Mobile Agent coordination is important in applications, which are both time and space bound, such as disaster recovery or demining. In such environments, each agent is required to be mobile to cover a region around it. In order to cover the entire area, all agents should collectively ensure full coverage of the area of interest, for which infrastructure for coordination across mobile agents may not be readily available. Knowing when to communicate and what to communicate becomes therefore becomes a key learning exercise. In such a scenario, it is interesting to study the evolution of a communication hierarchy in order to accomplish a multi-agent coordination task such as area coverage. In this paper, we propose a coordination hierarchy for a survivor rescue task and overlay a learning framework to teach a team of agents what to communicate and when to communicate while recovering the survivors efficiently. We then study the evolution of a communication hierarchy derived from the agent learning.
Read full abstract