抄録
Learning and evolving in artificial agent is an extremely difficult problem, but on the other hand, a challenging task. At present the studies mainly centered on single agent learning problem. In our case, we use simulated soccer to investigate multi-agent cooperative learning. Consider the fundamental differences in learning mechanism, existing reinforcement learning algorithms can be roughly classified into two classes-that based on evaluation functions and that of searching through policy space in direct. Genetic Programming developed from Genetic Algorithms is one of the most well known approaches that belong to the latter. In this paper, we give detailed algorithm as well as data construction description that are necessary for learning single agent strategies at first. In the following sections, we extend developed methods into multiple robot domains moreover. We investigate and contrast three different solutions-single agent learning, simple team learning and sub-group learning and conclude the paper with some actual experiments and result analyses.
