2-409 チェイン長さの相互関係に注目したマルチエージェント強化学習

概要

論文の詳細を見る
In multi-agent systems, one of the purposes is to find the way to acquire the cooperative behavior. In the process of the acquisition of the cooperative behavior, multiple agents learn simultaneously. Therefore, the certainty of partner action that is the ground of learning promotes the multi-agent learning. The macro rule was developed in order to be stabilized and to maintain the rule group applied continuously serially. It is generally used as a method of settlement for the perceptual aliasing problem. In the process of the acquisition of the cooperative behavior, however, a macro rule regards a partner agent's action also as a significant unit. This research shows that the cooperation nature of the length of the generated macro rule works effectively in the acquisition of the cooperative behavior in a pursuit problem.
社団法人日本機械学会の論文
2002-11-14