Distributional Actor-Critic Ensemble for Uncertainty-Aware Continuous Control

TL;DR

This paper introduces an uncertainty-aware reinforcement learning algorithm that extends DDPG by disentangling epistemic and aleatoric uncertainties, leading to improved performance in continuous control tasks.

Contribution

It proposes a novel distributional actor-critic ensemble method that separately estimates epistemic and aleatoric uncertainties within DDPG for better exploration and risk-sensitive policies.

Findings

Outperforms vanilla DDPG in robotic control benchmarks

Enhances exploration efficiency through epistemic uncertainty

Learns risk-sensitive policies using aleatoric uncertainty

Abstract

Uncertainty quantification is one of the central challenges for machine learning in real-world applications. In reinforcement learning, an agent confronts two kinds of uncertainty, called epistemic uncertainty and aleatoric uncertainty. Disentangling and evaluating these uncertainties simultaneously stands a chance of improving the agent's final performance, accelerating training, and facilitating quality assurance after deployment. In this work, we propose an uncertainty-aware reinforcement learning algorithm for continuous control tasks that extends the Deep Deterministic Policy Gradient algorithm (DDPG). It exploits epistemic uncertainty to accelerate exploration and aleatoric uncertainty to learn a risk-sensitive policy. We conduct numerical experiments showing that our variant of DDPG outperforms vanilla DDPG without uncertainty estimation in benchmark tasks on robotic control and power-grid optimization.