Q-based Variational Inverse Reinforcement Learning

2026-08-17

AI Summary

This study addresses the challenge of simultaneously achieving scalability and uncertainty quantification in inverse reinforcement learning by proposing QVIRL. By learning a variational distribution over optimal Q-values to recover the reward posterior, QVIRL constitutes the first Bayesian inverse reinforcement learning framework that supports raw pixel inputs while providing uncertainty quantification. The method effectively integrates variational inference with active learning, demonstrating superior performance across multiple benchmark tasks and ATARI games. Consequently, QVIRL successfully enables high-dimensional pixel-level training and efficient sample acquisition, significantly enhancing both the scalability and practical applicability of the algorithm.

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 behaviour. 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 optimal Q-values. Unlike previous approaches, QVIRL combines scalability with uncertainty quantification, important for safety-critical applications as well as active learning. We demonstrate QVIRL's strong performance in apprenticeship learning across various tasks, including gridworlds, Lunar Lander, the Highway Environment, and two ATARI games both with static expert data and with active learning. It is the first method for Bayesian IRL that demonstrates training from raw pixel observations.

Problem

• Inverse Reinforcement Learning
• Uncertainty Quantification
• Scalability
• Human Preferences

Innovation

• Q-based Variational IRL
• Bayesian Inverse Reinforcement Learning
• Uncertainty Quantification
• Raw Pixel Observations
• Active Learning