Topside Optical Processing for Global Unmanned Navy (TOPGUN)

Subsequent Phase II Proposal, extension of Phase I contract N00014-13-P-1152.Under the ACTUV program, DARPA developed a strategically critical autonomous anti-submarine warfare vessel, called Sea Hunter. One key to Sea Hunter’s success is an ability to avoid collisions with manned vessels by following the International Regulations for Preventing Collisions at Sea (COLREGS). However, autonomous COLREGS compliant navigation presents two main challenges: reliably detecting other vessels and classifying them according to their COLREGS class, to give certain vessel types (e.g., sailboats and fishing boats) the “right of way.”We propose to develop and integrate a system for automatic detection and classification supporting COLREGS compliance to address this need. Our approach builds on two previous efforts: 1) our Phase I Taxiing Operations via Gesture Understanding (TOPGUN-I) SBIR, wherein we developed a data-driven Object Detection Framework (ODF); and 2) our CROWSNEST effort under the ACTUV program, wherein we used the ODF for fast and reliable ship detection and convolutional neural networks for robustly classifying ships into the appropriate COLREGS classes based on their appearance. Our proposed TOPGUN effort focuses on collecting additional training data, refining existing algorithms, hardening system software, and supporting testing in representative environments using the final Sea Hunter sensors.