Long-Distance Real-World Navigation of the Legged-Wheeled Robot Go2-W Using Deep Reinforcement Learning

2026-06-19

AI Summary

This work addresses the challenge of long-range autonomous navigation for legged-wheeled robots in complex real-world environments by proposing an integrated approach that combines proprioceptive locomotion control with autonomous navigation based on deep reinforcement learning. The method extends a purely proprioception-based policy to the 16-degree-of-freedom commercial robot Go2-W and incorporates a load distribution mechanism to mitigate hip joint overheating, substantially enhancing the system’s endurance. Validated in the Tsukuba Challenge 2025, the system successfully completed a 2.8-kilometer uninterrupted autonomous run across diverse terrains—including sidewalks, parks, and staircases—demonstrating robust reliability, durability, and generalization capability in real-world conditions.

Abstract

Legged-wheeled robots have long been studied for their potential to combine the efficient flat-ground mobility of wheels with the rough-terrain capability of legs. However, examples of their application to long-range autonomous navigation in real environments remain limited. This paper reports our effort to build a deep reinforcement learning (DRL) based locomotion controller and an autonomous navigation system for the commercially available legged-wheeled robot Go2-W, and to apply them to long-range autonomous navigation in a real environment. For locomotion control, we extended a proprioception-only policy, which we had previously developed for quadruped robots, to the 16-DoF legged-wheeled robot. We also found that wheeled locomotion concentrates the load on the hip joints and causes heat concentration that hinders sustained travel, and obtained a policy that suppresses it by distributing the load. We evaluated the system at the Tsukuba Challenge 2025, demonstrating that it can autonomously traverse an approximately 2.8 km route including sidewalks, a park, and stairs without stopping due to overheating.

Problem

• legged-wheeled robot
• long-distance navigation
• autonomous navigation
• real-world environment
• joint overheating

Innovation

• legged-wheeled robot
• deep reinforcement learning
• proprioception-only policy
• load distribution
• long-distance autonomous navigation