Constraint-Aware Diffusion Priors for High-Fidelity and Versatile Quadruped Locomotion

TL;DR

This paper introduces Diff-CAST, a diffusion model-based motion prior framework that enhances quadruped locomotion by improving data scalability, reducing mode collapse, and ensuring hardware safety during complex maneuvers.

Contribution

The paper presents Diff-CAST, a novel diffusion-guided motion prior that replaces GAN discriminators, and a comprehensive Sim2Real architecture with SACC and Constrained RL for safe, high-fidelity quadruped locomotion.

Findings

Diff-CAST mitigates mode collapse in motion modeling.

Enables seamless skill transitions in quadruped locomotion.

Ensures robust, hardware-safe locomotion during complex maneuvers.

Abstract

Reinforcement learning combined with imitation learning has significantly advanced biomimetic quadrupedal locomotion. However, scaling these frameworks to massive, multi-source datasets exposes fundamental bottlenecks. First, traditional GAN-based discriminators are prone to mode collapse, struggling to capture diverse motion distributions from uncurated datasets. Second, existing kinematic priors suffer from out-of-distribution (OOD) tracking conflicts, leading to severe unintended heading drifts during complex maneuvers. Furthermore, deploying unconstrained priors to physical hardware poses critical safety risks by disregarding actuator dynamics. To overcome these challenges, we propose Diff-CAST (Diffusion-guided Constraint-Aware Symmetric Tracking), a novel motion prior framework leveraging the multi-modal distribution modeling capabilities of diffusion models for stylistic rewards. Diff-CAST effectively replaces traditional GAN discriminators, unlocking robust data scaling on heterogeneous collections. To ensure high-fidelity intent execution and reliable real-world deployment, we introduce a comprehensive Sim2Real architecture integrating Symmetric Augmented Command Conditioning (SACC) for drift-free tracking, and Constrained RL for hardware safety. Experiments on a quadruped demonstrate that Diff-CAST mitigates mode collapse, enables seamless transitions between diverse skills, and ensures robust, hardware-compliant locomotion.