Plasticity in sensory{motor coordination

This paper explores a novel challenge for dynamic recurrent networks; an environment where plasticity in sensory{motor coordination is required. The paper provides a brief review of some previous work looking at plasticity in sensory{ motor coordination for networks with explicit learning algorithms, and describes existing work in the eld of adaptive behaviour where dynamic neural networks are utilised. Experimental work is described in which a simulated robot is evolved tòtrack' and approach a light source despite having previously unencoun-tered mappings applied to its motor outputs, relying only on the inherent dynamics of the network controller, without the use of architectural adaptations such as weight changes. Preliminary analysis of this network controller is presented.