Despite significant progress in the Belief-desire-intention theories of teamwork and their applications, existing multiagent teamwork research suffers from at least two serious shortcomings. First, there is no quantitative analysis of the optimality of its prescriptions for agent behavior in teamwork. Second, there is no analysis of the complexity of the general teamwork problem or specific team coordination decisions. To address these shortcomings, we propose a novel framework, the COMmunicative Multiagent Team Decision Problem (COM-MTDP). COM-MTDP is based on the decision-theoretic work in economic team theory, but with significant extensions to introduce dynamics and interagent communication. As a result, at its core, COM-MTDP is driven by communicating partially observable markov decision processes (POMDPs). The projected impact of the development of COM-MTDP will be a new, general framework for theoretical and empirical analysis of teamwork. Thus, rather than simply providing teamwork heuristics, researchers could analyze the computational complexities of the optimal team problem, and analyze the complexity-optimality tradeoffs in proposed team coordination heuristics and implementations. Such analysis would lead to improved, robust, general team coordination algorithms that explicitly reason about such tradeoffs. Finally, COM-MTDP ideas could potentially be more generally applicable (e.g., in analysis of multiagent contracting algorithms.)
