Frontiers in Evolutionary Robotics (Repost)

Some problems can be efficiently solved only by teams consisting of cooperative autonomous players (robots). Many researchers have developed methods that do not require human designers to define specific behaviors of players for each problem. The work reported in this chapter focuses on the techniques of evolutionary computation, which has been regarded as one of the most promising approaches to solving such complex problems. However, in using evolutionary computation for generating players performing tasks cooperatively, one faces fundamental and difficult decisions, including the one regarding the so-called credit assignment problem (Haynes et al., 1995). For example, if we can only evaluate the global performance of each team, how do we divide up the team’s performance among the participating players? We believe that there are some correlations among design decisions, and therefore a comprehensive evaluation of them is essential, although several researchers have proposed evolutionary methods for evolving teams performing specific tasks. This chapter is organized as follows. In Section 2, we list three fundamental decisions and possible options in each decision in designing a method for evolving a cooperative team. We find that there are 18 typical combinations available. Then, in Section 3, we describe the ultimately simplified soccer game played on a one-dimensional field as a testbed for comparative evaluation of these 18 candidate methods. Section 4 reports on the results of the comparative evaluation of these methods, and Section 5 summarizes the work.