Signal Systems Corporation (SSC) will utilize recent advances in deep reinforcement learning to develop novel ping strategies under "Disruptive Autonomy Against Reactive Targets" (DAART). SSC will leverage its extensive experience in developing ping control algorithms, deep learning, and simulations for acoustic air anti-submarine warfare (ASW) systems to develop an intelligent ping controller which learns to maximize the success of search missions. SSC will increase the fidelity of the simulation by developing a reactive target model which utilizes local environmental knowledge and observations of direct blasts to plot an evasive course and set a speed and depth which minimize detection while moving towards a goal. Key SSC innovations include adapting deep reinforcement learning research to the acoustic air ASW domain, and developing a flexible approach that easily incorporates new capabilities, and uses knowledge from one system to decrease training times when moving to new systems.
Benefit: Improved search performance, better performance evaluation, improved automation for air ASW. Commercial applications: radar scheduling, automated mission planning
Keywords: Reinforcement Learning, Reinforcement Learning, Sonar, Anti-Submarine Warfare, Artificial Intelligence, Automation, Deep Learning, Multistatic, Planning
