Professional Documents
Culture Documents
232 900 1 PB
232 900 1 PB
141
Raghunath et.al/GPS Datum Conversion and Kalman Filtering For Reducing Positional Errors
locate position with respect to a map. Map coordinates use independent (of each other), white, and with normal
a 2-D Cartesian system, in which the two axes are known as probability distributions
eastings and northings. The conversion from WGS-84 to P (W) – N (0, Q) --------------------- (3)
UTM is done using the direct conversion algorithm by P (V) – N (0, R) -------------------- (4)
Steven Dutch [12]. The algorithm gives UTM datum which The process noise covariance Q and measurement
is accurate to within less than a meter within a given grid noise covariance R matrices as in equations 3 & 4 might
zone. change with each time step or measurement, however here
we assume they are constant.
FRANSON COORDINATES
The nn matrix A in the difference equation (1)
A coordinate from GPS generally uses the WGS84
relates the state at the previous time step K-1 to the state at
datum. But maps and many GIS applications use other
the current step K, in the absence of either a driving
coordinate systems like NAD83 and UTM. Franson
function or process noise. Note that in practice A might
Coordinate transforms or CoordTrans is a software that can
change with each time step, but here we assume it is
be used to convert between all the coordinate systems on
Earth with high accuracy. In this software the text file with constant. The nl matrix B relates the optional control
input u to the state x. The mn matrix H in the
l
the datum can be directly used for the conversion. A special
feature of this software is to indicate specific location in a measurement equation (2) relates the state to the
pictorial format [13]. measurement ZK. In practice H might change with each time
Franson CoordTrans can convert single step or measurement, but here we assume it is constant.
coordinates, large files of coordinates and even live The Kalman filter estimates a process by using a form of
coordinates from a GPS. The WGS-84 datum which has to feedback control: the filter estimates the process state at
be converted can be in degree minutes or degree minutes some time and then obtains feedback in the form of (noisy)
and seconds form. The datum which is obtained after measurements. As such, the equations for the Kalman filter
conversion, will be more accurate then the datum which is fall into two groups: time update equations and
obtained by using the direct conversion algorithm. The measurement update equations as shown in figure 1.
datum which is obtained from this software is considered Discrete Kalman filter time update equations (5 & 6) are
as a reference datum for the comparison and calculation of given as
the errors. xˆ -k Axˆk 1 Buk
---------- 5 & 6
KALMAN FILTER Pk APk 1 AT Q
A Kalman filter is a mathematical toolbox which
computes the state estimate of a process with high Time update equations project the state and
efficiency in the presence of noise by minimizing the linear covariance estimates forward from time step k-1 to step k.
mean square error. It employs a Predictor-corrector type A and B are from equation (1), while Q is from is from
estimation to achieve a optimum result. The Kalman filter equation (.3). Initial conditions for the filter are discussed
follows a recursive algorihm which refines the estimate in a in the earlier references. Discrete Kalman filter
series of passes. In this paper a kalman filter is used to measurement update equations (7, 8 & 9) are given below.
reduce the ionospheric errors and the errors introduced K k Pk H T ( HPk H T R) 1
during coordinate conversion, thus improving the overall
range measurement accuracy. xˆk xk K k ( Z k Hxˆk ) -------7, 8 & 9
The best estimate of the position with a Kalman filter can Pk ( I K k H ) Pk
be obtained if the system is linear with Gaussian errors.
The most distinctive feature of Kalman filter is the use of a The time update equations are responsible for
recursive data processing algorithm to compute the projecting forward (in time) the current state and error
solution. Each update estimate is computed from the covariance estimates to obtain the a priori estimates for the
previous estimate and the input data. This only requires next time step. The measurement update equations are
the storage of the previous estimate. responsible for the feedback--i.e. for incorporating a new
The filter processes measurements to deduce a measurement into the a priori estimate to obtain an
minimum error estimate of the system by utilizing the improved a posteriori estimate. The time update equations
knowledge of the system, measurement dynamics and can also be considered of as predictor equations, while the
statistics of the system, noise measurement errors and measurement update equations can be corrector equations.
initial condition information. The Kalman filtered datum The first task during the measurement update is to
has less variance and mean error when compared to the compute the Kalman gain, Kk. The next step is to actually
datum obtained from direct conversion algorithm. measure the process to obtain Zk and then to generate an a
posteriori state estimate by incorporating the measurement
Mathematical formulation of Kalman Filter as in equation (8). The final step is to obtain an a posteriori
The Kalman filter estimates the state x of
n
error covariance estimate via equation (9). After each time
a discrete-time controlled process that is governed by the and measurement update pair, the process is repeated with
linear stochastic difference equation as in equation 1. the previous a posteriori estimates used to project or
XK = A X K -1 + B U K + W K -1 ------------- (1) predict the new a priori estimates.
With a measurement x
m
that is (as stated in RESULTS AND DISCUSSIONS
equation 2). The area of Ameerpet, Hyderabad,in the state of
ZK = HX K +V K ----------------------------- (2) Andhra Pradesh, India is the location at which the datum
The random variables WK and VK represent the process and has been collected. Ameerpet is a densely populated area
measurement noise (respectively). They are assumed to be with heavy traffic and tall buildings. The datum is collected
throughout the day at three different times i.e in the
142
Raghunath et.al/GPS Datum Conversion and Kalman Filtering For Reducing Positional Errors
Figure 1: Eastings obtained from software (Franson Figure 4: Eastings error between the software and the
CoordTrans) used for the conversion. conversion algorithm
The datum output which is obtained from the The error between the software and the kalman
reciver is also applied to the direct conversion algorithm filter datum is shown in figure 5. The diffrence between the
(program). The Eastings which is obtained from the software and the klaman filter is in the range of 19320 to
conversion is shown in figure 2. The range in which the 19350 meters. The error by using the klaman filter is is less
eastings are obtained are from 209350 to 209420 meters. when compared to the error between the software and the
conversion algorithm.
Figure 2: Eastings obtained from program (direct Figure 5: Eastings error between the software and the
conversion) used for the conversion. kalman filter algorithm
The UTM datum obatined from the conversion Figure 6 shows the comparision results of the
algorithm are smothened using the kalman filter algorithm errors obtained in figures 5 and 6. From this grpah it can be
is shown in figure 3. The easting datum which is obtained shown that the Kalman filter gives an output which is
from the direct conversion algorithm is applied to the smooth when compared to the output which is obtained
kalman filter. The output which is obtained will be ranging from the direct conversion algorithm.
from 209360 to 209440 meters.
143
Raghunath et.al/GPS Datum Conversion and Kalman Filtering For Reducing Positional Errors
144
Raghunath et.al/GPS Datum Conversion and Kalman Filtering For Reducing Positional Errors
145