TOYON RESEARCH CORPORATION — Department of Defense SBIR Phase I: MDA21-021

Description

Toyon Research Corporation proposes the application of Deep Reinforcement Learning to optimize salvo management against a raid scenario by leveraging a two-sided learning framework. In Phase I, Toyon will employ a simulator that models launch and flight of interceptors, missile launch and exo-atmospheric deployment of objects, and, finally, the success of any engagement. Existing battle simulator assets will be leveraged to develop the proof-of-concept simulator. Once available, the simulator will be packaged into a new training environment for use in Toyon’s Modular Autonomy Incubator (MAUI), a Deep Reinforcement Learning framework. Agents will be trained using adversarial multi-agent self-play. This state-of-the-art technique allows each agent to both learn and master complicated tasks from the experience gained while competing in a simulated environment. In Phase II, the simulation environment will be enhanced to support higher-fidelity models and increased functionality. Using the enhanced simulator, agents will be further trained over millions of simulated mission sets using Toyon’s MAUI framework and high-performance computing cluster. Approved for Public Release | 21-MDA-11013 (19 Nov 21)