Saved in:
Bibliographic Details
Main Authors: Zhang, Haozhuo, Sun, Jingkai, Caprio, Michele, Tang, Jian, Zhang, Shanghang, Zhang, Qiang, Pan, Wei
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.16943
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908867595599872
author Zhang, Haozhuo
Sun, Jingkai
Caprio, Michele
Tang, Jian
Zhang, Shanghang
Zhang, Qiang
Pan, Wei
author_facet Zhang, Haozhuo
Sun, Jingkai
Caprio, Michele
Tang, Jian
Zhang, Shanghang
Zhang, Qiang
Pan, Wei
contents We introduce LHM-Humanoid, a benchmark and learning framework for long-horizon whole-body humanoid loco-manipulation in diverse, cluttered scenes. In our setting, multiple objects are displaced from their intended locations and may obstruct navigation; a humanoid agent must repeatedly (i) walk to a target, (ii) pick it up with diverse whole-body postures under balance constraints, (iii) carry it while navigating around obstacles, and (iv) place it at a designated goal -- all within a single continuous episode and without any environment reset. This task simultaneously demands cross-scene generalization and unified one-policy control: layouts, obstacle arrangements, object category/mass/shape/color and object start/goal poses vary substantially even within a room category, requiring a single general policy that directly outputs actions rather than invoking pre-trained skill libraries. Our dataset spans four room types (bedroom, living room, kitchen, and warehouse), comprising 350 diverse scenes/tasks with 79 objects (25 movable targets). Since no scene-specific ground-truth motion sequences are provided, we learn goal-conditioned teacher policies via reinforcement learning and distill them into a single end-to-end student policy using DAgger. We further distill this unified policy into a vision-language-action (VLA) model driven by egocentric RGB observations and natural language. Experiments in Isaac Gym demonstrate that LHM-Humanoid substantially outperforms end-to-end RL baselines and prior humanoid loco-manipulation methods on both seen and unseen scenes, exhibiting strong long-horizon robustness and cross-scene generalization.
format Preprint
id arxiv_https___arxiv_org_abs_2508_16943
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LHM-Humanoid: Learning a Unified Policy for Long-Horizon Humanoid Whole-Body Loco-Manipulation in Diverse Messy Environments
Zhang, Haozhuo
Sun, Jingkai
Caprio, Michele
Tang, Jian
Zhang, Shanghang
Zhang, Qiang
Pan, Wei
Robotics
Artificial Intelligence
We introduce LHM-Humanoid, a benchmark and learning framework for long-horizon whole-body humanoid loco-manipulation in diverse, cluttered scenes. In our setting, multiple objects are displaced from their intended locations and may obstruct navigation; a humanoid agent must repeatedly (i) walk to a target, (ii) pick it up with diverse whole-body postures under balance constraints, (iii) carry it while navigating around obstacles, and (iv) place it at a designated goal -- all within a single continuous episode and without any environment reset. This task simultaneously demands cross-scene generalization and unified one-policy control: layouts, obstacle arrangements, object category/mass/shape/color and object start/goal poses vary substantially even within a room category, requiring a single general policy that directly outputs actions rather than invoking pre-trained skill libraries. Our dataset spans four room types (bedroom, living room, kitchen, and warehouse), comprising 350 diverse scenes/tasks with 79 objects (25 movable targets). Since no scene-specific ground-truth motion sequences are provided, we learn goal-conditioned teacher policies via reinforcement learning and distill them into a single end-to-end student policy using DAgger. We further distill this unified policy into a vision-language-action (VLA) model driven by egocentric RGB observations and natural language. Experiments in Isaac Gym demonstrate that LHM-Humanoid substantially outperforms end-to-end RL baselines and prior humanoid loco-manipulation methods on both seen and unseen scenes, exhibiting strong long-horizon robustness and cross-scene generalization.
title LHM-Humanoid: Learning a Unified Policy for Long-Horizon Humanoid Whole-Body Loco-Manipulation in Diverse Messy Environments
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2508.16943