A Bayesian Approach to Robust Inverse Reinforcement Learning
This work addresses robust policy estimation in offline inverse reinforcement learning, which is incremental as it builds on existing model-based IRL approaches with a novel Bayesian formulation.
The paper tackles the problem of offline model-based inverse reinforcement learning by proposing a Bayesian framework that simultaneously estimates the expert's reward function and subjective model of environment dynamics, showing that the estimated policy achieves robust performance when the expert is believed to have an accurate model and outperforms state-of-the-art offline IRL algorithms in MuJoCo environments.
We consider a Bayesian approach to offline model-based inverse reinforcement learning (IRL). The proposed framework differs from existing offline model-based IRL approaches by performing simultaneous estimation of the expert's reward function and subjective model of environment dynamics. We make use of a class of prior distributions which parameterizes how accurate the expert's model of the environment is to develop efficient algorithms to estimate the expert's reward and subjective dynamics in high-dimensional settings. Our analysis reveals a novel insight that the estimated policy exhibits robust performance when the expert is believed (a priori) to have a highly accurate model of the environment. We verify this observation in the MuJoCo environments and show that our algorithms outperform state-of-the-art offline IRL algorithms.