Langevin Soft Actor-Critic: Efficient Exploration through Uncertainty-Driven Critic Learning

TL;DR

Langevin Soft Actor Critic (LSAC) introduces an uncertainty-driven exploration method for continuous control RL, leveraging Langevin Monte Carlo and Thompson sampling to improve sample efficiency and outperform existing algorithms.

Contribution

The paper presents LSAC, a novel RL algorithm that integrates Langevin Monte Carlo-based Thompson sampling for critic learning, enabling efficient exploration in continuous action spaces.

Findings

LSAC outperforms mainstream RL algorithms in continuous control tasks.

First successful application of LMC-based Thompson sampling in continuous control RL.

Demonstrates improved sample efficiency through uncertainty-driven exploration.

Abstract

Existing actor-critic algorithms, which are popular for continuous control reinforcement learning (RL) tasks, suffer from poor sample efficiency due to lack of principled exploration mechanism within them. Motivated by the success of Thompson sampling for efficient exploration in RL, we propose a novel model-free RL algorithm, Langevin Soft Actor Critic (LSAC), which prioritizes enhancing critic learning through uncertainty estimation over policy optimization. LSAC employs three key innovations: approximate Thompson sampling through distributional Langevin Monte Carlo (LMC) based $Q$ updates, parallel tempering for exploring multiple modes of the posterior of the $Q$ function, and diffusion synthesized state-action samples regularized with $Q$ action gradients. Our extensive experiments demonstrate that LSAC outperforms or matches the performance of mainstream model-free RL algorithms for continuous control tasks. Notably, LSAC marks the first successful application of an LMC based Thompson sampling in continuous control tasks with continuous action spaces.