Volving a Hierarchical C Ontrol S Ystem for C O - Operating a Utonomous Robots

We seek to achieve co-operating adaptive autonomous robots capable of functioning in a semistructured environment. Such an adaptive capacity implies a context sensitive control structure which has either been learnt or acquired through an evolution process. A method is therefore presented which uses a genetic algorithm to optimise the global strategy of two co-operating autonomous robots. The control system has two levels, a primary reactive level of control composed of the Behaviour Synthesis Architecture [1] which is continuously modified by the secondary layer, based on a Fuzzy Logic controller. The Fuzzy rule base has been evolved by a genetic algorithm and aims to adjust the relative utility of each behaviour in the B.S.A, to allow the robot to ‘focus’ its response on the current set of sensory stimuli, reflecting the attention span displayed by most natural animats. This allows the robot to ‘balance’ multiple conflicting behaviours and goals in a dynamic environment.