Learning Controllers for Complex Behavioral Systems

摘要

Biological control systems routinely guide complex dynamical systems throughcomplicated tasks such as running or diving. Conventional control techniques, however, stumble with these problems, which have complex dynamics, many degrees of freedom, and a task which is often only partially specified. To address problems like these, we are using a biologically inspired, hierarchical control structure, in which controllers composed of radial basis function networks learn the controls required at each level of the hierarchy. Through learning and proper encoding of behaviors and controls, some of these difficulties in controlling complex systems can be overcome.