Soft Decomposed Policy-Critic: Bridging the Gap for Effective Continuous Control with Discrete RL

TL;DR

This paper introduces the SDPC architecture that combines discrete RL techniques with soft actor-critic methods to effectively handle continuous control tasks, overcoming the dimensional explosion problem.

Contribution

The paper proposes a novel SDPC framework that discretizes action spaces independently and integrates soft RL with actor-critic methods for continuous control.

Findings

SDPC outperforms state-of-the-art continuous RL algorithms in Mujoco and BipedalWalker tasks.

The approach effectively addresses the dimensional explosion in continuous control.

Empirical validation shows superior performance of SDPC in various benchmarks.

Abstract

Discrete reinforcement learning (RL) algorithms have demonstrated exceptional performance in solving sequential decision tasks with discrete action spaces, such as Atari games. However, their effectiveness is hindered when applied to continuous control problems due to the challenge of dimensional explosion. In this paper, we present the Soft Decomposed Policy-Critic (SDPC) architecture, which combines soft RL and actor-critic techniques with discrete RL methods to overcome this limitation. SDPC discretizes each action dimension independently and employs a shared critic network to maximize the soft $Q$-function. This novel approach enables SDPC to support two types of policies: decomposed actors that lead to the Soft Decomposed Actor-Critic (SDAC) algorithm, and decomposed $Q$-networks that generate Boltzmann soft exploration policies, resulting in the Soft Decomposed-Critic Q (SDCQ) algorithm. Through extensive experiments, we demonstrate that our proposed approach outperforms state-of-the-art continuous RL algorithms in a variety of continuous control tasks, including Mujoco's Humanoid and Box2d's BipedalWalker. These empirical results validate the effectiveness of the SDPC architecture in addressing the challenges associated with continuous control.