Adversarial Imitation via Variational Inverse Reinforcement Learning
This work addresses the problem of robust imitation learning for AI agents in complex environments, offering improved generalization over existing inverse reinforcement learning algorithms.
The paper tackles learning rewards and policies from expert demonstrations under unknown dynamics by introducing empowerment-regularized maximum-entropy inverse reinforcement learning, achieving near-optimal rewards and policies that outperform state-of-the-art methods on high-dimensional control tasks and transfer learning scenarios.
We consider a problem of learning the reward and policy from expert examples under unknown dynamics. Our proposed method builds on the framework of generative adversarial networks and introduces the empowerment-regularized maximum-entropy inverse reinforcement learning to learn near-optimal rewards and policies. Empowerment-based regularization prevents the policy from overfitting to expert demonstrations, which advantageously leads to more generalized behaviors that result in learning near-optimal rewards. Our method simultaneously learns empowerment through variational information maximization along with the reward and policy under the adversarial learning formulation. We evaluate our approach on various high-dimensional complex control tasks. We also test our learned rewards in challenging transfer learning problems where training and testing environments are made to be different from each other in terms of dynamics or structure. The results show that our proposed method not only learns near-optimal rewards and policies that are matching expert behavior but also performs significantly better than state-of-the-art inverse reinforcement learning algorithms.