next up previous
Next: References Up: Towards Collaborative and Previous: Multi-agent Learning Methods

Discussion and Conclusion

 

Robotic Soccer is a rich domain for the study of multi-agent learning issues. There are opportunities to study both collaborative and adversarial situations. However, in order to study these situations, the agents must first learn some basic behaviors. Similar to human soccer players, they can first learn to make contact with a moving ball, then learn to aim it, and only then start thinking about trying to beat an opponent and about team-level strategies.

Having achieved some success at learning a low-level behavior we are currently working on extending it to limited multi-agent scenarios. Our future research agenda includes improving the low-level behaviors while simultaneously working on collaborative and adversarial learning issues. We will collect several different behaviors learned in a similar manner to the shooting described in this paper. Then we will use them to learn how to work as a part of a team against hostile opponents. We are currently in the process of implementing our learning techniques in the simulator that will be used in RoboCup '97 [Noda1995].



Peter Stone
Wed Jan 17 10:57:03 EST 1996