Automatic agent design based on gate growth-application to wall following problem

Abstract

Looks at an evolutionary approach to robotics. A reactive architecture leads robots moving in real world adaptively. However, designing adequate complete rules is practically impossible, since the interaction is unlikely to be completely known. To approach this difficult design problem, a new approach "gate growth" is presented. This methodology consists of two components: a cellular automaton (CA) as the connection algorithm and a NAND array as the device model. Through GAs operation, the rules forming CA is evolved to achieve the given tasks. In the application to a wall following problem, the possibility of building task achieving behavior is shown.