Piloting Intelligence using Logic, Optics, and Tracking (PILOT)

The Navy has worked for years to develop uncrewed surface vessels (USVs) that can operate in open oceans, but less work has gone into autonomous operations in harbors. Harbor piloting is a challenge due to constrained channels, static obstacles, and dynamic and dense ship traffic. In harbors, sensing is a challenge due to clutter and complex backgrounds, and path planners must frequently update paths to match constantly changing situations. To provide the perception and responsive path planning necessary for uncrewed harbor piloting, we propose Piloting Intelligence using Logic, Optics, and Tracking (PILOT). PILOT is a modular harbor piloting software system that integrates digital charts and tide information, real-time perception, and expert knowledge of ship dynamics and navigation rules (e.g., COLREGS) to plan appropriate, responsive, and safe paths through harbors. PILOT’s perception system uses deep learning, projective geometry, and probabilistic tracking with electro-optical (EO) and infrared (IR) imagery to effectively detect, classify, localize, and track surface contacts, day or night, even in cluttered backgrounds. PILOT’s autonomy develops path plans using adaptive, behavior-based autonomy built on multi-objective optimization. PILOT will be Modular Open Systems Approach (MOSA)-compliant, and the interfaces will be designed for Unmanned Maritime Autonomy Architecture (UMAA) compliance to facilitate transition.