Co-evolutionary learning and hierarchical fuzzy control for the inverted pendulum

In this paper we examine the problem of learning a two layer hierarchical fuzzy controller for the control of the inverted pendulum (with nonlinear dynamics). The fuzzy rules are learned using cooperative co-evolution, whereby two distinct evolutionary populations are used: one defining the first fuzzy layer and the other defining the second fuzzy layer. We compare the results from the co-evolutionary algorithm with the results from a classical evolutionary algorithm.