Multi-sensor and Acoustic Contact Localization through Artificial Intelligence/Machine Learning

This proposal outlines the development of an advanced AI/ML-based TMA algorithm to estimate a contactís bearing, range, course, and speed using only information from promoted passive sensor contact followers.ÝThe objective of Phase I is the development of TRL-3 proof of concept AI/ML TMA algorithm that improves accuracy of and confidence in state estimation over that of current capabilities by a reduction of ~20% in estimation error using promoted CFs, ship and array parameters, and environmental data. Phase II will be the completion of a TRL-6 ready to integrate prototype AI/ML TMA algorithm. Additional capabilities to include simplification and improved visualization of eventual information provided to operators in the development of a tactical picture, selection of effective TMA maneuvers, extension to non-promoted CFs, and extension to other systems such as the AN/SQQ-89 or use on unmanned surface or undersea vessels (USVs or UUVs) will be developed in Phase III if awarded.ÝThis proposal delivers a comprehensive solution to the Navyís need for advanced TMA capabilities, leveraging the strengths of PINNs, BNNs, RNNs, and transformers. The incremental training approach, combined with physical constraints and uncertainty quantification, ensures robust and reliable decision-making in complex acoustic environments. Furthermore, the CF generation tool will be used to create high-fidelity synthetic acoustic data for initial (Phase I) algorithm training and, if needed, to supplement real-world data in Phase II, ensuring robustness across diverse environments and scenarios. The target APB will be APB-27 or 29. The proposed AI/ML systemís adaptability positions it for future adoption across Navy ASW systems and unmanned sensor platforms. With the potential for expansion to provide indications of the best sources of data, assessment of overall localization uncertainty across solution sources to support tactical decisions.