Towards Flexible Teamwork in Persistent Teams: Extended Report

摘要

Teamwork is a critical capability in multi-agent environments. Many such environments mandate that the agents and agent-teams must be persistent i.e., exist over long periods of time. Agents in such persistent teams are bound together by their long-term common interests and goals. This paper focuses on flexible teamwork in such persistent teams. Unfortunately, while previous work has investigated flexible teamwork, persistent teams remain unexplored. For flexible teamwork, one promising approach that has emerged is model-based, i.e., providing agents with general models of teamwork that explicitly specify their commitments in teamwork. Such models enable agents to autonomously reason about coordination. Unfortunately, for persistent teams, such models may lead to coordination and communication actions that while locally optimal, are highly problematic for the team's long-term goals. We present a decision-theoretic technique based on Markov decision processes to enable persistent teams to overcome such limitations of the model-based approach. In particular, agents reason about expected team utilities of future team states that are projected to result from actions recommended by the teamwork model, as well as lower-cost (or higher-cost) variations on these actions. To accommodate real-time constraints, this reasoning is done in an any-time fashion. Implemented examples from an analytic search tree and some real-world domains are presented.