Pessimistic Bootstrapping for Uncertainty-Driven Offline Reinforcement Learning

TL;DR

This paper introduces PBRL, a novel offline RL algorithm that uses uncertainty quantification and pessimistic updates based on bootstrapped Q-function disagreement, improving performance without explicit policy constraints.

Contribution

The paper proposes a purely uncertainty-driven offline RL method using bootstrapped Q-function disagreement for pessimistic updates, with a new OOD sampling technique and theoretical guarantees.

Findings

PBRL outperforms state-of-the-art algorithms on D4RL benchmarks.

The method provides provable uncertainty quantification in linear MDPs.

PBRL avoids explicit policy constraints, enabling better generalization.

Abstract

Offline Reinforcement Learning (RL) aims to learn policies from previously collected datasets without exploring the environment. Directly applying off-policy algorithms to offline RL usually fails due to the extrapolation error caused by the out-of-distribution (OOD) actions. Previous methods tackle such problem by penalizing the Q-values of OOD actions or constraining the trained policy to be close to the behavior policy. Nevertheless, such methods typically prevent the generalization of value functions beyond the offline data and also lack precise characterization of OOD data. In this paper, we propose Pessimistic Bootstrapping for offline RL (PBRL), a purely uncertainty-driven offline algorithm without explicit policy constraints. Specifically, PBRL conducts uncertainty quantification via the disagreement of bootstrapped Q-functions, and performs pessimistic updates by penalizing the value function based on the estimated uncertainty. To tackle the extrapolating error, we further propose a novel OOD sampling method. We show that such OOD sampling and pessimistic bootstrapping yields provable uncertainty quantifier in linear MDPs, thus providing the theoretical underpinning for PBRL. Extensive experiments on D4RL benchmark show that PBRL has better performance compared to the state-of-the-art algorithms.