← Back to all news
arXiv AI By Shengzhang Xu, Yuning Guo, Weiyi Hu, Yifeng Xu, Zhongyu Li, He Kong

REACH: An RL-based Hierarchical Humanoid Crawling Framework for Adaptable Confined Space Traversal

Read the original on arXiv AI →

REACH is a reinforcement‑learning based hierarchical framework that allows a Unitree G1 humanoid robot to navigate confined spaces by switching between variable‑height bipedal walking and crawling. The system learns policies that adjust body height from 1.27 m down to 0.20 m, uses a history‑conditioned latent representation to handle contact‑rich crawling dynamics, and incorporates a bidirectional transition policy to connect the two locomotion modes. A LiDAR‑based planner generates collision‑free 3D paths and proactive height targets, and real‑world tests confirm accurate height and velocity tracking, robust terrain traversal, and reliable mode transitions.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv AI
Jun 17

OmniRetarget: Interaction-Preserving Data Generation for Humanoid Whole-Body Loco-Manipulation and Scene Interaction

arXiv:2509. 26633v3 Announce Type: replace-cross Abstract: A dominant paradigm for teaching humanoid robots complex skills is to retarget human motions as kinematic references to train reinforcement learning (RL) policies.

By Lujie Yang, Xiaoyu Huang, Zhen Wu, Angjoo Kanazawa, Pieter Abbeel, Carmelo Sferrazza, C. Karen Liu, Rocky Duan, Guanya Shi
agentsreinforcement-learningrobotics
More like this →
arXiv AI
Sep 17

Language-Guided Terrain-Adaptive Neural MPC for Autonomous Traversal of Articulated Tracked Robots

The paper introduces ASTRIL-MPC, a language‑guided neural model predictive control framework that enables articulated tracked robots to navigate complex, contact‑rich urban environments such as stairwells and cluttered interiors. By combining a learned kinematics model that predicts short‑horizon state changes, an optimization‑based planner with multi‑objective costs, and a large language model that safely updates control weights, the system achieves up to 71% better traversal quality than non‑adaptive NMPC and 67% better than a PPO baseline, while eliminating collision impacts during descent. Real‑robot trials over four indoor obstacles confirm the method’s transferability to physical contact‑rich traversal.

By Zhenfeng Gan, Yanbo Chen, Lirong Che, Yongyi Ma, Rongkai Zhu, Xueqian Wang
llmsagentsreinforcement-learningroboticssafety
More like this →