Developing Autonomy for USVs by Using Virtual Environments [ETC/UMD Team]
The Department of Defense and the Department of the Navy have significant interest in developing long endurance, autonomously controlled Unmanned Surface Vehicles [USVs] that can operate in complex, dynamic and unpredictable ocean environments. In addition, DOD and DoN desire these USVs to both be able to anticipate, respond, and preempt non-cooperating and even hostile behaviors. One example is the use of autonomous USVs to defend of high value vessels from unknown/hostile intelligent intruders.
Under ONR funding our team developed a virtual environment physics based meta-models, six degree-of freedom dynamics models, USV geometric model simplification and computational acceleration techniques (speed up computations by factor of 28) including use of parallelized Graphical Processing Units [GPU], trajectory planners enabling safe, efficient, and dynamically feasible trajectory planning in high sea states resulting in significant autonomy synthesis capabilities. These capabilities also enabled evaluation of human vs. human and human vs. autonomy scenarios. We successfully demonstrated the use of evolutionary computation and statistical reasoning to automatically synthesize USV blocking behavior against a human-competitive deceptive intruder, and an intruder with probabilistic actions. The computer synthesized algorithmís behavior was comparable to that obtained hand-coded behavior.
Our modular software virtual environment included real-time rendering of ocean environments, modeling of boats, terrain, and shorelines. The automated model simplification applies to GPU computing and general CPU computing. Simplification is based on clustering and temporal coherence technique. We are able to compute trajectories for various sea states and incorporate look-ahead into trajectory planning under motion uncertainty for USVs. We are currently funded to extend this work and our COLREGS [Coast Guardís Collision Avoidance regulation] algorithms and implement those on Florida Atlantic Universityís [FAU] experimental Unmanned Surface Vehicle.
10400 O'Donnell Place, Suite 202
St. Charles, MD 20603
Copyright 2012, ETC, all rights reserved.
Send comments & questions to the ETC Webmaster
ETC, as an independent nonprofit Equal Opportunity Employer, is classified as a Section 501(c)(3) organization under the Internal Revenue Code, and is recognized as a publicly supported organization. Individuals, corporations, companies, associations, and foundations are eligible to support the work of ETC through tax-deductible gifts