Depth-First Coupled Sensor Configuration and Path-Planning in Unknown Static Environments

We address path-planning for a mobile agent in an unknown static environment. The environment is observed by a sensor network where each sensor has a configurable location and field of view. We propose a depth-first coupled sensor configuration and path-planning (DF-CSCP) iterative method, which iteratively finds an optimal sensor configuration (location and FoV), applies Gaussian Process Regression to construct a threat field estimate, and then finds a candidate optimal path with minimum expected threat exposure. The DF-CSCP method uses a two stage procedure, (1) Explore and (2) Exploit, to drive the uncertainty of the candidate path cost variance below a prespecified threshold. To maintain tractability of GPR with increasing number of measurements, we present a sparse-update scheme. The proposed method relies on novel task-driven information gain (TDIG) metrics, the maximization of which provides sensor configurations. The TDIG metric quantifies the importance of acquiring sensor data of highest relevance to the path-planning task. Through numerical studies, we demonstrate the technical results that the DF-CSCP algorithm finds near-optimal paths with significantly fewer sensor measurements compared to traditional information-maximization methods.