Minefield Search and Object Recognition for Autonomous Underwater Vehicles.

摘要

Autonomous Underwater Vehicles (AUV) are an outstanding minefield search platform. Because of their stealthy nature, AUVs can be deployed in a potential minefield without the enemy's knowledge. They also minimize dangerous exposure to manned and more expensive naval assets. This thesis explores two important and related aspects of AUV minefield search: exhaustive sensor coverage of a minefield through effective path planning and underwater object recognition using the vehicle's sensors. The minefield search algorithm does not require a priori knowledge of the world except for user-defined boundaries. It is a three-dimensional, prioritized graph search using a ladder based methodology and an A* optimal path planning algorithm. The minefield search algorithm effectively ignores areas which are blocked by obstacles, performs terrain following and avoids local minima problems encountered by other area search solutions. The algorithm is shown to be effective using a variety of graphical simulators. The object recognition algorithm provides autonomous classification of underwater objects. It uses geometric reasoning and line fitting of raw sonar data to form geometric primitives. These primitives are analyzed by a CLIPS language expert system using heuristic based rules. The resulting classifications may be used for higher level mission planning modules for effectively conducting the minefield search. Actual NPS AUV swimming pool test runs and graphic simulations are used to demonstrate this algorithm which was built in cooperation with Lieutenant Commander Donald P. Brutzman, USN.