Information Technology Reference
In-Depth Information
Optimization of Emotional Learning Approach
to Control Systems with Unstable Equilibrium
Mohammad Hadi Valipour, Khashayar Niki Maleki, and Saeed Shiry Ghidary
Abstract. The main problem concerning model free learning controllers in partic-
ular BELBIC (Brain Emotional Learning Based Intelligent controller), is attributed
to initial steps of learning process since the system performance is dramatically
low, because they produce inappropriate control commands. In this paper a new ap-
proach is proposed in order to control unstable systems or systems with unstable
equilibrium. This method is combination of one imitation phase to imitate a ba-
sic solution through a basic controller and two optimization phases based on PSO
(Particle Swarm Optimization) which are employed to find a new solution for stress
generation and to improve control signal gradually in reducing error. An inverted
pendulum system is opted as the test bed for evaluation. Evaluation measures in
simulation results show the improvement of error reduction and more robustness
than a basic tuned double-PID controller for this task.
1
Introduction
Traditional control methods are based on system identification, modeling, and de-
signing controller regarding to predefined goals for under control systems. As these
systems become more complex, on one hand their identification grow to be much
more difficult and sometimes impossible, and on the other hand there are some fac-
tors such as dynamics of system, uncertainty, and decay induced system changes,
that require redesigning or readjusting the controller [15]. Readjusting parameters
 
Search WWH ::




Custom Search