Adaptive Horizon Actor-Critic for Policy Learning in Contact-Rich Differentiable Simulation

TL;DR

This paper introduces Adaptive Horizon Actor-Critic, a novel model-based reinforcement learning algorithm that dynamically adjusts the simulation horizon to mitigate stiff dynamics errors, leading to improved policy learning in contact-rich environments.

Contribution

The paper proposes AHAC, an adaptive horizon method for FO-MBRL that reduces gradient errors caused by stiff dynamics, enhancing performance and scalability.

Findings

AHAC outperforms MFRL baselines by 40% in reward.

It scales efficiently to high-dimensional control tasks.

Demonstrates improved wall-clock-time efficiency.

Abstract

Model-Free Reinforcement Learning (MFRL), leveraging the policy gradient theorem, has demonstrated considerable success in continuous control tasks. However, these approaches are plagued by high gradient variance due to zeroth-order gradient estimation, resulting in suboptimal policies. Conversely, First-Order Model-Based Reinforcement Learning (FO-MBRL) methods employing differentiable simulation provide gradients with reduced variance but are susceptible to sampling error in scenarios involving stiff dynamics, such as physical contact. This paper investigates the source of this error and introduces Adaptive Horizon Actor-Critic (AHAC), an FO-MBRL algorithm that reduces gradient error by adapting the model-based horizon to avoid stiff dynamics. Empirical findings reveal that AHAC outperforms MFRL baselines, attaining 40% more reward across a set of locomotion tasks and efficiently scaling to high-dimensional control environments with improved wall-clock-time efficiency.