Using Genetic Programming to Evolve Robot Behaviours

Abstract

This paper presents the application of genetic programming (GP) to the task of evolving robot behaviours. The domain used here is the well-known wall-following problem. A set of programs were evolved that can successfully perform wall-following behaviours. The experiments involving different wall shapes were designed and implemented to investigate whether the solutions offered by GP are scalable. Experimental results show that GP is able to automatically produce algorithms for wall-following tasks. In addition, more complex wall shapes were introduced to verify that they do not affect our GP implementation detrimentally.