Applications of evolutionary learning in fuzzy logic and optimal control

In this paper we first review two evolutionary learning methods to learn a fuzzy knowledge base that is required to control a system not just from a single initial configuration (open loop control), but over a region of configuration states (closed loop control). It is applied to the control of two different systems, the control of an inverted pendulum (a nonlinear dynamic model), and control of a simulated point mass mobile robot to a fixed target in a two robot collision-avoidance problem, (kinematic model). For these problems appropriate multi-criteria objective/fitness functions incorporating constraint satisfaction need to be developed, together with appropriate operators to ensure effective genetic learning of the final knowledge base. This is seen also to be typically the case in seeking to apply evolutionary algorithms to find the solution of constrained optimisation problems of very high dimension, and for highly constrained continuous optimal control problems. We review some new evolutionary operators for the solution of these problems in the second half of the paper.