Planning Efficient Paths through Dynamic Flow Fields in Real World Domains

Abstract

This research addresses the problem of planning efficient paths for agents through flow fields in small real-world domains where vehicle dynamics and environmental uncertainty can significantly affect the optimality of a path. In particular, we consider the task of planning routes for small autonomous airboats deployed in various river domains so as to best take advantage of water currents to save energy and time. Existing planning techniques for flow fields were implemented on our airboat platform and evaluated on these domains with current models developed from the data gathered using a Nortek AD2CP-Glider acoustic doppler current profiler. The real-world performance of these algorithms were compared to theoretical estimates and several modifications are suggested to improve their performance in specific domains.