Beruflich Dokumente
Kultur Dokumente
net/publication/320041584
CITATIONS READS
0 982
2 authors:
Some of the authors of this publication are also working on these related projects:
Development of rapid manufacturing process using 3D printing and ultrasonic assisted sintering View project
All content following this page was uploaded by Gurminder Singh on 31 October 2018.
2729
IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI-2017)
2730
IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI-2017)
ℒ ℒ
− + = , (7) In this section, design of LQR controller is discussed for the
̇ ̇
cart pendulum system. As the objective of the controller, is to
stabilize the plant at the unstable equilibrium point, the problem
ℒ ℒ can be treated as a regulation problem as
̇ − + ̇ = 0. (8)
( )= ( )− ( ) ,
By substituting the values of Lagrange and Rayleigh’s
dissipation function from (5) and (6) into in (7) and (8), the where ( ), ( ), ( ) and are control input vector, state
equations of motion are calculated as
vector, desired state vector and control gain vector. For the
̈+ ̇ + regulation problem, ( ) = 0, which leads to
( + ) ̈− cos sin ̇ = , (9)
( )= − ( ). (13)
( + ) ̈− cos ̈− sin = 0. (10)
Linear Quadratic Regulator is the optimal control used to
Equation (9) and (10) represent the nonlinear equations of
minimize the cost associated with generating control inputs.
motion of the cart pendulum system. For stabilization of the cart
The cost function consists of two matrices, Q and R, the state
pendulum system around upright equilibrium point, the
weighting matrix and control cost matrix. For a linearized plant,
nonlinear system can be linearized by neglecting the square
the objective of LQR controller is to minimize the integration
term and taking sin = , cos = 1. The design of controller
of sum of these two matrices by using a feedback regulator. The
requires the mathematical model of the system in state-space
objective function is given as:
form. Thus, state-space representation of the linear system is
given as
̇( ) = ( ) + ( ), (11) ( )= ∫ ( ( ) ( )+ ( ) ( ) , (14)
( )= ( )+ ( ).
and K is the optimal state feedback control matrix given as
where ( ) = [ ̇ ̇ ] , is called as the state-vector, ( ) = = , (15)
[ ] as the input vector, ( ) is known as the input vector where, is the solution of the following algebraic Riccati
and A, B, C and D are called as state-weighing coefficient equation
matrices, which are given as
2731
IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI-2017)
0 1 0 0 0
̇ 0 −0.09 0.63 0 ̇ 0.89
= + ,
0 0 0 1 0
̇ 0 −0.23 27.82 0 ̇ 2.35
(17)
1 0 0 0 ̇ 0
= + .
0 0 1 0 0
̇
The eigenvalues of the state dynamic matrix A are Fig. 5: Step response using LQR controller.
[0, −0.08, −5.27, 5.27]. One of the four eigenvalues, one lies
at the origin position, which signifies that the system is going In the first test run, the control objective is to stabilize the
to unstable. But system is unstable, as one of the eigenvalue of pendulum at upright position i.e. = 0, while the cart
the open-loop system lies in the right half plane. Thus, in order maintains the reference position on the horizontal track i.e. =
to stabilize the system at the inverted position, a controller 0. First, the experimental run needs to be executed, as it will
needs to be designed in order to shift the eigenvalues in the left
decide the initial condition of , where the linear controller
half plane.
starts operating. At that instant, the GLIP setup is in its home
The cart pendulum system has four state variables (n), i.e.
̇ ] , where is the linear displacement, ̇ is the position (stable equilibrium point) of the pendulum, i.e. = 0
=[ ̇
and = 0. Now, the pendulum is lifted manually in the
linear velocity of the cart, is the angular displacement and ̇
counter-clockwise direction, until it comes in the linear stage of
is the angular velocity of the pendulum. Thus, n = 4 for
operation i.e. = ±20°. As the pendulum is lifted, at =
controllability matrix, which is given as
−19.5°, the linear LQR controller becomes active and stabilize
the pendulum at its unstable equilibrium point with an
=[ ],
overshoot of 2°, shown in Fig. 6. It can be observed from Fig.
7 that in order to stabilize the pendulum, the cart executes an
0 0.88 −0.07 1.48
0.88 −0.07 1.48 −0.26 overshoot of 0.15 m from its reference position and the closed-
= . (18) loop system settles within 2 seconds, as expected due to the
0 2.35 −0.20 65.59
2.35 −0.21 65.59 −6.14 choice for state weighting and control cost matrices.
After getting the initial condition of from the experimental
The rank of the controllability matrix is calculated as four. run, we switch to the simulation run for validation the response
Hence, the cart pendulum system is controllable. For both of the closed-loop system. Similar trend is observed in the
analytical and experimental purposes, the initial conditions are analytical results, performed on the Matlab environment, as
taken as both the cart and pendulum position settle with 2 seconds. It can
be seen in Fig. 6 and 7. That the overshoot for the simulation
2732
IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI-2017)
run in the case of ( ) and ( ) is found as 12°and 0.18 m, as LQR controller is able to stabilize the cart pendulum system
compared to 2° and 0.15 m respectively in the experimental at its unstable equilibrium point (upward position). The
results. This discrepancy in the overshoot results is due to the pendulum angle’s trend at upward position is shown in Fig. 9.
activation of the linear controller in real time. From the figure, it is observed that the pendulum oscillates in
the range of -0.6° to 0.6°.
VI. CONCLUSIONS
In this paper, the real-time control of the cart pendulum
system is investigated, which is underactuated in nature. With
only one actuator used in the system i.e. to control cart
movement, it is difficult to stabilize the pendulum at unstable
equilibrium point. LQR controller has been used to address this
stabilizing problem. LQR controller has been designed to
stabilize the cart pendulum system in 2 seconds at its desired
Fig. 6: Angular displacement of the pendulum by LQR position. The dynamics of the system has been derived by using
controller. Euler-Lagrange approach to ensure the accuracy of the
analytical results, which is validated with the close agreement
of the simulation results with those obtained with real-time
setup. Thus, the derived mathematical model has been
validated, which was not validated for this setup in [3]. Control
input magnitude has been compared, both experimentally and
analytically, and results are found in close agreement.
REFERENCES
[1] Y. Liu, H. Yu, “A survey of underactuated mechanical systems”,
IET Control Theory & Applications, Vol. 7, pp. 921-935, 2013.
[2] M. W. Dunnigan, “Computer based control assignment- digital
control of an inverted pendulum”, International Journal of Electrical
Fig. 7: Linear displacement of the cart by LQR controller. Engineering Education, Vol. 35, pp. 162-174, 1998.
[3] Googol Technology (HK) Limited, “Inverted pendulum
experimental manual suitable for GLIP series”, second edition,
The comparison of control input is shown in Fig. 8, where 2016.
the magnitude of maximum control input is found as 20 units [4] W. D. Chang, R. C. Hwang, J. G. Hsieh, “A self-tuning PID control
for the simulation run, whereas it is observed as 25 units for the for a class of nonlinear systems based on the Lyapunov approach”
Journal of Process Control, Vol. 12, pp. 233-242, 2002.
experimental run. Thus, it is observed that the maximum control [5] K. Ogata, “Modern Control Engineering”, Prentice-Hall, New York,
input is of the same order in both the cases, which validates the 2000.
efficacy of the mathematical model. [6] V. Kumar, J. Jerome, “Robust LQR controller design for stabilizing
and trajectory tracking of inverted pendulum”, Procedia
Engineering, Vol. 64, pp. 169-178, 2013.
[7] G. W. Linden, P. F. Lambrechts, “H control of an experimental
inverted pendulum with dry friction”, IEEE Control Systems, Vol.
93, pp. 44-50, 1993.
[8] J. Yi, N. Yubazaki, K. Hirota, “Upswing and stabilization control of
inverted pendulum system based on the SIRMs dynamically
connected fuzzy inference model” Fuzzy Sets and Systems, Vol.
122, pp. 139-152, 2001.
[9] C. W. Anderson, “Learning to control an inverted pendulum using
neural network”, IEEE Control Systems Magazine, Vol. 9, pp. 31-
37, 1989.
[10] S. Mori, H. Nishihara, K. Furuta, “Control of unstable mechanical
Fig. 8: Input to cart pendulum system by LQR controller. system- control of pendulum” International Journal of Control, Vol.
23, pp. 673-692, 1976.
[11] O. Boubaker, “The inverted pendulum benchmark in nonlinear
control theory”, International Journal of Advanced Robotic
Systems, Vol. 10, pp. 1-9, 2013.
[12] A. Tewari, “Modern control design with MATLAB & SIMULINK”,
Chichester Wiley, 2002.
2733