Co-evolution of cooperation in a pursuit evasion game

This research concems the comparison of different artificial evolution approaches to the design of cooperative behavior in a team of simulated mobile robots co-evolved against a second team. The first and second approaches, termed single pool and plasticiry, are characterized by robots that share a single genotype, though the plasticity approach includes a leaning mechanism. The third approach, termed multiple pools, is characterized by robots that use different genotypes. The application domain is a pursuit-evasion game in which a team of three robots termed pursuers, collectively work to immobilize one of the three robots of the other team, termed: evaders. Results indicate that the multiple p l s approach applied within a competitive co-evolution process yields superior performance comparative to the other approaches. Specifically, the coevolutionary process allows the multiple pools approach to 'bootstrap' complementary behavioral roles, facilitating the evolution of a stable cooperative pursuit strategy. evolutionary robotics experiments using two mobile robots. This co-evolution scenario was compared with single agent evolution. A comparatively faster evolution of more diverse behavioral strategies was observed with competitive coevolution. Floreano et al. [1998] extended this research with a physical implementation of this predator-prey co-evolution using Khepera robots. With notable exceptions such as [lha, 19961, [Haynes and Sen, 19971, [Yong and Miikkulainen, 20011, [Stone and Veloso, 19981 few researchers have investigated emergent cooperative behavior in a competitive co-evolutionary context, where two teams co-evolve cooperative behavior that ensures the survival of each team. This paper describes a comparison of three artificial evolution approaches for the synthesis of cooperative behaviour evaluated within a team of simulated Khepera robots [Mondada et al. 19931, co-evolved against a second team. For each artifiicial evolution approach, a team of three pursuers implementing initially random bebavior is competitively coevolved with a team of three evaders implementing an obstacle avoidance behavior. The pursuers and evaders are functionally different in terms of sensor and movement capabilities. The team of evaders is able IO move 40 percent faster than the team of pursuers. F'nrsuen are rewarded fitness proportional to how much they are able to slow down the evader team on average, and the evader team is rewarded fitness propoaional to how fast it is able to move