Neuro-Symbolic Partially Observable Stochastic Games: Modeling and Solving Continuous-State Multi-Agent Scenarios with Data-Driven Perception

The authors propose a novel model called one-sided neuro-symbolic partially observable stochastic games (NS-POSGs) that explicitly incorporates neural perception mechanisms for one of the agents in a continuous-state environment. They develop a solution method called one-sided NS-HSVI that exploits the piecewise constant structure of the model and leverages neural network pre-image analysis to construct finite polyhedral representations.