Maximal Distance Discounted and Weighted Revisit Period: A Utility Approach to Persistent Unmanned Surveillance

Autonomous unmanned vehicles are well suited for long-endurance, persistent intelligence, surveillance and reconnaissance (PISR) missions. In order to conduct missions, vehicles must implement a method of task selection. We propose the Maximal Distance Discounted & Weighted Revisit Period ([Formula: see text]) utility function as a solution. We derive [Formula: see text] as a zeroth-order approximation to an infinite horizon solution of PISR when formulated as a dynamic programming (DP) problem. We then use the DP solution to develop a heuristic utility function for autonomous task selections, with the goal of minimizing the prioritized revisit time to each task. Our function adapts to different task maps and task priorities, is scalable in the number of tasks, and is robust to the ad-hoc addition or removal of tasks. We demonstrate how the [Formula: see text] parameters influence vehicle behavior. We also prove that the policy results in steady-state task selections that are periodic and that such periodicity occurs regardless of initial conditions. We then demonstrate periodicity via numerical simulations on a set of test scenarios. We present a two-step heuristic methodology for selecting utility function parameters that deliver empirically good performance, which we demonstrate through a simulation-based comparison to a single-vehicle Traveling Salesman Problem (TSP) solution. The comparisons are based on four sample task maps designed to resemble operational scenarios.