An Empirical Study of Coaching
In simple terms, one can say that team coaching in adversarial domains consists of providing advice to distributed players to help the team to respond effectively to an adversary. We have been researching this problem to find that creating an autonomous coach is indeed a very challenging and fascinating endeavor. This paper reports on our extensive empirical study of coaching in simulated robotic soccer. We can view our coach as a special agent in our team. However, our coach is also capable of coaching other teams other than our own, as we use a recently developed universal coach language for simulated robotic soccer with a set of predefined primitives. We present three methods that extract models from past games and respond to an ongoing game: (i) formation learning, in which the coach captures a team’s formation by analyzing logs of past play; (ii) set-play planning, in which the coach uses a model of the adversary to direct the players to execute a specific plan; (iii) passing rule learning, in which the coach learns clusters in space and conditions that define passing behaviors. We discuss these techniques within the context of experimental results with different teams. We show that the techniques can impact the performance of teams and our results further illustrate the complexity of the coaching problem.
KeywordsMultiagent System Formation Learning Intelligent Tutor System Learning Agent Extensive Empirical Study
Unable to display preview. Download preview PDF.
- 1.P. Bakker and Y. Kuniyoshi. Robot see, robot do: An overview of robot imitation. In AISB96 Workshop on Learning in Robots and Animals, pages 3–11, Brighton, UK, 1996.Google Scholar
- 3.P. Cheeseman, J. Kelly, M. Self, J. Stutz, W. Taylor, and D. Freeman. Auto-class: A bayesian classification system. In ICML-88, pages 54–64, San Francisco, June 1988. Morgan Kaufmann.Google Scholar
- 4.J. Clouse. Learning from an automated training agent. In D. Gordon, editor, Working Notes of the ICML ′95 Workshop on Agents that Learn from Other Agents, Tahoe City, CA, 1995.Google Scholar
- 8.R. Maclin and J. W. Shavlik. Creating advice-taking reinforcement learners. Machine Learning, 22:251–282, 1996.Google Scholar
- 9.M. S. Miller, J. Yin, R. A. Volz, T. R. Ioerger, and J. Yen. Training teams with collaborative agents. In ITS-2000, pages 63–72, 2000.Google Scholar
- 11.M. Paolucci, D. D. Suthers, and A. Weiner. Automated advice-giving strategies for scientific inquiry. In ITS-96, pages 372–381, 1996.Google Scholar
- 12.J. R. Quinlan. C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo, CA, 1993.Google Scholar
- 13.T. Raines, M. Tambe, and S. Marsella. Automated assistant to aid humans in understanding team behaviors. In Agents-2000, 2000.Google Scholar
- 14.P. Riley and M. Veloso. Planning for distributed execution through use of probabilistic opponent models. In AIPS-2002, 2002. (to appear).Google Scholar
- 15.P. Riley, M. Veloso, and G. Kaminka. Towards any-team coaching in adversarial domains. In AAMAS-02, 2002. (extended abstract) (to appear).Google Scholar
- 17.C. Sammut, S. Hurst, D. Kedzier, and D. Michie. Learning to fly. In ICML-92, Aberdeen, 1992. Morgan Kaufmann.Google Scholar
- 20.D. Šuc and I. Bratko. Skill reconstruction as induction of LQ controllers with subgoals. In IJCAI-97, pages 914–919, 1997.Google Scholar