Favoring the Evolution of Adaptive Robots Through Environmental Differentiation

In this paper we demonstrate how differentiated environmental conditions facilitate the evolution of better solutions in agents evolved for the capacity to solve the double-pole navigation problem. Moreover, we propose a new evolutionary algorithm that operates on the basis of multiple environmental niches characterized by randomly different environmental conditions. The evolving agents are distributed in their own niche but have the possibility to colonize the other niches. Agents evolved through this method outperform both agents evolved in fixed environmental conditions and agents evolved in always varying environmental conditions.