Decentralized Interactive Planning and Sensing in an Unknown Spatiotemporal Threat Field

We address path-planning for a mobile vehicle, called the actor, to minimize exposure to an unknown spatiotemporally varying scalar threat field. Multiple sensor agents commanded by the actor take pointwise measurements and estimate the threat field using a consensus filter. The actor and sensors communicate over a network with fixed distance-based topology. We introduce a novel interactive planning and sensing algorithm, where sensor agents are iteratively reconfigured to minimize the actor’s cost. Two sensor reconfiguration policies are proposed: one that enforces all sensors to remain within the actor’s communication network, and the other that allows sensors to be deployed farther away. Through numerical simulation results we demonstrate that the proposed algorithm achieves near-optimal performance for the actor.