AMBUSH AI-based Ping Strategies in Multistatic Active Sonar

Triton proposes to develop AMBUSH, a neurosymbolic AI-based system that will enable the expandable, scalable, and systematic research of formal logic-based and reinforcement-learning algorithms for the exploration of intelligent ping strategies in realistic multistatic active sonar fields operating in pursuit of a reacting and intelligent target that engages in rationally decided detection evasion maneuvers. During Phase I, a proof-of-concept version of the AMBUSH AI system will be developed and demonstrated as a simulation-environment-based ping strategy evaluation tool. It will model a realistic multistatic active sonar field. The active sonobuoy field model and its programmable waveform capabilities will be augmented by high-fidelity location-specific underwater sound propagation modeling. The AMBUSH system’s AI will then generate and exercise multiple reinforcement learning-based techniques to automatically plan, schedule and monitor the progress of intelligent ping strategies in pursuit of its objective, which is to shorten the detect-to-kill chain time. Moreover, the target will be model-programmed so as to exhibit smart reactive behavior.