Evolving simple software agents: comparing genetic algorithm and genetic programming performance
Mark C. Sinclair · 1997
This paper investigates the relative efficiency of genetic algorithms and genetic programming in evolving simple software agents. The problem domain consists of an autonomous food-gathering agent placed on a square grid of hundred cells with food units spread evenly over the grid. Initial results show that evolving the agent using GP requires less effort than with GA. Nevertheless, further investigation revealed some interesting aspects. 1 Introduction The aim of this paper is to compare the relative efficiency of genetic algorithms (GAs) [1] and genetic programming (GP) [2] for the evolution of simple software agents. This comparison forms part of a larger project to evolve software agents for distributed routing and restoration in telecommunications networks [3]. Consequently, within the overall timescales of the larger project, we were only able to devote a modest amount of effort to the work reported here. As the basis for the comparison, we independently reimplemented a simplifie...