CONSTRUCTION OF OPTIMAL-PATH MAPS FOR HOMOGENEOUS-COST-REGION PATH-PLANNING PROBLEMS
ROBERT S. ALEXANDER · Calhoun: The Naval Postgraduate School Institutional Archive (Naval Postgraduate School) · 1989
Fast path-planning algorithms are needed for autonomous vehicles and tactical terrain-analysis tools. We explore a new approach using "optimal-path maps", that give the best path to a goal point from any given start point in cross-country two-dimensional terrain for a moving agent of negligible size. Such maps allow fast point-location algorithms at run-time to categorize die start point according to the behavior of the optimal path to the goal, from which the path can be reconstructed. We study terrain modelled by piecewise-linear roads and rivers, polygonal obstacles, and by convex polygonal homogeneous-cost areas ("weighted regions"). We explore two methods for constructing optimal-path maps, one based on wavefront-propagation point-to-point path planning, and a more exact divide-and-conquer algorithm that reasons about how optimal paths must behave. In the exact approach, boundaries caused by terrain features are characterized using analytical geometry and optimal-path principles, and partial optimal-path maps are merged into complete ones.