Collaboration, Self-Reflection, and Adaptation in Robot Communities: Using Multi-Agent Distributed Learning for Coordination Planning