Bayesian Multiagent Active Sensing and Localization via Decentralized Posterior Sampling

In many applications, including situational awareness and surveillance, it is often desirable that a small number of intelligent and mobile sensors efficiently cover a large area via active sensing. Moreover, for many national security applications, uncertainty quantification is critical for decision making. Additionally, power and bandwidth constraints and robustness to a single point of failure typically require fully distributed processing. In this paper, we provide a fully distributed Bayesian active sensing framework where the sensors move to collect the most informative measurements via distributed optimization of the Bayesian Fisher information as well as providing uncertainty quantification at every step via the full joint posterior distribution. We also provide extensive simulation results demonstrating the efficacy of our proposed approach for the localization of multiple passive targets.