Communication Decision in Decentralized Control of Coordinated System

Although instantaneous and free communication simplifies multi-agent partially observable Markov decision processes (POMDPs) to single-agent POMDPs, in practice, communication is not free and reducing the amount of communication is often desirable. We present a novel approach for using a DAG to maintain and reason the possible joint beliefs of the team at the less cost of memory, by which communication decisions are made in a decentralized style. Our algorithm, called DAG-DEC-COMM, makes a tradeoff between the performance achieved by the team and the amount of communication used in planning. The experimental results on the benchmark problem show that our approach drastically reduces the communication with a significant decrease in storing the possible joint beliefs while improving the performance of distributed execution.