Distributed information-theoretic target detection using physics-inspired motion coordination

In this paper we describe a distributed, information theoretic motion planning strategy for multi-agent target detection. Agents assimilate measurements into a likelihood-ratio tracker, which provides a probability distribution for potential target locations. Information from other agents is fused with the local agents' probability density using an Information Weighted Consensus Filter when in communication range. Each agent uses a physics-inspired motion planning strategy to reactively cover the domain and informatively gather measurements based on the posterior of the likelihood ratio tracker. The proposed strategy produces emergent behavior that optimally collects information about the environment in a reactive and scalable manner that is resilient to communication drop outs. The algorithm is tested in simulation to verify the performance.