You are on page 1of 4

NONLINEAR DYNAMIC SYSTEMS CONTROLLER BASED ON BLIND

SEARCH ALGORITHMS

César Daltoé Berci∗ Celso Pascoli Bottura∗


cesardaltoe@hotmail.com cpbottura@dmcsi.fee.unicamp.br


DMCSI - FEEC - UNICAMP
Av. Albert Einstein, 400 - sala 215
Campinas, Brasil

Intelligent Control, Nonlinear Systems, Blind Search Algo- r(t) u(t) y(t)
rithms Controller System 3

1 INTRODUCTION
Nonlinear X Linear Systems x(t)

Classical Methods For Nonlinear Control


Figure 1: Proposed Controller
Motivation, previous results in observability

Genetic Algorithms and its choice motivation state from x0 to x1

Possible other choices considerations: Particle Swarm, Ant 3 THE CONTROLLER


Colony Optimization, Artificial Immune Systems, Simulated
Anilim, and even brute force search. The main equation that defines the search problem.
[12, 10, 11]. [6, 5]:
[1, 3, 2, 4] [13, 8, 7, 9]
u(t) = 
2 THE CONTROL PROBLEM AND NON- (1)

T
arg min (G(u(t), t) − r(t)) (G(u(t), t) − r(t))
LINEAR CONTROLLABILITY CONDI-
TIONS
Future output yα (t0 ):
Particular Control Problem definition based on its empirical
sense.
yα (t0 , u) = y(t0 + α) | {u(t) = u(t0 ) ∀ t0 < t ≤ t + α}
Trajectories and Integral Curves (2)
Nonlinear Controllability Condition Sample time and discretization
Existence of a unique u(t) in the interval [t0,t1] that drive the Controller time-line

Revista Controle & Automação/Vol.XX no.X/Meses XXXX 1


System Input
35 System Output
0
y(t)
r(t)
30 −0.5

−1
Amplitude

Amplitude
25
−1.5

20 −2

−2.5
15

−3

10
0 0.5 1 1.5 2 −3.5
0 0.5 1 1.5 2
Time [s]
Time [s]
Figure 2: System input calculated by the proposed controller,
Figure 3: System output, sample time: α = 0.1s
sample time: α = 0.1s

System Output
0
Stability empirical results and assumptions y(t)
r(t)
−0.2
General discussion about the controller parameters choice
−0.4
when using genetic algorithms
Amplitude

−0.6

4 APPLICATION EXAMPLES −0.8

A nonlinear multi variable dynamical system, non globally −1

controllable −1.2

−1.4

ẋ1 = x21− x2 − u −1.6


ẋ2 = x1 x2 − 2x21 (3)
0 0.5 1
Time [s]
1.5 2

y = x1
Figure 4: System output, sample time: α = 0.005s
Region where the system is locally weak controllable and the
existence of a solution to the optimization problem proposed
System Input
Step response
8

Variations of this response with respect to sample time 6

Servo response and 2nd order predictor for system reference


Amplitude

Inverted Pendulum, problem description 2

ẋ1 = x2 h i
−u−0.25ẏ 2 sin(y)
−2
9.8 sin(y)+cos(y) 1.5
ẋ2 = (4)
0.5[ 34 −13 cos2 (y)] −4

y = x1 0 0.5 1 1.5 2 2.5 3 3.5 4


Time [s]

Step response, general considerations Figure 5: System Input calculated for a step reference signal
0.2
Servo response

2 Revista Controle & Automação/Vol.XX no.X/Meses XXXX


5 CONCLUSIONS
Capability of the proposed controller

Time parameter choice discussion


System Output
Multi variable systems
y(t)
r(t)
0.205
REFERENCES
0.2
[1] César Daltoé Berci. Observadores inteligentes de es-
Amplitude

tado: Propostas. Master’s thesis, Universidade Estadual


0.195 de Campinas, 2008.
[2] César Daltoé Berci and Celso Pascoli Bottura. Ob-
0.19
servador inteligente adaptativo neural não baseado em
modelo para sistemas não lineares. Proceedings of 7th
0.185
Brazilian Conference on Dynamics, Control and Appli-
0 0.5 1 1.5 2 2.5 3 3.5 cations. Presidente Prudente, Brasil, 7:209–215, 2008.
Time [s]
[3] César Daltoé Berci and Celso Pascoli Bottura. Ob-
Figure 6: System Outputy(t) and the reference signal r(t) servador inteligente para sistema não lineares baseado
em inteligência de enxame. Anais do XVII Congresso
Brasileiro de Automática. Juiz de Fora, Brasil, 2008.
[4] César Daltoé Berci and Celso Pascoli Bottura. Ob-
servador neural inteligente não baseado em modelo
para sistemas não lineares. Anais do XVII Congresso
Brasileiro de Automática. Juiz de Fora, MG, Brasil,
2008.
[5] César Daltoé Berci and Celso Pascoli Bottura. Con-
trole genético não-linear por realimentação de estado.
In CBA 2010, sep 2010.
System State
x1 (t) [6] César Daltoé Berci and Celso Pascoli Bottura. Controle
1 x2 (t)
não linear genético do pendulo invertido. In DINCON
2010, jun 2010.
0.5
[7] Lawrence Davis. Handbook of Genetic Algorithms. Van
Amplitude

Nostrand Reinhold, 1991.


0
[8] L.J. Eshelman and J.D. Schafter. Real-coded genetic
algorithm and interval schemata. Foundation of Genetic
−0.5
Algorithms 2, pages 375–382, 1993.
[9] David E. Goldberg. Genetic Algorithms in Search,
−1
Optimization & Machine Learning. Addison-Wesley,
0.5 1 1.5 2 2.5 3 3.5 4
Time [s] 1989.
[10] Petros A. Ioannou. Robust Adaptive Control. Prentice
Figure 7: System State Trajectory when applied the input Hall, 1996.
shown in Figure 5
[11] Shankara Sastry. Nonlinear Systems: Analysis, Stabil-
ity, and Control. Springer Verlag, 1999.
[12] J.T. Spooner, M. Maggiore, R. Ordóñez, and
K. Passino. Stable Adaptive Control and Estimation
for Nonlinear Systems. Wiley Interscience, 2002.

Revista Controle & Automação/Vol.XX no.X/Meses XXXX 3


[13] Alden H. Wright. Genetic algorithms for real parameter
optimization. Department of Computer Science Univer-
sity of Montana, pages 205–220, 1991.

4 Revista Controle & Automação/Vol.XX no.X/Meses XXXX

You might also like