ABSTRACT
Researchers have recognized that an approach with more potential for the development of cooperative control mechanisms is autonomous learn ing. Most of the learning research to date is in the area of multi agent learning (e.g., [504]). Chapter 3 of this book describes multi-agent systems from a machine-learning perspective. However, much less work has been done in the area of multiro&o£ learning, although the topic is gaining in creased interest. The multirobot learning problem is particularly challeng ing due to a number of issues, including: very large state spaces, uncertain credit assignment, limited training time, uncertainty in sensing and shared information, non-deterministic actions, difficulty in defining appropriate abstractions for learned information, and difficulty of merging information learned from different robot experiences. However, the potential benefits are significant, and include increased robustness, reduced complexity, and increased ease of adding new assets to the team.
