BG{ :< [ B?{ [ :<
:< { [ { B { :< [ { B { [ :<
:< V { [ BG{ :< [ B?{ [ :<3
W p p$2 $2 $?Wy
:C $% $- B!':< { $ > $;@ BZYN
:CJ6$%06$- B!':< $ > $;@ BZYN
:C { $% { $- { B!':<{R$ > {$;@{0BZYNy{
:C $% $- B!':<R$ > $;@0BZYNy
[ +6B?{E'NP
[ B?{E'NP
[ { B { 'NP
[ B?{E'NP
with W $- $2y{R$2y[X being the position of the four GPS satellites, their ranges to the stationary receiver at
given by 5+R$- $2 $- 8 . The parameters :5++$-%+$27 8 , 5 $% $2 8 , 5 $% $- 8 , 5 $- $- 8 , 5 ++$- $2 $- 8 B
{
{ { {

{
are elements of the spherical cone that intersect at to give the coordinates 50 $2 $? 8 of the receiver and the
{
stationary receiver range bias The equations above can be expanded as follows
W{
{
W{
W{
[ ,0
[ ,
[ , {
[ ,
with the variables 5 9$? $?
{
W {{
{{
W{{
W{{
$23 8
[ ,R%0
[ ,R
[ ,R {
[ ,R
{ W {{ [[
{ W{ [
{ W{ [
{
,R7
,R
,R {
,R
[[ {{
[ {
[ {

{
{
{{
{
{ { 'P
{ { 'P
{{ {{ 'P
{ { 'P
, the other terms being known constants, then W W{ $?p$2W{ $? $?{ 2$ $2W{ $?W X
{ {
said to be a set of monomials in the variables 5 6$2 $? $?WR 8
{
97
,+0
,
, {
,
7{
{
7{{
7{
is
Example A-3 (E. Grafarend and P. Lohse 1991, equation 1(3), p.93):
Consider equation 1(3) used to map the topographical surface point embedded into a three-dimensional Euclidean space @ onto a (reference) ellipsoid of revolution subject to the constrain that the projection point is a
point on the (reference) ellipsoid of revolution. Equation 1(3) given by
[ C:
[ :X
[ :u
{W{
[ B
[ {B
[ B
7{0 W{{
7{
7 { {
{7
{{ [
'KP
' P
'NP
{67{E'KP
where :<$-$7EB are known topographical coordinates from e.g. GPS, three-dimensional resection e.t.c., 56$- 8 the
semi-major and semi-minor axis respectively. Desired are the ellipsoidal Cartesian coordinates 50 'K$? ' > $? '\@
{
of the projected topographical point. 5 $? $2 $? R8 in this case are the variables whose monomials are given as
{
W { $2 $2 $?{ $? $2 $?W{ $? $2 X .
{ { {
Definition A-2 (polynomial):

A polynomial 9$0 $2 U in 9$0 $2 U with coefficients in the field
monomials with pairwise different terms expressed as
'^]
is a finite linear combination of
R R $ R 9$ R ':C R $ 0$? RTS B$_K':#K $ 0$;K U B
Example A-4: The equations
{ R, { {
{{ ,+ {- {
{ ,R
W{ .2U'P
W{ { -U'KP
W{ 2'P
in Example A-1 in page 97 are multivariate polynomials with the first equation being a multivariate polynomial
in two variables 50p$2 8 and a linear combination of the monomials W { $29% $2W{ X The second equation is a
{ {
{
multivariate polynomial in two variables 5 $? 8 and a linear combination of the monomials W { $2 $?W{ $ X
{
{ {
while the third equation is a multivariate polynomial in two variables 5 $2 8 and a linear combination of the
monomials W W{ $? $?{ NX
Example A-5: The four Equations
{
{
{
{
[ ,R
[ ,R
[ ,R { 9
[ ,R 9
W {{ [
W{{ [
W{{ [
W{{ [
,+%
,+
,+ {
,+
{ {{
{ {
{ {
{
[ ,7
[ ,
[ , {
[ ,
[[ WW {{
[ W{
[ W{

,+0
,
, { W
, W
{
{
{{
{
7{
7{
7{{
7{
{
{
{{
{
{
{
{{
{
'KP
' P
K
' P
K
' P
K
presented in Example A-2 are multivariate polynomials with each equation being a multivariate polynomial in four
variables 596$2 $? $?W 8 and a linear combination of the monomials W W{ $?9$?{ $2 $?W{ $? $2W{ $?3 X .
Example A-6: The four Equations
{ {
[ C:
[ :X
[ :u
{W{
[ B
[ {B
[ B
7{0W{{
7{7W'KP
7 { { 3'P
{7 'NP
{{ [ {67{E'KP
presented in Example A-3 of page 98 are multivariate polynomials with the first equation being a multivariate polynomial in two variables 5 $? R8 and a linear combination of the monomials 5 $2 +8 . The second
equation is a multivariate polynomial in two variables 5 $? 8 and a linear combination of the monomials
{ in two variables 50 $23 8 and a linear combina5 { $2 { 3 8 . The third equation is a multivariate polynomial
tion of the monomials 5 $? 3 8 , while the fourth equation is a multivariate polynomial in three variables
5 6$2 $? 8 and a linear combination of the monomials W { $2W{ $2W{ X .
98
Having defined the terms monomial and polynomial above, we next define the term polynomial ring upon
which the Grbner basis is computed using the B. Buchberger algorithm. In computing the Grbner bases, one
computes the Grbner basis of an Ideal ( ` p$ 0 ? U ) which belongs to a Polynomial ring of the field . First
the term "ring" is defined in order to understand what the term "Polynomial ring" means. We begin by considering the
definition of linear Algebra. For literature in linear Algebra we refer to T. Becker and V. Weispfenning (1993 Chapter
1, pp.1-60, 1998).
Definition A-3 (linear algebra):
Algebra can be defined as a set ^ of elements and a finite set a of operations. In linear algebra the elements
of the set ^ are vectors over the field of real numbers, while the set a is basically made up of two elements
of internal relation namely additive and multiplicative. An additional definition of the external relation
expounds on the term linear algebra as follows: A linear algebra over the field of real numbers consists of a
set b of objects, two internal relation elements (either "additive" or "multiplicative") and one external relation
as follows
:XPWpB-E'(gcKgdb e
b ? b
:XP WpB { '(g L g e
b ? b6fb N ? b
:XPWpB '(g:M`gdb ^b ? b
The three cases are outlined as follows:
b as a linear space in a vector space over $ an Abelian group

g
$i&h $;
j 6b
* With respect to the internal relation K (join),

written additively or multiplicatively:
Axiom
Additively
written
g Abelian
g group
Associativity
Identity
K!: $ih|g B!'(g h

gg : h|B j('
' :$h jB
(additive associativity)
g
g
m , g po '
d
m)
Inverse
my
Commutativity
m.
(additive identity,
neutral
g element)
g
g : [ B' o
(additive
g inverse)
g h 'th
Kk : $ih|gnBk '(g k h
m) gUgn: k h|k B j='
'
:#h j+B
(multiplicative
k gnk associativity)
g
m , g
d
)c'
(multiplicative identity,
neutral
element)
k gnkq
g
my
'sr
(multiplicative
inverse)
k gnk
kqg
m.
(additive commutativity,
Abelian axiom)
Multiplicatively
written
g Abelianglgroup
k
h \
' h
(multiplicative commutativity,
Abelian axiom)
with the triplet of axioms 5<m) $;md, $;my 8 or 5<m) $Amd, $;my 8 constituting the set of group axioms and 5<m.
the Abelian axioms. Examples of groups include
(a) The group of integer under addition.
(b) The group of non-zero rational number u under multiplication.
(c) The set of rotation about the origin in the Euclidean plane under the operation of composite function.
* With respect to the external relation L the following compatibility conditions are satisfied
D3: 1
'
distr.
w)
distr.
ws,
)c'
$ib
h 6g br$.7$;vg
L:C7$ B'(g
g&g : hBY'g : |h B&~Y'

' &8h' ~ h ~
K
1st additive gdistributivity
g
Ug :<g vB g ' g <: g vBY'

' & `
v '
8
~
v
2nd additive distributivity
:#9pnb#F.3# #u > B

99
w)
gnk
gnk
|g & g : k h|BY'
gnk : hB&
':C! B
h '
:#hb?B
1stk multiplicative
k
g distributivity
g
k
ws,
gUk :C v9Bg ' g :< k vB

' #: v B': ~? B v
2nd multiplicative distributivity
$;m. 8
* With respect to the internal relation M (meet) the following conditions are satisfied
gnx
i$ hg $;j@8br$.gn
x
M: $ih|B'(g
h
gnx
Axiom
Ass.
dist.
dist.
x{k gnx
k
gnx
k
w) n
g yk
x :$h j+BY
g|'
k : x h|B j
: |h B (
j '
:#h jB
dist.
ws,
m)
w)
gU:
h|B
j('
gnx
:$h
glx
gnx
:$x h j+gnB!'x
h x
: h B (
j '
j h j
g zy
gn x y
gnx
Comments
Associativity w.r.t
internal multiplication
jB
j
Left and Right

additive distributivity w.r.t
internal multiplication
Left and Right
multiplicative distributivity
w.r.t internal multiplication
g x
: h|g|BYx':C! B h
|h B&~Y'
:#hb?B
Left and Right distributivity

of internal and external
multiplication
Definition A-4 (ring):

A sub-algebra is called a ring with identity if the following two conditions encompassing (seven conditions)
hold:
(a) The set b is an Abelian group with respect to addition, i.e.four conditions 5<m) $;md, $;my $;m. 8 of
Abelian group hold.
x
x
(b) The set b is a semi-group with respect to multiplication; that is, 5<m) $Amd, 8 holds. In other words, the set
b comprises a monoid (i.e. a set with two operations, associativity and identity with respect to multiplication).
x
The last condition is the left and right additive distributivity with respect to internal multiplication 5<w) 8
which connects the Abelian Group and the monoid. b In total the four conditions forming the Abelian group (a)
and the three forming the semi-group in (b) add up to form seven conditions enclosed in a ring in Figure 6.
md,
m)
m)
my
m.
x
md,
Figure A1:
Ring
w)
Condition md, makes b a ring with identity while the inclusion of my makes the ring be known as the division
ring if every non-zero element of the xring has a multiplicative inverse. The ring becomes a commutative ring if it
has the commutative multiplicative m. . Examples of rings include:
* Field of real numbers , complex numbers C and rational numbers u . In particular, a ring becomes a field if
every non zero element of the ring has an inverse.
* Integers
100
* Polynomial function in variables over a ring b expressed as '}br 9p$0 $2 U .
In the present study, we will consider a ring to be commutative and to include identity element. The field will be
used in subsequent definition to refer he field of an arbitrary ring 5$PC@$PuO 8 Having defined a ring, we next
expound on the term Polynomial ring
Definition A-5 (polynomial ring):
~ b , a univariate polynomial is formed by assigning
Let us consider a ring b . If we consider an indeterminate p8
coefficients b to the indeterminate and obtaining the summation over finite number of distinct integers.
Thus
:<BY'e R R -$ R 6br$AK
P
R
is said to be a univariate polynomial over
{ <: 9B=' 2 $
b . If two polynomials be given such that
+p:<9B'
and
then two binary operation addition and multiplication can be defined on these polynomials
R6:CB { :<9B' 2$ 'K $2 6b

(b) Multiplicative: :<9B7t :CB'e 0p -$ 0y'
{
; u2R$2 db .
A collection of polynomials with these additive and multiplicative rules forms a commutative ring with zero
element and identity 1. A univariate polynomial :<9B obtained by assigning elements belonging to the ring
b to the variable 5 8 is called a polynomial ring and is expressed as :<BY}
' b In general the entire collection
of all polynomials in 6$ 0$? U with coefficients in the field that satisfy the definition of a ring above such
that the operations addition and multiplication can be carried out is called a polynomial ring. Designated &$
polynomial ring is represented by unknown variables over and is expressed as gh'
p$ 0$? U
Its elements are polynomials known as univariate when 'i) and multivariate otherwise. To complete the
such that:
(a) Addition:
definition of a polynomial ring, we distinguish it from a polynomial. A polynomial, belonging to a polynomial

ring, is the sum of a finite set of monomials (see Definition A-1, page 97 for the definition of a monomial). A
complete definition of a polynomial is given in Definition A-2, page 98.
Example A-7: Consider the equations of Example A-1 in page 97. The three equations
W{ , J { 9 { {{ - N
' P
W{{ ,R {- { { - 'P
W{ , % W{ - 'KP
are said to be polynomial elements of the polynomial ring in three variables

numbers expressed as gt'N 6$2 $?
.
9p$? { 2$ 7
over the field of real
{
{
{{
{
'NP
' P
N
' P
N
' P3
N
Example A-8: Consider equations of Example A-2 expressed below
{
{
{
{
[ ,
[ ,
[ , {
[ ,
{{
{{
{{
{{
[ ,+%0
[ ,+
[ ,+ {
[ ,+
{ WW{{
{ W{
{ W{
{
[ ,7
[ ,
[ , {
[ ,
[[ WW {{
[ W{
[ W{

,RR
,R
,R {
,R
{
{
{{
{
{
{
{{
{
{
{
{{
{
The four equations are said to be polynomial elements of the polynomial ring in four variables
the field of real numbers expressed as gh'N $? $2 $? .
$? { 2$ 2$
over
Example A-9: Consider equation in Example A-3, page 98 expressed below
[ <:
[ :C
[ :F
7{{
[ B
[ {B
[ B
{ W{{
7{093c'P
7 {0 { 3'NP
{ 3
'KP
{{ [ {7{E'PJ
The four equations are said to be polynomial elements of the polynomial ring in four variables
over the field of real numbers expressed as gh' 9$? $? $23
101
9$? { 2$ ?$ 3
In the case of univariate polynomials, the format of presentation is usually that of either ascending or descending orders
with respect to the power of the variable involved. The situation is however complicated in the case of multivariate
polynomials where an ordering system has to be defined. We mention here the three commonly used monomial
ordering systems namely; lexicographic ordering, graded lexicographic ordering and graded reverse lexicographic
ordering. First we define the monomial ordering before considering the three types of monomial ordering.
Definition A-6 (monomial ordering):
U
A monomial ordering on 96$00$? U is any relation > on or equivalently any relation on the set R
U
satisfying the following conditions:
U
(a) is total (or linear) ordering on
U
(b) If K
L and M` , then K MpL M
U
(c) > is a well ordering on .
U
This condition is satisfied if and only if every strictly decreasing sequence in eventually terminates.
$K
Definition A-6a (Lexicographic ordering):

This is akin to the ordering of words used in dictionaries. If we define a polynomial in three variables as
' $ > $;@ and specify an ordering > }@ , then any term with will supersede that of > which in tern
supersedes that of @3 If the powers of the variables for respective monomials are given as K':#K|p$ 0$;K U B and
U
U
L ' :$L6$ 0$iL U B$;K!$iLK $ then KG{ L if in the vector difference K [ LNb $ the most left non-zero
entry is positive. For the same variable (e.g. ) this subsequently means R
Example A-10: > @[ is an example of lexicographic ordering.
As a second example, consider the polynomial ',W{ > [
order; ' [ > @ ,pW{ > [ > > @ > @ .
> @ > @+ [ > , we have the lexicographic
Definition A-6b (Graded lexicographic ordering):

In this case, the total degree of the monomials is taken into account. First, one considers which monomial has
the highest total degree before looking at the lexicographic ordering. This ordering looks at the left most (or
U
largest) variable of a monomial and favours the largest power. Let K!$iL $ then K{ L if K&'
U
K L!' L m K' L! , and K{ LY$ in K [ L $ the most left non zero entry is positive.

Example A-11: > E@{ > {E@Y:CJ$-J$%,RB8 :C13$-,J$2+B$;2lk:XJ$-J$-,+B7W') 6L:C13$-,.$-+B7' )+)nlWq1K [

Lb'L:F,.$0)+$ [ )6B . Since the left most term of the difference (2) is positive, the ordering is graded lexicographic.
As a second example, consider the polynomial ' ,W{ > [ R > @ > @+ [ > $ we have the graded
lexicographic order; ' [ R > @ ,W{ > [ > > @ > t@ .
Definition A-6c (Graded reverse lexicographic ordering):

In this case, the total degree of the monomials is taken into account as in the case of graded lexicographic
ordering. First, one considers which monomial has the highest total degree before looking at the lexicographic
ordering. In contrast to the graded lexicographic ordering, one looks at the right most (or largest)
U variable of
a monomial and favours the smallest power. Let
K!$iL
U $
then
K P -P
if
K&' K L!'
m K&+' L , and K{P -P L , and in K [ L U the right most non zero entry is negative.

Example A-12: > E@{{ P > {E@Y:CJ$-J$%,RBz{ -P :C13$-,J$2+B$;2lk:XJ$23$-,+B+') 4:C13$-,J$2+B0R')+)nlWq1K [

L ' :F,.$0)+$ [ )6B . Since the right most term of the difference (-1) is negative, the ordering is graded reverse lexi

cographic. As a second example, consider the polynomial ' ,p{ > [ R > @ > @+ [ > , we have the
graded reverse lexicographic order: ',W{ > [ > @ [ > > @ > p@ .
L

If we consider a non-zero polynomial

additional terms can be defined:
' R R R
Definition A-7
Multidegree of : Multideg ( )=nZ9:IK
in
$00 $2 U
P+B
R 'K
Leading coefficient of : LC ( )= [ D -+<
fE D E-E< (with coefficient 1)
Leading monomial of : LM ( )=
102
and fix the monomial order, the following
Leading term of : LT ( )=LC ( ) LM ( )

Example A-13: Consider the polynomial
50 > @ 8 , we have
Multideg ( )=(5,1,4)
LC ( )= -3

LM ( )= > @

LT ( )= [ > @
',pW{ > [ R > @ > @+ [ >
with respect to lexicographic order
The definitions of polynomial ordering above have been adopted from D. Cox et. al. (1997 pp.52-58). Most computer
algebra makes use of the lexicographic ordering (D. Cox et al. 1997 pp. 56-57, J. H. Davenport et al. 1988 p. 72).
The present study restricts itself to lexicographic ordering.
103
Appendix A-2:
Simplexes for Stuttgart Central

Figure A2: Combinatorial simplexes for the for test network Stuttgart Central
104
Figure A3: Combinatorial simplexes for the for test network Stuttgart Central continues
105
106
107
Figure A6: Combinatorial simplexes for the test network Stuttgart Central continues
Appendix A-3:
Error propagation
Nonlinear random effect model (univariate)

Consider a function
specified as follows:
Case 1
>
'*:I@B
where
> is a scalar valued observation and @ a random effect. Three cases can be
(= assumed to be known):
By Taylor series expansion we have
:I@BY':$=6B ) ) :$=6B:I@ [ =6B ,) :#=pB:#@ [ =6B { p :CB

) :$=6B75:#@ [ =6 B { 8 :X+B
75 > 8 '}756:I@B 8 ':$=6B
,
leading to (cf. E. Grafarend and B. Schaffrin 1983, p.470)
75: >=[ 75 >

hence 75 >=[
75 >
75 >
F
8 ':$ B { #: B?H{ p :C+By
8 B { 8 '}75 :#=B0:#@ [ =pB , ) $: =pB:I@ [ = B {

:X+B [
[ ,) :$ BGH { [ :X+B { 8 $
8 >d[ 75 > 8 8 is given by

H { ' { :$ B?H { [ ) { :$ BGH :# Bi75:I@ [ B 8
) { 75:#@ [ B 8 :X+B7
Finally if @ is quasi-normally distributed, we have from E. Grafarend and B. Schaffrin (1983, p.468) that
'}75:I@ [ BG 8 'NP and '}75:#@ [ B 8 'N { 'NRH leading to
{
F
H{ ' {+:$=pBGH{ { {+:$=pBGH
X: +B
108
Case 2
(= unknown, but known as a fixed effect approximation (this implies in E. Grafarend and B. Schaffrin
= )):
1983, p.470 that 'G
:#@BY'N:# B ) ) :$ B:I@ [ B , ) #: B:#@ [ B {

:X+B
using
't
:$0 [ Bd3@ [
'\@ [ :$ [ 6B
:#@BY'N :# B :# B0:#@ [ =pB :# B:$= [ B
{P $: BI: @ [ =6BG{ { :$ B0:$= [ BG{
#: B:#@ [ =6B:#= [ B :X+B
we have
and
8 '\75p:#@B 8 'N:#6B :$6B:# [ B ,) :$0BGH {

,) :$pB:# [ B { p :C+B
leading to 75 >=[ 75 > 8 >d[ 75 > 8 8 as
H { ' { $: BGH { #: iB 75I: @ [ =B 8 ,R :$ B?H { :#= [ B

{ :#pB7 5 :#@ [ B 8 { :$06Bi75:I@ [ B 8 :# [ B [
[ { $: pBGH { #: pBGH { $: [ 6B { p :CB
and with @ being quasi-normally distributed, thus s

' 75#: @ [ B? 8 'P and '75:I@ [ B 8 '
{ 'RH $ we have
{
75 >
F
H{ ' {+:$6BGH{ ,R #: 6B?H{ :# [ pB { { :$0BGH R{ :#pBGH{ :$ [ 6BG{ p :CB
with the first and third terms (on the right hand side) being the right hand side terms of case 1 (cf. E. Grafarend and
B. Schaffrin 1983, p.470).
Case 3
( unknown, but @ known as a random effect approximation):

:#@B|':#=B ) ) :$=pB:I@ [ = 6B , ) :#=B0:#@ [ = pB {

_) :#=p B:#@ [ =6B p #: B
changing
and the initial bias

leads to
[ =c'}@ [ ='G@ U[ 75<@ 8 [ :$= [ 75:@ 8 B

[ :$= [ 75<@ 8 BY'\75<@ 8 [ =c'(g'L
7 5:@ 8 WL
[ '\@ [
If we also change :$ B'N:I@ 6B :#@6B:$ [ @ B :X,+B7$p :# BY' #: @B :I@ 6B:# [ @ B

:X+B
etc, the derivatives became random effects and cannot any longer be separated in 756 :$ B :I@ [ B 8 '
:$ B75:#@ [ B 8 etc. Consider
#: @ [ B { '#: @ [ 7<5 @ 8 B { L { ,JI: @ [ 7:5 @ 8 B L3
we have
#@B|':# B :$ B0:#@ [ 75<@ 8 B :# BL3

:
{ :$ B :I@ [ 75:@ 8 B?{ { :# BL { :# B0:#@ [ 75<@ 8 B-LW :X+B

)
)
75 > 8 '$: =6B :#=6BL :$=pBGH i{ :#=BL { p :CB

,
,
109
8 >d[ 75 > 8 8 as
H{ 'N {+:$ BGH{ :$ B75:#@ [ 75<@ 8 BG 8
,R :#=B?Hi{ L {+:$=pB75:#@ [ 75<@ 8 B 8
{R:# B75:#@ [ 75<@ 8 BG 8 L3 {+:$ BGHi{ L {
{+:$ BGH i [ { {R:# B75:#@ [ 75<@ 8 B?{ 8 Hi{ p C: B
and with @ being quasi-normally distributed, thus '75:I@ [ 75<@ 8 BG 8
' P
75<@ 8 B 8 'N {{ 'NRH $ we have

F
H { ' {R:$=pB?H{ ,R :#=B?H{ L { {+:$=pBGH {+:$= BGH{ L { p :C+B
leading to 75 >=[
75 >
and
'75:#@ [
with the first and third terms (on the right hand side) being the right hand side terms of case 1 (cf. E. Grafarend and
B. Schaffrin 1983, p.470).
Appendix A-4:
C. F. Gauss combinatorial approach
110
111
112
References
Aduol, F. W. O. (1989): Intergrierte geodtische Netzanalyse mit stochastischer Vorinformation ber Schwerefeld
und Referenzellipsoid, DGK, Reihe C, Heft Nr. 351.
Abel, J. S. and Chaffee, J. W. (1991): Existence and uniqueness of GPS solutions, IEEE Transactions on Aerospace
and Electronic Systems 27 (1991) 952-956.
Ashkenazi V. and Grist, S. (1982): Inter-comparisson of 3-D geodetic network adjustment models, International
sypmosium on geodetic networks and computations of the I. A. G. Munich, August 31 to September 5, 1981.
Awange, L. J. (1999): Partial procrustes solution of the threedimensional orientation problem from GPS/LPS observations, Quo vadis geodesia...? Festschrift to E. W. Grafarend on the occasion of his 60th birthday, Eds. F.
Krumm and V. S. Schwarze, Report Nr. 1999.6-1
Awange L. J. and Grafarend, E. (2002). Algebraic solution of GPS pseudo-ranging equations, Journal of GPS
Solution, Volume 5(3), 2002.
Bajaj, C., Garity, T. and Waren, J. (1988): On the applications of multi-equational resultants, Department of Computer Science, Purdue University, Technical Report CSD-TR-826, pp. 1-22, 1988.
Bancroft, S. (1985): An algebraic solution of the GPS equations, IEEE Transaction on Aerospace and Electronic
Systems AES-21 (1985) 56-59.
Bhr, H. G. (1988): A quadratic approach to the non-linear treatment of non-redundant observations Manuscripta
Geodaetica, 13 (1988) 191-197.
Bhr, H. G. (1991): Einfach berbestimmtes ebenes Einschneiden, differentialgeometrisch analysiert, Zeitschrift fr
Vermessungswesen 116 (1991) 545-552.
Bartelme, N. and Meissl, P. (1975): Ein einfaches, rasches und numerisch stabiles Verfahren zur Bestimmung des
krzesten Abstandes eines Punktes von einem sphroidischen Rotationsellipsoid, Allgemeine VermessungsNachrichten 82 (1975) 436-439.
Becker, T. and Weispfenning, V. (1993): Grbner bases. A computational approach to commutative algebra. Graduate Text in Mathematics 141, Springer-Verlag, New York 1993.
Becker, T. and Weispfenning, V. (1998): Grbner bases. A computational approach to commutative algebra. Graduate Text in Mathematics 141, 2nd Edition, Springer-Verlag, New York 1998.
Benning, W.(1974): Der krzeste Abstand eines in rechtwinkligen Koordinaten gegebenen Auenpunktes vom Ellipsoid, Allgemeine Vermessungs-Nachrichten 81 (1974) 429-433.
Benning, W.(1987): Iterative ellipsoidische Lotfupunktberechung, Allgemeine Vermessungs-Nachrichten 94 (1987)
256-260.
Blaha, G. and Besette, R. P. (1989): Nonlinear least squares method via an isomorphic geometrical setup, Bulletin
Geodesique 63 (1989) 115-138.
Borre, K. and Lauritzen, S. L. (1983): Some geometric aspects of adjustment, In. Festschrift to Torben Krarup, Eds.
E. Kejlso et al., Geod. Inst., No. 58, pp. 70-90.
Borg, I. and Groenen, P. (1997): Modern multidimensional scaling, Springer-Verlag, New York 1997.
113
Borkowski, K. M. (1987): Transformation of geocentric to geodetic coordinates without approximation, Astrophys.

Space. Sci. 139 (1987) 1-4.
Borkowski, K. M. (1989): Accurate algorithm to transform geocentric to geodetic coordinates, Bull. Geod. 63 (1989)
50-56.
Bowring, B. R.(1976): Transformation from spatial to geographical coordinates, Survey Review 23 (1976) 323-327.
Bowring, B. R.(1985): The accuracy of geodetic latitude and height equations, Survey Review 28 (1985) 202-206.
Brandsttter, G. (1974): Notiz zur analytischen Lsung des ebenen Rckwrtsschnittes, sterreichische Zeitschrift
fr Vermessungswesen 61 (1974) 134-136.
Brokken, F. B. (1983): Orthogonal Procrustes rotation maximizing congruence, Psychometrika 48 (1983) 343-352.
Brunner, F. K. (1979): On the analysis of geodetic networks for the determination of the incremental strain tensor,
Survey Review 25 (1979) 146-162.
Buchberger, B. (1965): An algorithm for finding a basis for the residue class ring of a zero dimensional polynomial
ideal (German), Ph.D. thesis, University of Innsbruck, Institute of Mathematics.
Buchberger, B. (1970): Ein algorithmisches Kriterium fr die Lsbarkeit eines algebraischen Gleichungsystems,
Aequationes Mathematicae 4 (1970) 374-383.
Buchberger, B. (1979): A criterion for detecting unnecessary reductions in the construction of Groebner bases.
Proceedings of the 1979 European Symposium on Symbolic and Algebraic computation, Springer lecture notes
in Computer Science 72, Springer-Verlag, pp. 3-21, Berlin-Heidelberg-New York 1979.
Buchberger, B. (1999): Personal communication.
Buchberger, B. and Winkler, F. (1998): Groebner bases and applications. London mathematical society lecture note
series 251, Cambridge university press, 1998.
Canny, J. F. (1988): The complexity of robot motion planning, ACM doctoral dissertation award, MIT Press, 1988.
Canny, J. F., Kaltofen, E. and Yagati, L. (1989): Solving systems of nonlinear polynomial equations faster, Proceedings of the International Symposium on Symbolic and Algebraic Computations ISSAC, July 17-19, Portland,
Oregon 1989. pp.121-128.
Cattani, E., Dickenstein, A. and Sturmfels, B. (1998): Residues and resultants, J. Math. Sci. University of Tokyo 5
(1998) 119-148.
Chaffee, J. and Abel, J. (1994): On the exact solutions of the pseudorange equations, IEEE Transactions on Aerospace
and Electronic Systems 30 (1994) 1021-1030.
Cheng, C-L. and Van Ness J. W. (1999): Statistical regression with measurement error, Oxford University Press, 198
Madison Avenue, New York 1999.
Chrystal, G. (1964): Textbook of Algebra (Vol. 1), Chelsea, New York 1964.
Cox, D. A. (1998): Introduction to Grbner bases, Proceedings of Symposia in Applied Mathematics 53 (1998) 1-24.
Cox, D., Little, J. and OShea, D. (1997): Ideals, Varieties, and Algorithms. An introduction to computational
algebraic geometry and commutative algebra, Springer-Verlag, New York 1997.
Cox, D., Little, J. and OShea, D. (1998): Using algebraic geometry. Graduate Text in Mathematics 185, SpringerVerlag, New York 1998.
Cox, T. F. and Cox, M. A. (1994): Multidimensional scaling, St. Edmundsbury Press, St. Edmunds, Suffolk 1994.
Croceto, N.(1993): Point projection of topographic surface onto the reference ellipsoid of revolution in geocentric
Cartesian coordinates, Survey Review 32 (1993) 233-238.
Crosilla, F. (1983a): A criterion matrix for a second order design of control networks, Bull. Geod. 57 (1983a)
226-239.
114
Crosilla, F. (1983b): Procrustean transformation as a tool for the construction of a criterion Matrix for control networks, Manuscripta Geodetica 8 (1983b) 343-370.
Dach, R. (2000): Einflu von Auflasteffekten auf Przise GPS-Messungen, DGK, Reihe C, Heft Nr. 519.
Davenport, J. H., Siret, Y. and Tournier, E. (1988): Computer algebra. Systems and algorithms for algebraic computation, Academic Press Ltd., St. Edmundsbury, London 1988.
Dixon, A. L. (1908): The elimination of three quantics in two independent variables, Proc. London Mathematical
Society series 2 6 (1908) 468-478.
Dryden, I. L. (1998): General registration and shape analysis. STAT 96/03. Stochastic Geometry: Likelihood and
Computation, by W. S. Kendell, O. Barndoff-Nielsen and M. N. N. van Lieshout, Chapman and Hall/CRC Press,
333-364, Boca Raton 1998.
Finsterwalder, S. and Scheufele, W. (1937): Das Rckwartseinschneiden im Raum, Sebastian Finsterwalder zum 75
Geburtstage, pp. 86-100, Verlag Hebert Wichmann, Berlin 1937.
Fischler, M. A. and Bolles, R. C. (1981): Random Sample Consensus: A Paradigm for Medell Fitting with Application to Image Analysis and Automated Cartography, Communications of the ACM 24 (1981) 381-395.
Fitzgibbon, A., Pilu, M. and Fisher, R. B. (1999): Direct least squares fitting of ellipses, IEEE Transactions on
Pattern Analysis and Machine Intelligence 21 (1999) 476-480.
Fotiou, A.(1998): A pair of closed expressions to transform geocentric to geodetic coordinates, Zeitschrift fr Vermessungswesen 123 (1998) 133-135.
Frhlich, H. and Hansen, H. H. (1976): Zur Lotfupunktrechnung bei rotationsellipsoidischer Bezugsflche, Allgemeine Vermessungs-Nachrichten 83 (1976) 175-179.
Fukushima, T. (1999): Fast transform from geocentric to geodetic coordinates, Journal of Geodesy 73 (1999) 603610.
Gander, W., Golub, G. H. and Strebel, R. (1994): Least-Squares fitting of circles and ellipses, BIT No. 43 (1994)
558-578.
Gao, X. S. and Chou S. C. (1990): Implicitization of rational parametric equations. Technical report, Department of
Computer Science TR-90-34, University of Texas, Austin 1990.
Gelfand, I. M., Kapranov, M. M. and Zelevinsky, A. V. (1990): Generalized Euler Integrals and A-Hypergeometry
Functions, Advances in Mathematics 84 (1990) 255-271.
Gelfand, I. M., Kapranov, M. M. and Zelevinsky, A. V. (1994): Discriminants, resultants and multidimensional
determinants, Birkhuser, Boston 1994.
Gotthardt, E. (1940): Zur Unbestimmtheit des rumlichen Rckwrtseinschnittes, Mitteilungen der Ges. f. Photogrammetry e.V., Jnner 1940, Heft 5.
Gotthardt, E. (1974): Ein neuer gefhrlicher Ort zum rumlichen Rckwrtseinschneiden, Bildm. u. Luftbildw.,
1974.
Grafarend, E. (1981): Die Beobachtungsgleichungen der dreidimensionalen Geodsie im Geometrie- und Schwereraum. Ein Beitrag zur operationellen Geodsie, Zeitschrift fr Vermessungswesen 106 (1981) 411-429.
Grafarend, E. (1975): Three dimensional Geodesy 1. The holonomity problem, Zeitschrift fr Vermessungswesen
100 (1975) 269-280.
Grafarend, E. (1985): Variance-Covariance component estimation; theoretical results and geodetic applications,
Statistics & Decisions, Supplement IssueNo. 2 (1985) 407-441.
Grafarend, E. (1988): Azimuth transport and the problem of orientation within geodetic traverses and geodetic
networks, Vermessung, Photogrammetrie, Kulturtechnik 86 (1988) 132-150.
Grafarend, E. (1989): Photogrammetrische Positionierung. Festschrift fr Prof. Dr.-Ing. Dr. h.c Friedrich Ackermann zum 60. Geburtstag, Institut fr Photogrammetrie, Univeritt Stuttgart, Heft 14, pp.44-55, Stuttgart
1989.
115
Grafarend, E. (1991): Application of Geodesy to Engineering. IAG-Symposium No. 108, Eds. K. Linkwitz, V.
Eisele, H. J. Mnicke, Springer-Verlag, Berlin-Heidelberg-New York 1991.
Grafarend, E. and Kunz, J. (1965): Der Rckwrtseinschnitt mit dem Vermessungskreisel, Bergbauwissenschaften
12 (1965) 285-297.
Grafarend, E. and Schaffrin, B. (1974): Unbiased Freenet adjustment, Survey Review 22 (1974) 200-218.
Grafarend, E. and Richter, B. (1977): Generalized Laplace condition, Bull. Geod. 51 (1977) 287-293.
Grafarend, E. and Sanso, F. (1985): Optimization and design of geodetic networks, Springer-Verlag, Berlin-HeidelbergNew York-Tokyo 1985.
Grafarend, E. and Schaffrin, B. (1989): The geometry of nonlinear adjustment-the planar trisection problem-,
Festschrift to T. Krarup edited by E. Kejlso, K. Poder and C. C. Tscherning, pp. 149-172, Denmark 1989.
Grafarend, E. and Schaffrin, B. (1991): The planar trisection problem and the impact of curvature on non-linear least
-squares estimation, Computational statistics & data analysis 12 (1991) 187-199.
Grafarend, E. and Lohse, P. (1991): The minimal distance mapping of the topographic surface onto the (reference)
ellipsoid of revolution, Manuscripta Geodaetica 16 (1991) 92-110.
Grafarend, E. and Schaffrin, B. (1993): Ausgleichungsrechnung in Linearen Modellen, B. I. Wissenschaftsverlag,
Mannheim 1983.
Grafarend, E. and Keller, W. (1995): Setup of observational functionals in gravity space as well as in geometry space,
Manuscripta Geodetica 20 (1995) 301-325.
Grafarend, E. and Shan, J. (1996): Closed-form solution of the nonlinear pseudo-ranging equations (GPS), ARTIFICIAL SATELLITES, Planetary Geodesy No 28, Special issue on the 30th anniversary of the Department of
Planetary Geodesy, Vol. 31 No 3, pp. 133-147, Warszawa 1996.
Grafarend, E. and Shan, J. (1997a): Closed-form solution of P4P or the three-dimensional resection problem in terms
of Mbius barycentric coordinates, Journal of Geodesy 71 (1997) 217-231.
Grafarend, E., and Shan, J. (1997b): Closed form solution of the twin P4P or the combined three dimensional
resection-intersection problem in terms of Mbius barycentric coordinates, Journal of Geodesy 71 (1997) 232239.
Grafarend, E. and Okeke, F. (1998): Transformation of conformal coordinates of type Mercator from global datum
(WGS 84) to local datum (regional, national), Marine Geodesy 21 (1998) 169-180.
Grafarend, E. and Syfuss, R. (1998): Transformation of conformal coordinates of type Gauss-Krger or UTM from
local datum (regional, national, European) to global datum (WGS 84) part 1: The transformation equations,
Allgemeine Vermessungs-Nachrichten 105 (1998) 134-141.
Grafarend, E. and Ardalan, A. (1999): World geodetic datum 2000, Journal of Geodesy 73 (1999) 611-623.
Grafarend, E. W. and Awange, J. L. (2000): Determination of vertical deflections by GPS/LPS measurements,
Zeitschrift fr Vermessungswesen 125 (2000) 279-288.
Grafarend, E., Lohse, P. and Schaffrin, B. (1989): Dreidimensionaler Rckwrtsschnitt, Zeitschrift fr Vermessungswesen 114 (1989) 61-67,127-137,172-175,225-234,278-287.
Grafarend, E., Krumm, F. and Okeke, F. (1995): Curvilinear geodetic datum transformation, Zeitschrift fr Vermessungswesen 120 (1995) 334-350.
Grafarend E., Syffus, R. and You, R. J. (1995): Projective heights in geometry and gravity space, Allgemeine
Vermessungs-Nachrichten 102 (1995) 382-402.
Grafarend, E., Heister, H., Kelm, R., Knopff, H. and Schaffrin, B. (1979): Optimierung Geodtischer Messoperationen, Herbat Wichmann Verlag, Karlsruhe 1979.
Green, B. (1952): The orthogonal approximation of an oblique structure in factor analysis, Psychometrika 17 (1952)
429-440.
116
Grewal, M. S., Weill, L. R. and Andrews, A. P. (2001): Global Positioning Systems, Inertial Navigation and Integration, John Wiley & Sons, New York 2001.
Grunert, J. A . (1841): Das Pothenotsche Problem in erweiterter Gestalt; nebst Bemerkungen ber seine Anwendungen in der Geodsie, Grunerts Archiv fr Mathematik und Phsyik 1 pp. 238-241, 1841.
Guckenheimer, J., Myers, M. and Sturmfels, B. (1997): Computing Hopf birfucations, SIAM J. Numerical Analysis
34 (1997) 1-21.
Gulliksson, M. (1995a): The partial Procrustes problem - A first look, Department of Computing Science, Umea
University, Report UMINF-95.11, Sweden 1995.
Gullikson, M. (1995b): Algorithms for the partial Procrustes problem, Department of Industrial Technology, Mid
Sweden University s-891 18, Report 1995:27, Ornskoldsvik, Sweden 1995.
Gullikson, M. and Sderkvist, I. (1995): Surface fitting and parameter estimation with nonlinear least squares,
Guolin, L. (2000): Nonlinear curvature measures of strength and nonlinear diagnosis, Allgemein VermessungsNachrichten 107 (2000) 109-111.
Hanselman, D. and Littlefield, B. (1997): The student edition of Matlab, Prentice-Hall, New Jersey 1997.
Hammer, E. (1896): Zur graphischen Ausgleichung beim trigonometrischen Einschneiden von Punkten, Zeitschrift
fr Vermessungswesen 25 (1896) 611-636.
Han, S. C., Kwon, J. H. and Jekeli, C. (2001): Accurate absolute GPS positioning through satellite clock error
estimation, Journal of Geodesy 75 (2001) 33-43.
Haralick, R. M., Lee, C., Ottenberg, K., and Nlle, M. (1991): Analysis and Solution of the Three Point Perspective
Pose Estimation Problem, Proc. IEEE Org. on Computer Vision and Pattern Recognition, pp. 592-598, 1991.
Haralick, R. M., Lee, C., Ottenberg, K., and Nlle, M. (1994): Review and Analysis of Solution of the Three Point
Perspective Pose Estimation Problem, International Journal of Computer Vision 13 3 (1994) 331-356.
Heck, B.(1987): Rechenverfahren und Auswertemodelle der Landesvermessung, Wichmann Verlag, Karlsruhe 1987.
Heikkinen, M.(1982): Geschlossene Formeln zur Berechnung rumlicher geodtischer Koordinaten aus rechtwinkligen Koordinaten, Zeitschrift fr Vermessungswesen 107 (1982) 207-211.
Hein, G. (1982a): A contribution to 3D-operational Geodesy (Principles and observation equations of terrestrial
type), DGK Reihe B, Heft Nr. 258/VII pp.31-64.
Hein, G. (1982b): A contribution to 3D-operational Geodesy part 2, DGK Reihe B, Heft Nr. 258/VII pp.65-85.
Heiskanen, W. A. and Moritz, H. (1976): Physical Geodesy, Freeman and Company, London 1976.
Hirvonen, R. and Moritz, H. (1963): Practical computation of gravity at high altitudes, Institute of Geodesy, Photogrammetry and Cartography, Ohio State University, Report No. 27, Ohio 1963.
Hofman-Wellenhof, B., Lichtenegger, H. and Collins, J. (1994): Global Positioning System: Theory and practice,
Springer-Verlag, Wien 1994.
Horaud, R., Conio, B. and Leboulleux, O. (1989): An Analytical Solution for the Perspective 4-Point Problem,
Computer Vision, Graphics and Image Processing 47 (1989) 33-44.
Hornoch, A. T. (1950): ber die Zurckfhrung der Methode der kleinsten Quadrate auf das Prinzip des arithmetischen Mittels, sterr. Zeitschrift fr Vermessungswesen 38 (1950) 13-18.
Ireland, K. and Rosen, M. (1990): A classical introduction to modern number theory, Springer-Verlag, New York
1990.
Jacobi, C. G. I. (1841): Deformatione et proprietatibus determinantum, Crelles Journal fr die reine und angewandte
Mathematik. Bd. 22, 1841.
Kahmen, H. and Faig, W. (1988): Surveying, Walter de Gruyter, Berlin 1988.
117
Killian, K. (1990): Der gefhrliche Ort des berbestimmten rumlichen Rckwrtseinschneidens, st.Zeitschrift fr
Vermessungswesen und Photogrammetry 78 (1990) 1-12.
Klein, U. (1997): Analyse und Vergleich unterschiedlicher Modelle der dreidimensionale, DGK, Reihe C, Heft Nr.
479.
Kleusberg, A. (1994): Die direkte Lsung des rumlichen Hyperbelschnitts, Zeitschrift fr Vermessungswesen 119
(1994) 188-192.
Kleusberg, A. (1999): Analytical GPS navigation solution, Quo vadis geodesia...? Festschrift for E. W. Grafarend on
the occasion of his 60th birthday, Eds. F. Krumm and V. S. Schwarze, Report Nr. 1999.6-1.
Koch, K. R. (1999): Parameter estimation and hypothesis testing in linear models, 2nd Edition, Springer-Verlag,
Berlin-Heidelberg 1999.
Krarup, T. (1982): Nonlinear adjustment and curvature, In. Forty years of thought, Delft, pp. 145-159, 1982.
Krause, L. O. (1987): A direct solution of GPS-Type Navigation equations, IEEE Transactions on Aerospace and
Electronic Systems 23 (1987) 225-232.
Krishna, S. and Manocha, D. (1995): Numerical algorithms for evaluating one-dimensional algebraic sets, Proceedings of the International Symposium on Symbolic and Algebraic Computation ISSAC, July 10-12, pp. 59-67,
Montreal, Canada, 1995.
Kubik, K. K. (1967): Iterative Methoden zur Lsunge des nichtlinearen Ausgleichungsproblemes, Zeitschrift fr
Kurz, S. (1996): Positionierung mittels Rckwartsschnitt in drei Dimensionen. Studienarbeit, Geodtisches Institut,
University of Stuttgart, Stuttgart 1996.
Lapaine, M.(1990): A new direct solution of the transformation problem of Cartesian into ellipsoidal coordinates.
Determination of the geoid: Present and future. Eds. R. H. Rapp and F. Sanso, pp. 395-404. Springer-Verlag,
New York 1990.
Leick, A. (1995): GPS satellite surveying, 2nd Edition, John Wiley & Sons, New York 1995.
Lichtenegger, H. (1995): Eine direkte Lsung des rumlichen Bogenschnitts, sterreichische Zeitschrift fr Vermessung und Geoinformation 83 (1995) 224-226.
Lin, K. C. and Wang, J. (1995): Transformation from geocentric to geodetic coordinates using Newtons iteration,
Bulletin Geodesique 69 (1995) 300-303.
Linnainmaa, S., Harwood, D. and Davis, L. S. (1988): Pose Determination of a Three-Dimensional Object Using
Triangle Pairs, IEEE transaction on pattern analysis and Machine intelligence 105 (1988) 634-647.
Lohse, P. (1990): Dreidimensionaler Rckwrtsschnit. Ein Algorithmus zur Streckenberechnung ohne Hauptachsentransformation, Zeitschrift fr Vermessungswesen 115 (1990) 162-167.
Lohse, P. (1994): Ausgleichungsrechnung in nichtlinearen Modellen, DGK, Reihe C, Heft Nr. 429.
Loskowski, P.(1991): Is Newtons iteration faster than simple iteration for transformation between geocentric and
geodetic coordinates? Bulletin Geodesique 65 (1991) 14-17.
Lyubeznik, G. (1995): Minimal resultant system, Journal of Algebra 177 (1995) 612-616.
Macaulay, F. (1902): On some formulae in elimination, Proceeding in London Mathematical Society pp. 3-27, 1902.
Macaulay, F. (1916): The algebraic theory of modular systems, Cambridge Tracts in Mathematics 19, Cambridge
University Press, Cambridge 1916.
Macaulay, F. (1921): Note on the resultant of a number of polynomials of the same degree, Proceeding in London
Mathematical Society 21 (1921) 14-21.
Manocha, D. (1992): Algebraic and Numeric techniques for modeling and Robotics, Ph.D. thesis, Computer Science
Division, Department of Electrical Engineering and Computer Science, University of California, Berkeley 1992.
Manocha, D. (1993): Efficient algorithms for multipolynomial resultant, the Computer Journal 36 (1993) 485-496.
118
Manocha, D. (1994a): Algorithms for computing selected solutions of polynomial equations, Extended abstract
appearing in the proceedings of the ACM ISSAC 94.
Manocha, D. (1994b): Computing selected solutions of polynomial equations, Proceedings of the International Sypmosium on Symbolic and Algebraic Computations ISSAC, July 20-22, pp.1-8, Oxford 1994.
Manocha, D. (1994c): Solving systems of polynomial equations, IEEE Computer Graphics and application, pp.
46-55, March 1994.
Manocha, D. (1998): Numerical methods for solving polynomial equations, Proceedings of Symposia in Applied
Mathematics 53 (1998) 41-66.
Manocha, D. and Canny, J. (1991): Efficient techniques for multipolynomial resultant algorithms, Proceedings of the
International Symposium on Symbolic Computations, July 15-17, 1991, pp. 86-95, Bonn 1991.
Manocha, D. and Canny, J. (1992): Multipolynomial resultant and linear algebra, Proceedings of the International
Symposium on Symbolic and Algebraic Computations ISSAC, July 27-29, 1992, pp. 158-167, Berkeley 1992.
Manocha, D. and Canny, J. (1993): Multipolynomial resultant algorithms, Journal of Symbolic Computations 15
(1993) 99-122.
Mathar, R. (1997): Multidimensionale Skalierung, B. G. Teubner Verlag, Stuttgart 1997.
Mathes, A. (1998): GPS und GLONASS als Teil eines hybrid Mesystems in der geodsie am Beispiel des Systems
HIGGINS, Dissertationen, DGK, Reihe C, Nr. 500.
Mautz, R. (2001): Zur Lsung nichtlinearer Ausgleichungsprobleme bei der Bestimmung von Frequenzen in Zeitreihen, DGK, Reihe C, Nr. 532.
Meissl, P. (1982): Least squares adjustment. A modern approach, Mitteilungen der geodtischen Institut der Technischen Universitt Craz, Folge 43.
Merrit, E. L. (1949): Explicit Three-point resection in space, Phot. Eng. 15 (1949) 649-665.
Mittermayer, E. (1972): A generalization of least squares adjustment of free networks, Bull. Geod. No. 104 (1972)
139-155.
Mller, H. M. (1998): Grbner bases and numerical analysis, In Groebner bases and applications. Eds. B. Buchberger
and F. Winkler, London mathematical society lecture note series 251. Cambridge university press, pp.159-177,
1998.
Morgan, A. P. (1992): Polynomial continuation and its relationship to the symbolic reduction of polynomial systems,
In Symbolic and Numerical Computations for Artificial Intelligence, pp. 23-45, 1992.
Mller, F. J. (1925): Direkte (Exakte) Lsungen des einfachen Rckwarscnittseinschneidens im Raum. 1 Teil,
Zeitschrift fr Vermessungswesen 37 (1925) 249-255, 265-272, 349-353, 365-370, 569-580.
Mukherjee, K.(1996): Robust estimation in nonlinear regression via minimum distance method. Mathematical methods of statistics, Vol 5, No. 1, Allerton Press. Inc., New York 1996.
Musyoka, S. M. (1999): Ein Modell fr ein vierdimensionales integriertes regionales geodtisches Netz, Ph. D.
thesis, Department of Geodesy, University of Karlsruhe (Internet Publication), 1999.
Nicholson, W. K. (1999): Introduction to abstract algebra, Second Edition, John Wiley & Sons, New York-Chichesterweinheim-Brisbane-Singapore 1999.
Okeke, F. I. (1998): The curvilinear datum transformation model, DGK, Reihe C, Heft Nr. 481.
Ozone, M. I.(1985): Non-iterative solution of the equations, Surveying and Mapping 45 (1985) 169-171.
Paul, M. K. (1973): A note on computation of geodetic coordinates from geocentric (Cartesian) coordinates, Bull.
Geod. No. 108 (1973)135-139.
Penev, P.(1978): The transformation of rectangular coordinates into geographical by closed formulas Geo. Map.
Photo, 20 (1978) 175-177.
Perelmuter, A. (1979): Adjustment of free networks, Bull. Geod. 53 (1979) 291-295.
119
Pick, M.(1985): Closed formulae for transformation of Cartesian coordinates into a system of geodetic coordinates,
Studia geoph. et geod. 29 (1985) 653-666.
Pistone, G. and Wynn, H. P. (1996): Generalized confounding with Grbner bases, Biometrika 83 (1996) 112-119.
Pope, A. (1982): Two approaches to non-linear least squares adjustments, The Canadian Surveyor 28 (1982) 663669.
Rao, C. R. (1967): Least squares theory using an estimated dispersion matrix and its application to measurement of
signals, Procedure of the Fifth Barkeley Symposium, Barkeley 1967.
Rao, C. R. (1971): Estimation of variance and covariance components - MINQUE theory, Journal of Multivariate
Analysis 1 (1971) 257-275.
Rao, C. R. (1973): Representation of the best linear unbiased estimators in the Gauss-Markov model with singular
dispersion matrix, Journal of multivariate analysis 3 (1973) 276-292.
Rao, C. R. (1978): Choice of the best linear estimators in the Gauss-Markov model with singular dispersion matrix,
Comm. Stat. Theory Meth. A7 (13) (1978) 1199-1208.
Rao, C. R. and Kleffe, J. (1979): Variance and covariance components estimation and applications, Technical Report
No. 181, Ohio State University, Dept. of Statistics, Columbus, Ohio.
Richter, B. (1986): Entwurf eines nichtrelativistischen geodtisch-astronomischen Bezugssystems, DGK, Reihe C,
Heft Nr. 322.
Rinner, K (1962): ber die Genauigkeit des rumlichen Bogenschnittes, Zeitschrift fr Vermessungswesen 87 (1962)
361-374.
Ritt, J. F. (1950): Differential algebra, AMS colloquium publications.
Runge, C. (1900): Graphische Ausgleichung beim Rckwtseinchneiden, Zeitschrift fr Vermessungswesen 29
(1900) 581-588.
Saito, T. (1973): The non-linear least squares of condition equations, Bull. Geod. 110 (1973) 367-395.
Salmon, G. (1876): Lessons Introductory to modern higher algebra, Hodges, Foster and Co., Dublin 1876.
Schaffrin, B. (1983): Varianz-Kovarianz-Komponenten-Schtzung bei der Ausgleichung heterogener Wiederholungsmessungen, DGK, Reihe C, Heft Nr.282.
Schek, H. J. and Maier, P. (1976): Nichtlineare Normalgleichungen zur Bestimmung der Unbekannten und deren
Kovarianzmatrix, Zeitschrift fr Vermessungswesen 101 (1976) 140-159.
Schnemann, P. H. (1996): Generalized solution of the orthogonal Procrustes problem, Psychometrika 31 (1996)
1-10.
Schwarze, V. S. (1995): Satellitengeodtische Positionierung in der relativistischen Raum-Zeit, DGK, Reihe C, Heft
Nr.449.
Shut, G. H. (1958/59): Construction of orthogonal matrices and their application in analytical Photogrammetrie,
Photogrammetria XV (1958/59) 149-162.
Singer, P., Strbel, D,. Hrdt, R., Bahndorf, J. and Linkwitz, K. (1993): Direkte Lsung des rmlichen Bogenschnitts, Zeitschrift fr Vermessungswesen 118 (1993) 20-24.
Sjberg, L. E.(1999): An efficient iterative solution to transform rectangular geocentric coordinates to geodetic
coordinates, Zeitschrift fr Vermessungswesen 124 (1999) 295-297.
Soler, T. and Hothem, L. D. (1989): Important parameters used in geodetic transformations, Journal of Surveying
Engineering 115 (1989) 414-417.
Strang, G. and Borre, K. (1997): Linear Algebra, Geodesy and GPS, Wellesley Cambridge Press, Wellesley 1997.
Sturmfels, B. (1994): Multigraded resultant of Sylvester type, Journal of Algebra 163 (1994) 115-127.
Sturmfels, B. (1996): Grbner bases and convex polytopes, American Mathematical Society, Providence 1996.
120
Sturmfels, B. (1998): Introduction to resultants, Proceedings of Symposia in Applied Mathematics 53 (1998) 25-39.
Snkel, H.(1999): Ein nicht-iteratives Verfahren zur Transformation geodtischer Koordinaten, ster. Zeitschrift fr
Teunissen, P. J. G (1990): Nonlinear least squares, Manuscripta Geodaetica 15 (1990) 137-150.
Teunissen, P. J. G and Knickmeyer, E. H. (1988): Non-linearity and least squares, CISM Journal ASCGC 42 (1988)
321-330.
Thompson, E. H. (1959a): A method for the construction of orthogonal matrices, Photogrammetria III (1959) 55-59.
Thompson, E. H. (1959b): An exact linear solution of the absolute orientation. Photogrammetria XV (1959) 163179.
Torge, W. (1991): Geodesy. 2nd Edition. Walter de Gruyter, Berlin 1991.
Trefethen, L. N. and Bau, D. (1997): Numerical linear algebra, SIAM, Philadelphia.
Van Der Waerden, B. L. (1950): Modern Algebra, 3rd Edition, F. Ungar Publishing Co., New York 1950.
Vanicek, P. and Krakiwsky, E. J. (1982): Geodesy: The concepts. North-Holland Publishing Company, AmsterdamNew York-Oxford 1982.
Van Mierlo, J. (1988): Rckwrtschnitt mit Streckenverhltnissen, Algemain Vermessungs Nachrichten 95 (1988)
310-314.
Vasconcelos, W. V. (1998): Computational methods in commutative algebra and algebraic geometry, Springer-Verlag,
Berlin-Heidelberg 1998.
Vincenty, T. (1978): Vergleich zweier Verfahren zur Berechnung der geodtischen Breite und Hhe aus rechtwinkligen koorninaten, Allgemeine Vermessungs-Nachrichten 85 (1978) 269-270.
Vincenty, T. (1980): Zur rumlich-ellipsoidischen Koordinaten-Transformation, Zeitschrift fr Vermessungswesen,
105 (1980) 519-521.
Weiss, J. (1993): Resultant methods for the inverse kinematics problem, Computational Kinematics (Eds.) J. Angeles
et al., Kluwer Academic Publishers, Netherlands 1993.
Wellisch, S. (1910): Theorie und Praxis der Ausgleichsrechnung. Bd. II: Probleme der Ausgleichsrechnung, 1910.
Werkmeister, P. (1916): Trigonometrische Punktbestimmung durch einfaches Einschneiden mit Hilfe von Vertikalwinkeln,
Werkmeister, P. (1920): ber die Genauigkeit trigonometrischer Punktbestimmungen, Zeitschrift fr Vermessungswesen 49 (1920) 401-412,433-456.
Werner, D. (1913): Punktbestimmung durch Vertikalwinkelmessung, Zeitschrift fr Vermessungswesen 42 (1913)
241-253.
Winkler, F. (1996): A polynomial algorithm in computer algebra, Springer-Verlag, Wien 1996.
Wu, W. (1984): On the decision problem and mechanization of the theorem proving elementary geometry, Scientia
Sinica 21 (1984) 150-172.
You, R. J.(2000): Transformation of Cartesian to geodetic coordinates without iterations, Journal of Surveying Engineering 126 (2000) 1-7.
Zaiser, J. (1986): Begrndung Beobachtungsgleichungen und Ergebnisse fr dreidimensionales geometrisch-physikalishes
Modell der Geodsie, Zeitschrift fr Vermessungswesen 111 (1986) 190-196.
Zhang, S. (1994): Anwendung der Drehmatrix in Hamilton normierten Quaternionenen bei der Bndelblock Ausgleichung, Zeitschrift fr Vermessungswesen 119 (1994) 203-211.
Zippel, R.(1993): Effective polynomial computation, Kluwer Academic Publications, Boston 1993.
121

Gröbner Bases, Multipolynomial Resultants

Uploaded by

Copyright:

Available Formats

Gröbner Bases, Multipolynomial Resultants

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gröbner Bases, Multipolynomial Resultants

Uploaded by

Copyright:

Available Formats

Grbner bases, multipolynomial resultants

and the Gauss-Jacobi combinatorial algorithms

Von der Fakultt fr

Geodtisches Institut der Universitt Stuttgart

1-1 Statement of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1-2 Solution of the problem

2-1 Linear and nonlinear Gauss-Markov models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2-2 The Gauss-Jacobi combinatorial algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2-3 Solution of nonlinear Gauss-Markov model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Construction of the minimal combinatorial subsets . . . . . . . . . . . . . . . . . . . . . . .

Closed form solution of the minimal combinatorial subsets . . . . . . . . . . . . . . . . . . .

Closed form solution of nonlinear equations by Grbner bases

Closed form solution of nonlinear equations by Multipolynomial Resultants . . . .

Adjustment of the combinatorial minimal subsets solutions . . . . . . . . . . . . . . . . . . .

2-4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Solution of selected Geodetic problems

3-1 LPS and GPS positioning systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Local Positioning Systems (LPS) . . . . . . . . . . . . . . . . . . . . . . . . . . .

Local datum choice in an LPS 3-D network . . . . . . . . . . . . . . . . . . . . .

Relationship between Global and Local Level Reference Frames

3-2 Selected geodetic problems

Threedimensional resection problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

The minimum distance mapping problem . . . . . . . . . . . . . . . . . . . . . . . . . . . .

GPS positioning with observations of type pseudo-ranges

The pseudo-ranging four-point problem in GPS positioning . . . . . . . . . . . . .

Overdetermined GPS positioning with observations of type pseudo-ranges . . . . .

Linearized least squares versus Gauss-Jacobi combinatorial algorithm . . . . . . .

4-2 Closed form solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4-3 Overdetermined solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5-1 Geocentric Cartesian to Gauss ellipsoidal coordinates . . . . . . . . . . . . . . . . . . . . . . . . . .

5-2 7-parameter datum transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6 Summary and Conclusions

Simplexes for Stuttgart Central . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

Error propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

C. F. Gauss combinatorial approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

Tetrahedron for closed form 3d-resection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Computed PDOP for respective combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Scatter of the 15 Gauss-Jacobi combinatorial solutions ( ) around the adjusted value ( ). . . . . . . .

Deviation of positional norm from the mean value . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Graph of the Test network Stuttgart Central . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Scatter of combinatorial solutions for data set 11 Table (4.13) . . . . . . . . . . . . . . . . . . . . . .

Cartesian coordinates of topographic points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Computed polynomial coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Computed ellipsoidal cartesian coordinates and the Lagrange factor . . . . . . . . . . . . . . . . . .

Geocentric coordinates of four GPS satellites and the pseudo-ranging observations . . . . . . . . . .

Geocentric coordinates of six GPS satellites and the pseudo-range observations . . . . . . . . . . . .

Possible combinations and the computed PDOP . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Computed coefficients of the combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Computed combinatorial solution points in a polyhedron . . . . . . . . . . . . . . . . . . . . . . . .

Computed receiver position and stationary receiver range bias . . . . . . . . . . . . . . . . . . . . .

3.11 Computed residuals, squares of residuals and error norm . . . . . . . . . . . . . . . . . . . . . . . .

3.12 Computed positional norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

GPS Coordinates in the Global Reference Frame

Second data set: set 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Third data set: set 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fourth data set: set 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fifth data set: set 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Sixth data set: set 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

seventh data set: set 7

4.10 Eigth data set: set 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Lr{ 1q |Rzyv{ vq Bq;#'yv{ vq Bq%|R

yv{1 |Rzyv{n&| yv{|

* _ r{2n}| +r{C| n 1C *

q * q + q q *

& FBl,^030202FB z, 0

! ! YL03020 -&or.' ]^Y_103020 ] /) 00

Y C03030$ & o.' ] Y 103030#]

^ YNYZ 03020cA ( *t

YL]^Y_030203] XQ;030203 F]^Y_020303] Xf

Definition 2-8 (Grbner basis in terms of

T &6I 4 Y }*i]

Example 2-5 ( BLac ) ):

' I 42( I 43}*z]