Estimation Error Correction in Deep Reinforcement Learning for Deterministic Actor-Critic Methods

TL;DR

This paper introduces a novel deep Q-learning method that reduces estimation bias in deep reinforcement learning, especially under high-variance signals, leading to improved performance in continuous control tasks.

Contribution

A parameter-free deep Q-learning variant combining Clipped Double Q-learning and Maxmin Q-learning to better estimate value functions in actor-critic methods.

Findings

Improved performance on OpenAI Gym continuous control tasks

Reduces overestimation and underestimation biases

Achieves state-of-the-art results in tested environments

Abstract

In value-based deep reinforcement learning methods, approximation of value functions induces overestimation bias and leads to suboptimal policies. We show that in deep actor-critic methods that aim to overcome the overestimation bias, if the reinforcement signals received by the agent have a high variance, a significant underestimation bias arises. To minimize the underestimation, we introduce a parameter-free, novel deep Q-learning variant. Our Q-value update rule combines the notions behind Clipped Double Q-learning and Maxmin Q-learning by computing the critic objective through the nested combination of maximum and minimum operators to bound the approximate value estimates. We evaluate our modification on the suite of several OpenAI Gym continuous control tasks, improving the state-of-the-art in every environment tested.