Adaptive Sensor Fusion for Realistic and Autonomous Seabed Models

As autonomy, duration, and complexity of UUV missions increases, so too does the need for access to higher fidelity simulation and planning tools to ensure mission critical success. Advanced UUV fleets are critical for maintaining future subsea military dominance, and the availability of suitable simulation environments for technology and autonomy development is limited, in part due to the significant manual labor required to process and generate simulation models. An adaptive data processing method that uses a variety of multi-modal seafloor sensor and survey data, autonomously generates continuous and environmentally accurate three-dimensional seafloor models, and intelligently inserts and applies man-made or structured obstacles, will not only provide faster and more efficient seafloor modeling but will allow for accelerated development of the autonomous subsea vehicles critical to our subsea forces and operations. The Makai team propose to address this problem by developing a simulation environment synthesis that uses raw data plus adaptive and machine learning software to generate a realistic and continuous three-dimensional high-fidelity seafloor model. The proposed effort will leverage MakaiÆs Digital Terrain Models (DTM) used in the worldÆs leading submarine cable planning software MakaiPlan, the state-of-the-art in hydrographic data processing, and team membersÆ expertise developing embedded, machine learning (ML) algorithms.