ADAPTEAM: Teaming and Information Sharing Among Adaptive Battlefield Agents

Abstract : The ADAPTEAM project is to research and develop new techniques for Agent Adaptation in multi-agent systems. Such techniques must deal with the problems of team representation for agile organizational changes, collaboration among teammates for best knowledge sharing and task distribution, monitoring and detecting deficiencies in team performance, and rectifying and improving team knowledge and organizational structures. During the period of 07/97-03/01, the project has developed a graph representation for agent teams, where an adaptable organization is represented by roles (nodes), role-relationships (edges), and role-assignment (task distribution). A negotiation technique is developed for detecting and resolving conflict and inconsistent beliefs among agents and inspired a new research project for the Autonomous Negotiating Targets (ANT) program in DARPA/ITO. A distributed algorithm for monitoring team performance is developed and shown to be sound and complete (see the PhD dissertation at http://www.cs.cmu.edu/-galk/Publications). Furthermore, a distributed technique for dynamic task re-allocation is developed and experimented in distributed constraint satisfaction problems. Finally, the PI invented a new technique called 'Digital Hormones' for distributed control in self-reconfigurable systems. A U.S. Patent Disclosure (USC File #3157) has been applied for this technique.