Learning Environment with Surrogate Simulation to Optimize Naval Planning (LESSON-Plan)

Modern Warfighters in the Navy must make time-sensitive decisions on which courses of action (COAs) to take when executing a find¢fix¢track¢target¢engage¢assess (F2T2EA) kill chain. However, due to a myriad of complex environmental and mission factors/constraints, determining an optimal COA is challenging and time consuming, which leads to delays in the execution of kill chains. These delays can jeopardize missions and put human lives at risk, especially in time-sensitive situations where decisions must be made quickly and confidently. To expedite these kill chains, the Maritime Tactical Command and Control (MTC2) system is currently building upon modern artificial intelligence and machine learning (AI/ML) techniques to automate the COA assessment process. Unfortunately, contemporary AI/ML approaches are either not fast enough to support time sensitive missions, or too inaccurate to provide meaningful insights. Furthermore, many state-of-the-art approaches fail to develop trust with the end user, leading to AI hesitancy. To address these hurdles, the Learning Environment with Surrogate Simulation to Optimize Naval Planning (LESSON-Plan) uses novel techniques in hybrid AI, extending a combination of subject matter expertise in conjunction with reinforcement learning (RL), to train both a surrogate simulation model and a COA model that determines optimal COAs given mission and environmental data.