MUJICA: Multi-skill Unified Joint Integration of Control Architecture for Wheeled-Legged Robots

2026-05-13

AI Summary

This work addresses the challenges of robustness and adaptability faced by wheeled-legged robots operating in complex terrains under proprioceptive noise and motor constraints. The authors propose a unified reinforcement learning framework that relies solely on proprioceptive inputs and, for the first time, jointly trains multiple low-level skills—including omnidirectional locomotion, platform climbing, and fall recovery—within a single policy. A high-level skill selector enables seamless, environment-adaptive transitions among these behaviors. By accurately modeling DC motor constraints and employing indicator variables to distinguish skills, the approach achieves efficient sim-to-real transfer on the Unitree Go2-W platform, significantly improving task success rates and environmental adaptability in unstructured settings.

Abstract

Wheeled-legged robots hold promise for traversing complex terrains and offer superior mobility compared to legged robots. However, wheeled-legged robots must effectively balance both wheeled driving and legged control. Furthermore, due to noisy proprioceptive sensing and real-world motor constraints, realizing robust and adaptive locomotion at peak performance of motors remains challenging. We propose the Multi-skill Unified Joint Integration of Control Architecture (MUJICA), a unified, fully proprioceptive control framework for wheeled-legged robots that integrates diverse low-level skills-including omnidirectional moving, high platform climbing, and fall recovery-within a single policy. All skills, distinguished by unique indicator variables, are trained jointly with accurate DC-motor constraint modeling. Additionally, a high-level skill selector is learned to dynamically choose the optimal skill based solely on proprioceptions, enabling adaptive responses to the surrounding environment. Therefore, MUJICA enhances sim-to-real robustness and enables seamless transitions across diverse locomotion modes, facilitating autonomous adjustment to the environment. We validate our framework in both simulation and real-world experiments on the Unitree Go2-W robot, demonstrating significant improvements in adaptability and task success in unstructured environments.

Problem

• wheeled-legged robots
• locomotion control
• proprioceptive sensing
• motor constraints
• adaptive locomotion

Innovation

• wheeled-legged robots
• unified control architecture
• proprioceptive sensing
• motor constraint modeling
• skill selection