Q-based Variational Inverse Reinforcement Learning

Ondrej Bajgar, Konstantinos Gatsis, Alessandro Abate, Michael A Osborne

Keywords: inverse reinforcement learning, IRL, imitation learning

TL;DR: A new method for Bayesian inverse reinforcement learning based on variational inference over Q-values.

Abstract: The development of safe and beneficial AI requires that systems can learn and act in accordance with human preferences. However, explicitly specifying these preferences by hand is often infeasible. Inverse reinforcement learning (IRL) addresses this challenge by inferring preferences, represented as reward functions, from expert behavior. We introduce Q-based Variational IRL (QVIRL), a novel Bayesian IRL method that recovers a posterior distribution over rewards from expert demonstrations via primarily learning a variational distribution over Q-values. Unlike previous approaches, QVIRL combines scalability with uncertainty quantification, important for safety-critical applications. We demonstrate QVIRL's strong performance in apprenticeship learning across various tasks, including classical control problems and safe navigation in the Safety Gymnasium suite, where the method's uncertainty quantification allows us to produce safer policies.

Primary Area: probabilistic methods (Bayesian methods, variational inference, sampling, UQ, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 9740