Learning to Design Games: Strategic Environments in Deep Reinforcement Learning

In typical reinforcement learning (RL), the environment is assumed given and the goal of the learning is to identify an optimal policy for the agent taking actions through its interactions with the environment. In this paper, we extend this setting by considering the environment is not given, but controllable and learnable through its interaction with the agent at the same time. Theoretically, we find a dual Markov decision process (MDP) w.r.t. the environment to that w.r.t. the agent, and solving the dual MDP-policy pair yields a policy gradient solution to optimizing the parametrized environment. Furthermore, environments with discontinuous parameters are addressed by a proposed general generative framework. While the idea is illustrated by an extended two-agent rock-paper-scissors game, our experiments on a Maze game design task show the effectiveness of the proposed algorithm in generating diverse and challenging Mazes against different agents with various settings.