Stabilizing Off-Policy Q-Learning via Bootstrapping Error Reduction
This addresses the challenge of sample-efficient learning from prior experience for reinforcement learning practitioners, offering a more robust method but is incremental as it builds on existing off-policy techniques.
The paper tackles the problem of instability in off-policy reinforcement learning due to bootstrapping error, proposing the BEAR algorithm that reduces this error and learns robustly from fixed off-policy data, achieving improved performance on continuous control tasks.
Off-policy reinforcement learning aims to leverage experience collected from prior policies for sample-efficient learning. However, in practice, commonly used off-policy approximate dynamic programming methods based on Q-learning and actor-critic methods are highly sensitive to the data distribution, and can make only limited progress without collecting additional on-policy data. As a step towards more robust off-policy algorithms, we study the setting where the off-policy experience is fixed and there is no further interaction with the environment. We identify bootstrapping error as a key source of instability in current methods. Bootstrapping error is due to bootstrapping from actions that lie outside of the training data distribution, and it accumulates via the Bellman backup operator. We theoretically analyze bootstrapping error, and demonstrate how carefully constraining action selection in the backup can mitigate it. Based on our analysis, we propose a practical algorithm, bootstrapping error accumulation reduction (BEAR). We demonstrate that BEAR is able to learn robustly from different off-policy distributions, including random and suboptimal demonstrations, on a range of continuous control tasks.