Evolutionary Robotics

Abstract

In this chapter we discuss evolutionary robotics (ER), where evolutionary algorithms are employed to design robots. Our emphasis lies on the evolutionary aspects, not on robotics per se. Therefore, we only briefly discuss the ER approaches that work with conventional evolutionary algorithms to optimize some robotic features and pay more attention to systems that can give rise to a new kind of evolutionary algorithms. In particular, we consider groups of mobile robots whose features evolve in real-time, for example, a swarm of Mars explorers or ‘robot moles’ mining ore deep under the surface. In such settings the group of robots is a population itself, which leads to interesting interactions between the robotic and the evolutionary components of the whole system. For robotics, this new kind of ER offers the ability to evolve controllers as well as morphology in partially unknown and changing environments on the fly. For evolutionary computing, autonomous mobile robots provide a special substrate for implementing and studying artificial evolutionary processes in physical entities going beyond the digital systems of today’s evolutionary computing.