A fuzzy controller for Flakey, the robot

We present a fuzzy controller for Flakey, an autonomous mobile robot. This controller is based on the notion of ''behavior'': a specific conrtol regime aiming at the achievement of one specific goal. Behaviors are imlpemented by a set of fuzzy rules. Yhe controller includes behavior for reliable obstacle avoidance, as well as for purposeful motion, like following a wall, or reaching a given location. Many behavior can be simultaneously active: the fuzzy controller blends these behaviors using the mechanisms of fuzzy logic. By blending reactive and purposeful behaviors, the controller can lead Flakey to achieve a given goal while smoothly reacting to unexpected events. By blending many consecutive purposeful behaviors, the controller can execute a full plan. Context-dependent blending of different behaviors is the key to relaible operation in an unstructured, real world environment. The performance of our fuzzy controller was demonstrated at the First International Robots Competition of AAAI, held in July, 1992 at San Jose, California, where Flakey was commended for its smooth and reliable reactivity, and placed overall second.