Andrew Garland and Richard Alterman
Learning in a dynamic, uncertain, multi-agent setting is a challenging task. In this work, we demonstrate techniques to convert noisy run-time activity into procedures useful for future problem-solving activities. The cooperative procedures created by this conversion process are stored in an agent resource called collective memory. We show how this memory enables agents to learn cooperative procedures beyond the scope of their first-principles planner. Finally, we give experimental results demonstrating that collective memory improves agent performance.