Abstract
The central issue in the design of cooperative multi-agent systems is how to coordinate the behavior of the agents to meet the goals of the designer. Traditionally, this had been accomplished by hand-coding the coordination strategies. However, this task is complex due to the interactions that can take place among agents. Recent work in the area has focused on how strategies can be learned. Yet, many of these systems mainly put emphasis on the learning strategies of individual agent. So they suffer from the complexity, convergence and time-consuming problems and so on. This paper presents a novel learning method named Behavior-based On-line Learning on Multi-agent Domain. As tests bench a kind of multiple micro-robots soccer-playing system is built up in our laboratory. The construction of such experimental system involves kinds of challenges. In this paper we want to focus on how to evolve the coordination among agents automatically based on our new learning model. At first we give a brief introduce of our experimental system and present learning model in common sense, and then explain how to apply it to the experimental system. In end we will give some results showing that the proposed approach is feasible to guide the design of common multiple agents system.
