International Journal of Emerging Technology and Advanced Engineering

Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)

Observer Based Nonlinear Control of Robotic Manipulator

Using Backstepping Approach
Vishakha Jalaik1, B. B. Sharma2
1
Department of Electrical Engineering, National Institute of Technology, Hamirpur, Himachal Pradesh, India – 177005
2
Faculty, Department of Electrical Engineering, National Institute of Technology, Hamirpur, Himachal Pradesh, India - 177005
Abstract- In this paper, control of robot manipulator with Observer design for systems like robot manipulators are
unknown system states is addressed using backstepping necessitated because these systems are in constant exposure
approach. The systematic backstepping procedure is used in to the surrounding environment, thus leading to parameter
conjunction with Luenberger observers to achieve estimates of variations in such systems. Due to this, observer design
the unknown states. First, assuming that all the states of
becomes important for the control of these manipulators.
system are known, controller using the proposed approach
has been designed. Next, construction of Luenberger observer Many nonlinear control design strategies have been
has been proposed while establishing the stability at every presented for such systems such as output feedback control,
stage using Lyapunov approach. Simulation results for both fuzzy control etc. [11]-[12]. Here, system is represented in
the cases have been presented in the end to show the efficacy terms of polynomial fuzzy model in closed loop with the
of the proposed approach. polynomial fuzzy controller. Lyapunov stability criterion is
used for determining stability of the overall system. It is an
Keywords – Backstepping, Luenberger observer, Nonlinear effective approach for control of nonlinear system but
control, Observer-based control, Robot manipulator. computational complexity is increased. Adaptive observer-
based fuzzy tracking control was proposed in [13]–[14]. In
I. INTRODUCTION comparison with the fuzzy controller without adaptive
In practice, majority of the systems are nonlinear in capability, the adaptive fuzzy controller is more robust. In
nature and controlling the behavior of these systems is a [15], control mechanism on the basis of sliding surface
challenging task. Moreover, because of complexity of theory is provided. A continuous feedback control
systems and underlying cost of sensors, all the states of the mechanism is used which results in the state trajectory
system are not always known or available for online sliding along a time-varying sliding surface in the state
measurement. In such cases, in order to provide a precise space. But tracking is only achieved within prescribed
control input that will provide us with desired output, these limits due to nonlinearities and uncertainties of system
states have to be estimated. For this purpose, observers are states and parameters. Strategy based on backstepping
used which basically provide us with the estimate of these methodology serves as a robust technique for such systems
unknown states based on the known outputs and inputs. having nonlinearities and parametric uncertainties [16].
Over the last few decades, lot of work has been reported on Backstepping approach is also used in conjunction with
observers. Various linear and nonlinear control tools like other nonlinear approaches like sliding mode, adaptive
backstepping, variable structure control, OGY method, control etc. [6, 8, 18].
adaptive control, observer-based control etc. are utilized to In the present paper, the problem of tracking control of
design controllers for stabilization, tracking and single link robotic manipulator with actuator dynamics is
synchronization of various nonlinear systems [1–10]. addressed using systematic backstepping procedure. The
The applications of all these control design techniques uncertainties associated with the measurement of states of
are also extended to robotic systems. In robotics, the system are handled by adapting Luenberger observer
a manipulator is a device used to manipulate materials structure. The controller utilizes the estimates of uncertain
without direct contact. It consists of a controller and a states to achieve the asymptotic tracking performance. The
manipulator arm. Their applications include Motion stability of the proposed controller is established by using
planning, remote handling, Micro-robots, Machine tools suitable embedded Lyapunov function at every stage of
etc. Performance of the manipulator depends upon backstepping procedure. In the end, the efficacy of the
controller to large extent. The robot performance is mainly proposed approach is shown by presenting extensive
influenced by the mechanical design and by the actuation simulations.
system. The control problem is to define the input signals
for the joints in order to achieve a predefined behavior for
the manipulator.
232
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
The paper is organized as follows. Section II represents y= (8)
the differential model of a single link robot manipulator
with actuator dynamics. Further, design procedure for Which is required to track desired trajectory while
controller based on backstepping approach assuming all the suitably designing control function u.
states to be known has been discussed and the simulation A. Design of Controller Using Backstepping approach:
results have been presented for this case. Section III
The main objective is to design a controller which
presents Luenberger observer based controller design
provides appropriate input signal to the robot joints so that
procedure using backstepping approach for system with
they perform desired operation of tracking given trajectory.
uncertainty. Simulation results for the controlled system
Let the system output is required to track reference
with proposed observer structure are presented. Finally,
signal . Then tracking error is given by:
conclusion is represented in section IV.
=y- (9)
II. BACKSTEPPING BASED CONTROLLER DESIGN FOR
SINGLE LINK MANIPULATOR WITH ACTUATOR DYNAMICS ̇ = ̇ ̇ - ̇ (10)

The dynamic description for a single link robot Where, x2 is considered as a fictitious control signal. Let
manipulator with actuator dynamics can be written as desired signal be then, new error variable is defined
follows [17]: as:

D ̈ +B ̇+N = (1) = - (11)

M ̇ H =u- ̇ (2) This definition modifies (10) as:

Where, „u‟ is the control input to be designed in such a ̇ + - ̇ (12)

way that given choice of variable meets out given Let, be selected as:
objective. Let the selection of state variables , &
= ̇ - (13)
be made as follows:
It leads to modified error dynamics of first stage as:
= ; = ̇ =
̇ = (14)
& following nonlinear functions be defined:
The Lyapunov function candidate at this stage is chosen
=-( as:

= (15)
= -( ) (3a)
The time derivative of while using equation (14)
Let functions be known invertible becomes:
linear/nonlinear functions in system description in (1) & (2)
and be denoted as ̇ = - ; >0 (16)

= ; = . (3b) So, the system is stabilized by assumed pseudo

controller . From definition of , the dynamics of second
The above system in state space form can be written as stage is written as:
̇ = ̇ - ̇
̇ = (4)
= + - ̇ (17)
̇ = + (5)
Let fictitious controller for this stage be and desired
̇ = + u (6) value be . Then, by defining a new error variable as:
In compact form above system can be written as: = - (18)
̇ = f(x) + g(x)u (7) The equation (17) can be modified as:
The system output is considered to be: ̇ + - ̇ (19)

233
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
Choosing desired value of to be This is a stable dynamics as per equation (29) while
selecting gains , > 0. So asymptotic tracking
= ( - - + ̇ ) (20) performance is achieved i.e. system output tracks
The dynamics in (19) becomes reference signal .

̇ - - + (21) B. Numerical Simulation for Backstepping approach

For simulation purpose, consider system parameters for
Let, Lyapunov function candidate for second stage be the single link manipulator system with actuator dynamics
selected as: in equation (4), (5) and (6) be as follows: D=1, M=0.05,
= + (22) B=1, Km=10, H=0.5, N=10. The control gains are selected
as K1=10, K2= 25& K3=25. The initial conditions for
̇ = - + (- - + ) (23) system states are assumed as
̇ = - - ; , > 0 (24) xo =

For the last stage error dynamics becomes: For present case F1, F2, G1 & G2 are taken as per
equation (3).
̇ = ̇ - ̇ = + ubs - ̇ (25) Fig. 1(a) shows the time evolution of system output and
reference signal. The figure clearly shows tracking of
Choosing Lyapunov function candidate for this stage
desired reference by the actual output of the system.
as:
backstepping approach
3

(26) refrence
x1
2.5
The time derivative while using dynamics of equation
(25) becomes 2

̇ =- ̇
1

- + (27)
yr,x

1.5

System will be stable if and only if 1

̇ =- (28) 0.5

̇ =- - - (29) 0
0 10 20 30 40 50 60 70 80 90 100
t

Where, , >0
(a)
In (29), ̇ is negative definite and system is
backstepping approach
asymptotically stable. 3
refrence
Controlled input obtained while establishing (29) is as x1
2.5
follows:
( ̇ -
2
= - ) (30)
1
yr,x

1.5
The Final stage dynamics becomes
̇ =- u- ̇
1
= + (31)
0.5
So combined error system can be written as:
̇ 0
0 1 2 3 4 5 6 7 8 9 10
t
̇ [ ̇ ] [ ][ ] (32) (b)
̇
Figure 1: (a) Time evolution of system output y = and reference
signal & (b) expanded version of (a) in time span [0 10].

234
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
backstepping approach III. OBSERVER BASED CONTROLLER DESIGN FOR SYSTEM
7
x2
OPERATION IN UNCERTAIN ENVIRONMENT
6
As discussed earlier, all the states of single link
5 manipulator system with actuator dynamics in real practice
4
are not always known. So, in order to provide full state
feedback an observer is needed to be designed for unknown
2

3 system states. Here, Luenberger observer structure has been

x

2 presented to develop observer for uncertain environmental

operations.
1
A. Proposed observer structure
0
The Luenberger observer for system mentioned in
-1 section 2 can be written as:
0 10 20 30 40 50 60 70 80 90 100
t
̂̇ ̂ ̂ ̂ ̂ ; (33)
(a)
The above system using suitably selected feedback
30
backstepping approach
controller ̂ leads to following dynamics
x3
25 ̂̇ Â+ ̂ ̂ (34)
20
where, observer output is taken as:
15

10
̂ ̂ (35)
Then tracking error „E‟ is:
3

5
x

0
= ̂ (36)
-5
and the corresponding error dynamics can be written as
-10

-15
̇ ̂ (37)
-20
0 10 20 30 40 50 60 70 80 90 100 In order to design an observer, the following
t assumptions must be made:
(b) Assumption-1: The unknown parameters and states of the
Figure 2 : Time evolution of system states (a) x2 (b) x3 system are bounded and (A, C) pair is observable, with A
as linear part of system dynamics and C is output matrix.
0.2 Assumption-2: The gain matrix 𝐿 satisfies the following
Tracking error(x1-yr)
0
condition:
-0.2 (A-LC)TP+P (A-LC) = -Q ; (38)
-0.4
where, A, B, C and L are system matrix, input matrix,
Tracking error

-0.6 output matrix and gain matrix, respectively and P and Q are
-0.8
positive matrices for observer system defined in equation
(33).
-1
Assumption-3: Nonlinearities of system follow Lipschitz
-1.2
conditions i.e.
-1.4

̂ x- ̂
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time

Figure 3: Time evolution of tracking error Where, is Lipschitz constant.

235
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
Consider Lyapunov function candidate for the error So the system is stabilized by the pseudo controller .
system in equation (37) as: From the definition of , error dynamics for second stage
V = ETPE (39) becomes:

The time derivative of the Lyapunov function candidate ̇ = ̂̇ - ̇ = ̂ +̂ ̂ 𝐿 ̇ (48)

is as follows:
Let, ̂ to be the fictitious control signal and be the
̇ ̇ PE+ P ̇ desired value. Then, a new auxiliary error variable is
defined as:
̂ (40)
=̂ (49)
Using conditions of assumption (2) & (3), one can get
Now, equation (46) is modified as:
̇ x- ̂
̇ ̂ +̂ ̂ 𝐿 ̇ (50)
‖ ‖( )‖ ‖ ‖ ‖ ‖ ‖ (41) Choosing desired value of next state to be
Suitable selection of P as positive definite matrix and = (- ̂ – + ̇ 𝐿 ) (51)
taking to be ( ) in assumption (2) and solving
for suitable gains , equation (38) can be shown to be It leads to:
negative definite leading to the asymptotic stability of ̇ – ̂ (52)
system. Hence, as time tends to infinity, error tends to zero
leading to the desired behavior of the system. Choosing Lyapunov function candidate for second stage
as :
B. Observer based backstepping approach
Let the observer system defined in equation (33) is = + (53)
required to track reference signal then tracking error is
given by ̇ = - + (̂ +̂ ̂ - ̇ + 𝐿 )

= ̂- (42) ̇ = - - ; , > 0 (54)

̇ = ̂ - ̇ (43) For the last stage error dynamics is given by:
Where, x2 is a fictitious control signal. Let desired signal ̇ ̂̇ ̇ (55)
be then a new error variable is given by
Choosing Lyapunov function candidate for last stage
=̂ 𝐿 - (44) as :
Let be chosen as: (56)
= ̇ - 𝐿 (42) ̇ =- - + ̇

This modifies equation (40) as: =- - + ̂ +̂ ̂ 𝐿 ̇ (57)

̇= 𝐿 ̇ (45) System will be stable if and only if
The Lyapunov function candidate for stage one is chosen ̇ = - (58)
as: Then,
= (46) ̇ =- - - (59)
Making use of equation (40) and (43), we get: Where, , >0
̇ = ̇ is negative and system is asymptotically stable.
- ; >0 (47)
Controlled input can be determined as follows:

236
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
̂ = ( ̇ - -̂ 𝐿 ) (60) 3
backstepping approach

refrence
Tracking error system can be written as: x1
2.5 x1^

̇= q (61)
2
Where,

yr,x_1,x1^
1.5

[ ̂ ] (62)
1

For error system to converge, should be Hurwitz 0.5

matrix. This can be achieved by pole placement technique.

0
Using suitable positive values of gains K1, K2, K3 , above 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t
error dynamics can be stabilized, hence leading to the
(b)
desired tracking performance for the system under
Figure 4: (a) Time evolution of system output y = , observer o/p
consideration.
and reference signal & (b) expanded version of (a) in time span [0
C. Simulation results for observer based backstepping 5].
approach
backstepping approach
7
Consider parameters for the manipulator system defined x2
by equation (7) and (33) as 6 x2^

D=1, M=0.05, B=1, Km=10, H=0.5, N=10 5

Let the gains be selected as K1=15, K2= 25, K3=25 and 4

x_2,x2^

L1=K1 ; L2=K2 ; L3 = K3 for simulation purpose. The initial 3

conditions are assumed as xo = [0.1, 6.28, 0, 0, 0, 0].

2
For present backstepping based observer case in uncertain
environment, F1, F2, G1, G2 are taken as per equation (3). 1

0
backstepping approach
3
refrence -1
0 10 20 30 40 50 60 70 80 90 100
x1 t
2.5 x1^
(a)
2 backstepping approach
7
x2
yr,x_1,x1^

6 x2^
1.5

1
4
x_2,x2^

3
0.5
2

1
0 10 20 30 40 50 60 70 80 90 100
t
0
(a)
-1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t

(b)
Figure 5: (a) Time evolution of x2 and ̂ & (b) expanded version of
(a) in time span [0 5].

237
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
backstepping approach
40 20
x3 e1
x3^ 10 e2
30 e3
0

20 -10

Tracking errors
x_3,x3^

-20
10

-30

0
-40

-50
-10

-60
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
-20 time
0 10 20 30 40 50 60 70 80 90 100
t
(b)
(a) Figure 7: Time evolution of tracking errors& (b) expanded version of
(a) in time span [0 5].
backstepping approach
40
x3 As it can be seen from the simulation result, observer
30
x3^
based system converges faster than the former system. By
comparing figure 3 and 7, we can see that tracking error
20 becomes zero faster in later approach. The presented results
show the efficacy of the proposed approach.
x_3,x3^

10

0
IV. CONCLUSION
Observers are the critical part of real world control
-10
system where all the states of the system are not known or
-20
not available for on line measurement. In this paper, design
0 0.5 1 1.5 2 2.5
t
3 3.5 4 4.5 5
of controller for robot manipulator system using
(b) backstepping approach combined with Luenberger observer
Figure 6: (a) Time evolution of system state x3 and ̂ & (b) expanded
is presented for the robotic manipulator system with
version of (a) in time span [0 5]. actuator dynamics. Simulation results for above mentioned
approaches have been presented and compared. Simulation
20 result shows the effectiveness of the proposed techniques.
e1
10 e2
e3
REFERENCES
0 [1] Bastin G. and Gevers M., 1988, “Stable Adaptive Observers for
Nonlinear Time Varying Systems”, IEEE. Trans. on Automatic
-10
Tracking errors

Control, vol. 33, no. 7, pp. 650-658.

-20 [2] Gauthier J. P., Hammouri H. and S. Othman, 1992, “A simple
Observer for Nonlinear Systems Applications to Bioreactors”, IEEE.
-30
Trans. Automatic Control, vol. 37, no. 6, pp. 874-879.
-40 [3] Khalil, H. 1996, Nonlinear Systems, New Jersey, Prentice Hall.
-50
[4] Boutayeb M., Darouach M., and Rafaralahy H., 2002, “Generalized
state space observers for chaotic synchronization and secure
-60
0 10 20 30 40 50 60 70 80 90 100
communication,” IEEE Trans. Circuits Syst. I, Vol. 49, No. 3.
time
[5] Sharma B. B. and Kar I. N., 2011, “Observer-based synchronization
(a) scheme for a class of chaotic systems using contraction theory”,
Nonlinear Dynamics, vol. 63, no. 3, pp. 429–445.

238
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 6, Issue 6, June 2016)
[6] Morgul O. and Solak E., 1996, “Observer based synchronization of [12] Huang H., Yan J., and Cheng T. , 2007, “Development and fuzzy
chaotic systems,” Phys. Rev. E., vol. 54, pp. 4802–4811. control of a pipe inspection robot,” IEEE Trans. Ind. Electron., vol.
[7] Grassi, G. and Mascolo S., 1997, “Nonlinear observer design to 57, no. 3, pp. 1088– 1095, [13] Liu Y. J. and Wang W., “Adaptive
synchronize hyperchaotic systems via a scalar signal”, IEEE Trans. fuzzy control for a class of uncertain nonaffine nonlinear systems,”
Circuits Syst.-I, vol. 44, no. 10, pp. 1011–1014. Inf. Sci., vol. 177, no. 18, pp. 3901–3917.
[8] Meena N., Sharma B. B., “Backstepping Algorithm with Sliding [13] Tong S. and Li Y., 2009, “Observer-based fuzzy adaptive control for
Mode Control for Magnetic Levitation System”, International strict feedback nonlinear systems,” Fuzzy Sets Syst., vol. 160, no.
Journal of Emerging Trends in Electrical and Electronics, vol. 10, 12, pp. 1749– 1764.
pp. 39-43, 2014. [14] J. J. Slotine, S. S. Sastry, 1983, “Tracking control of non-linear
[9] Agrawal V. and Sharma B. B., 2014, “An observer based approach systems using sliding surfaces, with application to robot
for multi-scroll chaotic system synchronization and secure manipulators”, International Journal of Control, Vol. 38, no. 2.
communication with multi-shift ciphering”, Int. Conf. on Advances [15] Krstic M., Kanellakopoulos I. and Kokotovic P., 1995, “Nonlinear
in Engineering and Technology Research (ICAETR), pp.1-6. and Adaptive Control Design” John Wiley and sons, New York.
[10] Sharma V, Agrawal V, Sharma B B and Nath R, 2016, “Unknown [16] Murray R. M., Li Z .and Sastry S. S., 1993, “A mathematical
input nonlinear observer design for continuous and discrete time introduction to robot manipulation”, CRC Press.
systems with input recovery scheme”, Nonlinear Dynamics, pp. 1- [17] Dochain, D. and Perrier, M., 2003, “Adaptive Backstepping
14, 2016, DOI: 10.1007/s11071-016-2713-5 Nonlinear Control of Bioprocesses”, International Federation of
[11] Lam H. K. and Li H., 2013, “Output-feedback tracking control for Automatic Control Proceedings, ADCHEM, 77-82
polynomial fuzzy-model-based control systems”, IEEE transactions
on industrial electronics, vol. 60, no. 12.

239