Acrobotics: A Generalist Approahc To Quadrupedal Robots' Parkour

2025-09-02

AI Summary

Quadrupedal robots face significant challenges—including slippage, instability, and difficulty in coordinating multiple locomotion tasks—when performing dynamic obstacle traversal (e.g., climbing, gap crossing, stair ascent) over unstructured terrain. Conventional model-based controllers suffer from poor generalizability, while expert-guided reinforcement learning (RL) approaches exhibit low sample efficiency. Method: We propose a general-purpose deep RL framework that jointly integrates dynamic motion modeling with environment interaction optimization, enabling shared-policy training across diverse locomotion tasks. Crucially, the framework requires no handcrafted action templates or pre-defined expert policies. Contribution/Results: Our method achieves performance on par with an expert-mixture model using only 25% of the agent count. Experiments demonstrate high success rates and strong robustness across complex, multi-modal obstacle scenarios, significantly improving both training efficiency and cross-task generalization capability.

Technology Category

• Intelligent Robots: Learning & Optimization for ROB
• Search and Optimization: Learning to Search
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Climbing, crouching, bridging gaps, and walking up stairs are just a few of the advantages that quadruped robots have over wheeled robots, making them more suitable for navigating rough and unstructured terrain. However, executing such manoeuvres requires precise temporal coordination and complex agent-environment interactions. Moreover, legged locomotion is inherently more prone to slippage and tripping, and the classical approach of modeling such cases to design a robust controller thus quickly becomes impractical. In contrast, reinforcement learning offers a compelling solution by enabling optimal control through trial and error. We present a generalist reinforcement learning algorithm for quadrupedal agents in dynamic motion scenarios. The learned policy rivals state-of-the-art specialist policies trained using a mixture of experts approach, while using only 25% as many agents during training. Our experiments also highlight the key components of the generalist locomotion policy and the primary factors contributing to its success.

Problem

• Developing generalist reinforcement learning for quadrupedal parkour locomotion
• Addressing precise coordination challenges in complex agent-environment interactions
• Overcoming impractical classical modeling of slippage and tripping in legged robots

Innovation

• Generalist reinforcement learning algorithm
• Dynamic motion scenarios for quadrupedal agents
• Rivals specialist policies with fewer agents