Learning Real-Time A* Path Planner for Unmanned Air Vehicle Target Sensing

This paper presents a path planner for sensing closely-spaced targets from a fixed-wing unmanned air vehicle (UAV) having a specified sensor footprint. The planner is based on the learning real-time A* (LRTA*) search algorithm and produces dynamically feasible paths that accomplish the sensing objectives in the shortest possible distance. A tree of candidate paths that span the area of interest is created by assembling primitive turn and straight sections of a specified step size in a sequential fashion from the starting position of the UAV. An LRTA* search of the tree produces feasible paths any time during its execution and minimum length paths if run to completion. The running time and path-length performance of the search are directly influenced by the operating parameters of the LRTA* algorithm. To improve the running time of the planner, a modified LRTA* search that terminates when there is no improvement in the path for a pre- defined number of iterations is implemented. The result is a path planner that produces short-distance paths in acceptably short running times.