PulseAugur
EN
LIVE 21:16:03

New AI framework enables humanoid robots to adapt locomotion to varied terrains

Researchers have developed CoRe-MoE, a novel two-stage reinforcement learning framework designed to enhance humanoid robot locomotion across diverse terrains. This approach decouples gait generation from terrain adaptation, first establishing a stable walking and running policy. Subsequently, a contrastive objective is used to train a specialized Mixture-of-Experts branch for terrain awareness, promoting expert specialization and improving adaptability. The framework has demonstrated superior performance in simulations and successful zero-shot deployment on a Unitree G1 robot, enabling robust navigation over various challenging environments. AI

IMPACT Enhances humanoid robot capabilities for complex navigation tasks, potentially leading to more versatile robotic applications.

RANK_REASON This is a research paper detailing a new AI framework for robotics.

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

COVERAGE [2]

  1. arXiv cs.AI TIER_1 English(EN) · Kailun Huang (Hong Kong University of Science and Technology), Zikang Xie (Hong Kong University of Science and Technology), Yanzhe Xie (Hong Kong University of Science and Technology), Panpan Liao (Guangdong University of Technology), Fanghai Zhang (Hong… ·

    CoRe-MoE: Contrastive Reweighted Mixture of Experts for Multi-Terrain Humanoid Locomotion with Gait Adaptation

    arXiv:2606.04718v1 Announce Type: cross Abstract: Humans primarily rely on walking and running to traverse complex terrains, without resorting to unnecessarily complex motion patterns. Similarly, humanoid robots should achieve smooth transitions between walking and running while …

  2. arXiv cs.AI TIER_1 English(EN) · Haohui Huang ·

    CoRe-MoE: Contrastive Reweighted Mixture of Experts for Multi-Terrain Humanoid Locomotion with Gait Adaptation

    Humans primarily rely on walking and running to traverse complex terrains, without resorting to unnecessarily complex motion patterns. Similarly, humanoid robots should achieve smooth transitions between walking and running while maintaining natural and stable locomotion. However…