Interactive planning and sensing in unknown static environments with task-driven sensor placement

We address planar path-planning for a mobile vehicle to traverse a planar workspace with minimum exposure to a spatially varying scalar field called the threat field. The threat field is unknown, and is estimated by a finite number of sensors that take pointwise noisy measurements. We propose an iterative sensor placement and path-planning algorithm. At each iteration, regions of interest in the workspace are identified near the path with minimum estimated cost, and the next set of sensor locations is determined to improve the confidence of threat field estimates in these regions of interest. This paper breaks new ground in dynamic data-driven autonomy, in that an explicit bidirectional interaction between a path-planning algorithm and a sensor network separate from the actor vehicle is established. We provide theoretical results and corroborative numerical simulation results on the convergence and optimality of the proposed iterative algorithm. We show that the proposed algorithm provides significant improvements over existing approaches, including information-maximizing sensor placement methods.