Communication, Diversity and Learning: Cornerstones of Swarm Behavior
This paper reviews research in three important areas concerning robot swarms: communication, diversity, and learning. Communication (or the lack of it) is a key design consideration for robot teams. Communication can enable certain types of coordination that would be impossible otherwise. However communication can also add unnecessary cost and complexity. Important research issues regarding communication concern what should be communicated, over what range, and when the communication should occur. We also consider how diverse behaviors might help or hinder a team, and how to measure diversity in the first place. Finally, we show how learning can provide a powerful means for enabling a team to master a task or adapt to changing conditions. We hypothesize that these three topics are critically interrelated in the context of learning swarms, and we suggest research directions to explore them.
KeywordsReward Function Learning Team Home Base Goal Communication Control Team
Unable to display preview. Download preview PDF.
- 4.Mataric, M.J.: Kin recognition, similarity, and group behavior. In: Proceedings of the Fifteenth Annual Cognitive Science Society Conference, pp. 705–710 (1993)Google Scholar
- 5.MacKenzie, D.C., Balch, T.: Making a Clean Sweep: Behavior Based Vacuuming. In: Working Notes of the 1993 AAAI Fall Symposium Instantiating Real-World Agents (1993)Google Scholar
- 7.Parker, L.: Heterogeneous Multi-Robot Cooperation. Massachusetts Institute of Technology Ph.D. Dissertation. Available as MIT Artificial Intelligence Laboratory Technical Report 1465 (1994)Google Scholar
- 9.Coordinating mobile robot group behavior using a model of interaction dynamics. In: Proceedings, The Third International Conference on Autonomous Agents (Agents 1999), pp. 100–107 (1999)Google Scholar
- 11.Watkins, C., Dayan, P.: Q-Learning. Machine Learning 8, 279–292Google Scholar
- 13.Zinkevich, M., Balch, T.: Symmetry in Markov Decision Processes and its Implications for Single Agent and Multiagent Learning. In: International Conference on Machine Learning, pp. 632–642 (2001)Google Scholar