Evolution Versus Design: Controlling Autonomous Robots

This paper sets out and justifies a methodology for the development of the control systems, or ‘cognitive architectures)) of autonomous mobile robots. It will be argued that the design b y hand of such control systems becomes prohibitively dificult as complexity increases. The alternative approach of artificial evolution is presented. It is argued that the most useful basic building blocks for an evolved cognitive archite(*t ? ure are adaptive noise tolerant neural networks raiher than programs. These networks may be recurrent, and should operate in real time. Evolution should be incremental, using an extended and modified version of genetic algorithms. Time constraints mean that architecture evaluations must be largely done an simulation. Results from a simulation are presented. The pitfalls of simulations compared with reality is discussed, together with the importance of incorporating noise.