Learning Complementary Multiagent Behaviors: A Case Study

As the reach of multiagent reinforcement learning extends to more and more complex tasks, it is likely that the diverse challenges posed by some of these tasks can only be addressed by combining the strengths of different learning methods. While this important aspect of learning is yet to receive theoretical analysis, useful insights can be gained from its applications to concrete tasks. This paper presents one such case study grounded in the robot soccer context. The task we consider is Keepaway, a popular benchmark for multiagent reinforcement learning. Whereas previous successful results in this domain have limited learning to an isolated, infrequent decision that amounts to a turn-taking behavior (passing), we expand the agents’ learning capability to also include a much more ubiquitous action (moving without the ball, or getting open), such that at any given time, multiple agents are executing learned behaviors simultaneously. We introduce a policy search method for learning â...

• Type: Conference
• Year: 2009
• Where: ROBOCUP
• Authors: Shivaram Kalyanakrishnan, Peter Stone
• Keywords: Different Learning Methods, Multiagent Reinforcement, Reinforcement Learning, ROBOCUP 2009, Robotics