Thousands of parallel GPU environments train a quadruped walking policy in minutes; legged_gym became a standard starter.数千个 GPU 并行环境几分钟就能训练出四足行走策略；legged_gym 成为标准入门框架。
@misc{rudin2021learning,
title = {{Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning}},
author = {Nikita Rudin and David Hoeller and Philipp Reist and Marco Hutter},
year = {2021},
eprint = {2109.11978},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2109.11978},
}
HALO trains humanoid skills on vision-language-motion data synthesised in a thousand procedural homes, so data scales with compute. Reports zero-shot transfer: 91% over 360 trials in six real environments.HALO 在上千个程序化生成的家庭场景中合成视觉-语言-动作数据来训练人形技能，数据量随算力增长。报告零微调迁移：六个未建图真实环境的 360 次试验成功率 91%。
@misc{hu2026perceptive,
title = {{Perceptive Humanoid Motion from Simulation}},
author = {Siyuan Hu and Abulikemu Abuduweili and Yue Meng and Shipeng Liu and Quentin Leboutet and Chen Chen and Matthias Müller and Vladlen Koltun},
year = {2026},
eprint = {2611.06908},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06908},
}
One motion-tracking recipe with shared hyperparameters learns agile skills (cartwheels, flip-kicks); a guided latent diffusion model composes them for new goals like joystick control, zero-shot on hardware.用同一套超参数的运动跟踪方案学会侧手翻、飞踢、冲刺等敏捷技能，再用带引导的潜空间扩散模型组合这些技能以完成摇杆遥控、避障等新目标，零样本迁移到真机。
@misc{liao2025beyondmimic,
title = {{BeyondMimic: From Motion Tracking to Versatile Humanoid Control via Guided Diffusion}},
author = {Qiayuan Liao and Takara E. Truong and Xiaoyu Huang and Yuman Gao and Guy Tevet and Koushil Sreenath and C. Karen Liu},
year = {2025},
eprint = {2508.08241},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2508.08241},
}
Tests scaling on humanoid control: more network, data and compute yield a generalist motion-tracking foundation model, with no manual reward engineering.检验规模定律在人形控制上的效果：扩大网络、数据和算力，得到通用动作跟踪基础模型，无需手工设计奖励。
@misc{luo2025sonic,
title = {{SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control}},
author = {Zhengyi Luo and Ye Yuan and Tingwu Wang and Chenran Li and Fernando Castañeda and Sirui Chen and Zi-Ang Cao and Jiefeng Li and David Minor and Qingwei Ben and Jinhyung Park and David Sami and Zi Wang and Xingye Da and Runyu Ding and Cyrus Hogg and Lina Song and Edy Lim and Eugene Jeong and Tairan He and Haoru Xue and Wenli Xiao and Simon Yuen and Jan Kautz and Yan Chang and Umar Iqbal and Linxi "Jim" Fan and Yuke Zhu},
year = {2025},
eprint = {2511.07820},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.07820},
}
iDP3: whole-upper-body teleop, a 25-DoF humanoid with 3D LiDAR, and an improved 3D Diffusion Policy for noisy human data, aiming at skills beyond a single scene.iDP3：上半身全身遥操作、带三维激光雷达的 25 自由度人形，加上改进的三维扩散策略，使技能不再局限于单一场景。
@misc{ze2024generalizable,
title = {{Generalizable Humanoid Manipulation with 3D Diffusion Policies}},
author = {Yanjie Ze and Zixuan Chen and Wenhao Wang and Tianyi Chen and Xialin He and Ying Yuan and Xue Bin Peng and Jiajun Wu},
year = {2024},
eprint = {2410.10803},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.10803},
}
@misc{radosavovic2023real,
title = {{Real-World Humanoid Locomotion with Reinforcement Learning}},
author = {Ilija Radosavovic and Tete Xiao and Bike Zhang and Trevor Darrell and Jitendra Malik and Koushil Sreenath},
year = {2023},
eprint = {2303.03381},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2303.03381},
}
Showed RL can imitate a wide range of motion clips, flips and spins included, while staying robust. The motion-imitation template much humanoid work builds on.证明强化学习能模仿包括空翻在内的大量动作片段并保持鲁棒，是后续人形动作模仿工作的基本范式。
@misc{peng2018deepmimic,
title = {{DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills}},
author = {Xue Bin Peng and Pieter Abbeel and Sergey Levine and Michiel van de Panne},
year = {2018},
eprint = {1804.02717},
archivePrefix = {arXiv},
primaryClass = {cs.GR},
url = {https://arxiv.org/abs/1804.02717},
}
Commercially licensed 3B-parameter humanoid VLA: a Cosmos-Reason2 backbone plans and a diffusion transformer outputs motor commands. NVIDIA reports training on about 20,854 hours of human egocentric video.可商用的 30 亿参数人形 VLA：Cosmos-Reason2 主干负责推理规划，扩散 Transformer 输出电机指令。NVIDIA 称用约 20,854 小时人类第一视角视频训练。
Rather than distilling per-motion experts or brute-forcing one RL policy, trains a humanoid tracking policy jointly with one differentiable world model of shared dynamics, fit from simulator logs (CoRL 2026 oral).SuperMimic 不再蒸馏逐动作专家或用大算力硬训单一 RL 策略，而是把人形机器人动作跟踪策略与一个从仿真日志拟合的、共享动力学的可微世界模型联合训练（CoRL 2026 口头报告）。
@misc{liao2026supermimic,
title = {{SuperMimic: Scaling Humanoid Whole-Body Control with One Differentiable World Model}},
author = {Qiayuan Liao and Yiyang Shao and Junfeng Long and Kevin Zakka and Roberto Horowitz and Pieter Abbeel and Koushil Sreenath},
year = {2026},
eprint = {2610.28480},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.28480},
}
CoRL 2026 Spotlight. Hardens whole-body end-effector tracking for legged manipulators against out-of-distribution inputs such as sensor noise or infeasible commands, using a frame-wise single-step safety manifold.CoRL 2026 Spotlight。提升足式移动操作臂全身末端跟踪对分布外输入（传感器噪声、不可行指令）的鲁棒性：用逐帧安全方案把无限时域安全约束转化为计算高效的单步流形约束。
@misc{cheng2026cmp,
title = {{CMP: Robust Whole-Body Tracking for Loco-Manipulation via Competence Manifold Projection}},
author = {Ziyang Cheng and Haoyu Wei and Hang Yin and Xiuwei Xu and Bingyao Yu and Jie Zhou and Jiwen Lu},
year = {2026},
eprint = {2604.07457},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.07457},
}
Learns a residual "delta action" model from real rollouts to fix sim physics, enabling agile athletic humanoid motions.从真实轨迹学习残差"增量动作"模型来修正仿真物理，实现敏捷的人形运动技能。
@misc{he2025asap,
title = {{ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills}},
author = {Tairan He and Jiawei Gao and Wenli Xiao and Yuanhang Zhang and Zi Wang and Jiashun Wang and Zhengyi Luo and Guanqi He and Nikhil Sobanbab and Chaoyi Pan and Zeji Yi and Guannan Qu and Kris Kitani and Jessica Hodgins and Linxi "Jim" Fan and Yuke Zhu and Changliu Liu and Guanya Shi},
year = {2025},
eprint = {2502.01143},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.01143},
}
Shows retargeting artifacts such as foot sliding hurt tracking policies, and offers GMR, a general retargeter. Its open-source repo is widely used for humanoid data.说明脚滑等重定向瑕疵会损害跟踪策略，并提出通用重定向方法 GMR；其开源仓库在人形数据处理中使用广泛。
@misc{araujo2025retargeting,
title = {{Retargeting Matters: General Motion Retargeting for Humanoid Motion Tracking}},
author = {Joao Pedro Araujo and Yanjie Ze and Pei Xu and Jiajun Wu and C. Karen Liu},
year = {2025},
eprint = {2510.02252},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.02252},
}
Retargeting that preserves human-object and human-scene interaction rather than only body pose, generating physically plausible data for humanoid loco-manipulation.不仅保持身体姿态，还保留人与物体、人与场景交互的重定向方法，为人形移动操作生成物理上合理的数据。
@misc{yang2025omniretarget,
title = {{OmniRetarget: Interaction-Preserving Data Generation for Humanoid Whole-Body Loco-Manipulation and Scene Interaction}},
author = {Lujie Yang and Xiaoyu Huang and Zhen Wu and Angjoo Kanazawa and Pieter Abbeel and Carmelo Sferrazza and C. Karen Liu and Rocky Duan and Guanya Shi},
year = {2025},
eprint = {2509.26633},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.26633},
}
Learns highly dynamic skills such as kung fu and dancing, through multi-step motion processing and adaptive motion tracking.通过多步动作处理和自适应动作跟踪，学习功夫、舞蹈等高动态技能。
@misc{xie2025kungfubot,
title = {{KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills}},
author = {Weiji Xie and Jinrui Han and Jiakun Zheng and Huanyu Li and Xinzhe Liu and Jiyuan Shi and Weinan Zhang and Chenjia Bai and Xuelong Li},
year = {2025},
eprint = {2506.12851},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.12851},
}
Open-source humanoid with 3D-printed modular gearboxes, buildable for under US$5,000; hardware designs, embedded code and training code are released.开源人形机器人，采用 3D 打印的模块化减速器，在美国可用不到 5000 美元搭建；硬件设计、嵌入式代码与训练框架均已开源。
@misc{chi2025demonstrating,
title = {{Demonstrating Berkeley Humanoid Lite: An Open-source, Accessible, and Customizable 3D-printed Humanoid Robot}},
author = {Yufeng Chi and Qiayuan Liao and Junfeng Long and Xiaoyu Huang and Sophia Shao and Borivoje Nikolic and Zhongyu Li and Koushil Sreenath},
year = {2025},
eprint = {2504.17249},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2504.17249},
}
Unitree's official starting point for RL locomotion on Go2, H1 and G1 — train in Isaac Gym, check sim-to-sim in MuJoCo, then deploy to the real robot.宇树官方的强化学习运动控制入门仓库，支持 Go2、H1、G1：在 Isaac Gym 中训练，在 MuJoCo 中做 sim2sim 验证，再部署到真机。
Isaac Gym–based RL framework for humanoid locomotion with a sim-to-sim check in MuJoCo before deploying zero-shot to real humanoids; a common starting point for humanoid RL.基于 Isaac Gym 的人形机器人运动强化学习框架，先在 MuJoCo 中做 sim2sim 验证，再零样本部署到真实人形机器人，是人形 RL 的常用起点。
@misc{cheng2024open,
title = {{Open-TeleVision: Teleoperation with Immersive Active Visual Feedback}},
author = {Xuxin Cheng and Jialong Li and Shiqi Yang and Ge Yang and Xiaolong Wang},
year = {2024},
eprint = {2407.01512},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2407.01512},
}
Distills specialist parkour skills into a single vision-based policy that climbs, leaps and crawls on low-cost quadrupeds.将多个跑酷专家技能蒸馏为单一视觉策略，让低成本四足机器人攀爬、跳跃和匍匐。
@misc{zhuang2023robot,
title = {{Robot Parkour Learning}},
author = {Ziwen Zhuang and Zipeng Fu and Jianren Wang and Christopher Atkeson and Soeren Schwertfeger and Chelsea Finn and Hang Zhao},
year = {2023},
eprint = {2309.05665},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2309.05665},
}
Humanoid controller that scales to ten thousand motion clips with no external stabilising forces and recovers from falls by itself, so resets are not needed.人形控制器可扩展到一万条动作片段且无需外部稳定力，并能自行从跌倒中恢复，无需重置。
@misc{luo2023perpetual,
title = {{Perpetual Humanoid Control for Real-time Simulated Avatars}},
author = {Zhengyi Luo and Jinkun Cao and Alexander Winkler and Kris Kitani and Weipeng Xu},
year = {2023},
eprint = {2305.06456},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2305.06456},
}
Trains one quadruped policy that exposes gait parameters (frequency, foot swing, posture) as commands, so a human can tune behavior at deploy time for new tasks; open sim-to-real stack for the Unitree Go1.训练一个把步态参数（频率、抬脚高度、姿态等）作为指令开放的四足策略，部署时可由人调节行为以适应新任务；附带面向 Unitree Go1 的开源仿真到真实工具链。
@misc{margolis2022walk,
title = {{Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior}},
author = {Gabriel B Margolis and Pulkit Agrawal},
year = {2022},
eprint = {2212.03238},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2212.03238},
}
Learns reusable skill embeddings for simulated characters by combining adversarial imitation and unsupervised RL, so new tasks need not start from scratch.结合对抗模仿与无监督强化学习，为仿真角色学习可复用的技能嵌入，新任务不必从零训练。
Open-source NMPC + whole-body control + state-estimation stack for quadrupeds built on OCS2 and ros-control, with a sim-to-real path; a popular starting point for model-based legged control.基于 OCS2 与 ros-control 的开源四足控制栈：非线性 MPC + 全身控制 + 状态估计，并支持 sim2real；基于模型的足式控制常用的起点。
C++ toolbox for nonlinear MPC on switched systems (SLQ, DDP, SQP) with robot examples for legged robots and mobile manipulators; the base of many open-source legged MPC + whole-body control stacks.面向切换系统的非线性 MPC C++ 工具箱（SLQ、DDP、SQP），附足式机器人与移动操作臂示例；许多开源足式 MPC + 全身控制方案都基于它。
An open-source DDP-based solver for contact-rich trajectory optimization, widely used for legged and humanoid motion generation and MPC.基于 DDP 的开源接触丰富轨迹优化求解器，广泛用于足式与人形机器人的运动生成和 MPC。
@misc{mastalli2019crocoddyl,
title = {{Crocoddyl: An Efficient and Versatile Framework for Multi-Contact Optimal Control}},
author = {Carlos Mastalli and Rohan Budhiraja and Wolfgang Merkt and Guilhem Saurel and Bilal Hammoud and Maximilien Naveau and Justin Carpentier and Ludovic Righetti and Sethu Vijayakumar and Nicolas Mansard},
year = {2019},
eprint = {1909.04947},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/1909.04947},
}
The open-source Solo quadruped and its modular torque-controlled actuators, built from off-the-shelf and 3D-printed parts so labs can reproduce a dynamic legged robot.开源四足机器人 Solo 及其模块化力矩控制关节，采用现成件和 3D 打印件，便于各实验室复现动态腿式机器人。
@misc{grimminger2019open,
title = {{An Open Torque-Controlled Modular Robot Architecture for Legged Locomotion Research}},
author = {Felix Grimminger and Avadesh Meduri and Majid Khadiv and Julian Viereck and Manuel Wüthrich and Maximilien Naveau and Vincent Berenz and Steve Heim and Felix Widmaier and Thomas Flayols and Jonathan Fiene and Alexander Badri-Spröwitz and Ludovic Righetti},
year = {2019},
eprint = {1910.00093},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/1910.00093},
}
Convex MPC for quadrupeds — simplified rigid-body dynamics turn ground-reaction-force planning into a small QP solved in real time; the classic model-based baseline behind many open-source quadruped controllers.四足机器人的凸 MPC：用简化刚体动力学把足端地面反作用力规划变成可实时求解的小规模 QP；众多开源四足控制器背后的经典基于模型的基线。
Unitree's open VLA for humanoid manipulation: a VLM further pretrained on robot data, fine-tuned on 12 open Unitree G1 datasets; the README reports one policy handling 12 task categories on a real G1.宇树开源的人形机器人操作 VLA：在 UnifoLM-VLM 基础上用机器人操作数据继续预训练，再在 12 个公开的 Unitree G1 数据集上微调；README 称单一策略可在真实 G1 上完成 12 类复杂操作任务，另提供 LIBERO 微调权重。
Humanoid VLA family pre-trained on 2,000+ h of egocentric human and robot data, predicting latent motion tokens for a whole-body controller; uses pseudo-proprioception and arm re-rendering to bridge the embodiment gap.人形机器人通用 VLA 系列：在 2,000 多小时第一视角人类与机器人数据上预训练，输出潜在动作 token 驱动预训练的全身控制器并附手部信号；通过仿真回放构造伪本体感知、重渲染手臂外观缩小人-机差异，报告 HumanoidArena 最优，并在 Unitree G1 上验证人类数据增益。
@misc{liu2026huge,
title = {{HuGe: Scaling Humanoid Generalist Vision-Language-Action Models with Egocentric Experience}},
author = {Yang Liu and Jiacheng Liu and Hengtao Li and Yuheng Huang and Zijian Yang and Ji Dai and Maoxin Wang and Yi Gao and Qishen Wang and Baohua Yin and Peiyang Li and Shuanghao Bai and Mingyang Sun and Pengxiang Ding and Jianqin Yin and Lixin Yang and Donglin Wang},
year = {2026},
eprint = {2610.34186},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.34186},
}
An open foundation model for humanoid loco-manipulation that argues against co-training on mixed human and humanoid data because of their kinematic differences.面向人形移动操作的开放基础模型，主张因运动学差异，不宜直接混合人类与人形数据共同训练。
@misc{wei2026open,
title = {{$Ψ_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation}},
author = {Songlin Wei and Hongyi Jing and Boqian Li and Zhenyu Zhao and Jiageng Mao and Zhenhao Ni and Sicheng He and Jie Liu and Xiawei Liu and Kaidi Kang and Sheng Zang and Weiduo Yuan and Marco Pavone and Di Huang and Yue Wang},
year = {2026},
eprint = {2603.12263},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.12263},
}
Retargets human demos under contact without a differentiable simulator: gradients come from archived sampling rollouts. Reports 92.8% success on 1,352 bimanual OakInk2 demos with no task-specific contact rewards.无需可微仿真器即可在接触动力学下重定向人类示教：从存档的采样轨迹中恢复梯度。报告在 1352 条双手 OakInk2 示教上成功率 92.8%，且不用任务专属接触奖励。
@misc{liu2026grasp,
title = {{GRASP: Gradient-Guided Sampling for Contact-Rich Physics-Based Motion Retargeting}},
author = {Yang Liu and Tong Wang and Yiming Bao and Heng Yang},
year = {2026},
eprint = {2611.05754},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05754},
}
Adds a full-body tactile reflex that reaches the motors within 10 ms beneath a learned whole-body policy; on a G1 it cuts first force peak 34.1% versus the same reflex delayed by 20 ms.在学习型全身策略之下加入全身触觉反射，10 毫秒内直达电机；在 G1 上，与延迟 20 毫秒的相同反射相比，首次力峰值降低 34.1%。
@misc{wang2026reflexarc,
title = {{ReflexArc: Bottom-Up Integration of Instantaneous Tactile Reflexes into Whole-Body Humanoid Control}},
author = {Mingyu Wang and Si Li and Ziqing Chen and He Wang and Li Yi},
year = {2026},
eprint = {2611.04424},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04424},
}
A lightweight foothold prediction model strips terrain irrelevant to the next steps from the elevation map before the humanoid policy sees it, aiming at faster training and better transfer to new terrain.不再把整张高程图输入人形机器人运动策略，而是先用轻量的落脚点预测模型去掉与接下来几步无关的地形信息，以加快训练并提升对新地形的泛化。
@misc{heyrman2026watch,
title = {{Watch Your Step: Anticipated Perception for Humanoid Locomotion}},
author = {Matthias Heyrman and Chenhao Li and Ioannis Dadiotis and Marco Hutter},
year = {2026},
eprint = {2610.29612},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.29612},
}
HumanVerse, 500 h of egocentric human video with body and hand motion, aligned to the Unitree G1 for mid-training a humanoid model; reports 90% on SIMPLE and 62.5% on real loco-manipulation tasks.构建 500 小时的第一视角人类数据集 HumanVerse（含同步的身体与手部动作），对齐到 Unitree G1 后用于人形基础模型 λ0 的中期训练；论文报告 SIMPLE 上成功率 90%、真实移动操作任务 62.5%，人类数据越多下游表现越好。
@misc{team2026scaling,
title = {{$λ_0$: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation}},
author = {Lambda-humanoid Team and : and Chongyang Xu and Jin Chen and Jinhui Ye and Yilun Chen and Yiming Jiang and Zhao Wu and Hu Wei and Lin Qu and Shifeng Zhang and Yisheng He and Yuming Jiang and Zenan Meng and Zheng Wu and Mu Xu and Steven Hoi},
year = {2026},
eprint = {2610.18950},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18950},
}
Skips reconstruction and retargeting: a diffusion model turns monocular human video straight into humanoid reference motion, post-trained with DPO on tracking outcomes; real-time variant teleoperates a G1 under 400 ms.跳过人体重建与重定向：扩散模型直接把单目人类视频生成为人形机器人参考动作，并以运动跟踪结果做 DPO 后训练；自回归版本可在 Unitree G1 上实现低于 400 ms 延迟的实时视频遥操作。
@misc{he2026vit,
title = {{ViT-H: Generative Video-To-Humanoid Control with Motion Tracking Feedback}},
author = {Zimo He and Tianyi Bian and Zizhong Yang and Hongjie Li and Nan Jiang and Jieming Cui and Baoxiong Jia and Yizhou Wang and Siyuan Huang and Yixin Zhu},
year = {2026},
eprint = {2610.32966},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32966},
}
A quadruped solves a giant Rubik's Cube with its feet: onboard RGB-D estimates cube pose and colours, a planner sequences skills, and RL whole-body skills turn layers. Reports three full autonomous solves.四足机器人只用脚完成巨型魔方：机载 RGB-D 估计魔方位姿与颜色状态，规划器把解法转成技能序列，强化学习训练的全身技能负责翻转魔方、一脚固定一脚推动转层。报告完成三次完全自主求解。
@misc{amiet2026solving,
title = {{Solving a Giant Rubik's Cube with a Quadruped Robot}},
author = {Lukas Amiet and Gabriel Sánchez-Blanco Repgen and Mayank Mittal and Victor Klemm and Andrei Cramariuc and Marco Hutter and Cesar Cadena},
year = {2026},
eprint = {2610.24610},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.24610},
}
FLoRe: robot-free demos from force-sensing UMI-style grippers + body trackers train a diffusion policy that outputs per-hand force commands for a compliant whole-body controller; reports 80% vs. 26.7% for position-only.FLoRe：用带力传感的 UMI 式手持夹爪与人体追踪器采集无机器人示教，训练扩散策略输出每只手的力指令，由柔顺全身控制器执行；报告在三个接触密集任务上成功率 80%，仅位置控制为 26.7%。
@misc{li2026learning,
title = {{Learning Force-Reactive Humanoid Loco-Manipulation}},
author = {Sihang Li and Xiaofeng Guo and Jinkun Cao and Charlie Cheng and Ruihan Yang and Chen Feng and Guanya Shi and Haozhi Qi},
year = {2026},
eprint = {2610.35862},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35862},
}
Real-to-sim-to-real: smartphone scans become physics-ready sim scenes, whole-body motions are corrected by rollouts, giving 4,800 demos; a sim-only GR00T N1.7 on Unitree G1 reports 66.7% vs. 46.7% trained on real data.真实到仿真再到真实：用手机拍摄重建可物理仿真的场景，并根据仿真回放修正全身动作，生成 4800 条示教；仅用仿真数据训练的 GR00T N1.7 在宇树 G1 上报告平均成功率 66.7%，真实数据训练为 46.7%。
@misc{yang2026zero,
title = {{Zero-HLM: Generalizable Humanoid Loco-Manipulation without Real Robot Data}},
author = {Dianyi Yang and Mengya Liu and Zhumei Wang and Yu Liu and Xiaochen Zhang and Yixuan Li and Le Ma and Yuhang Yang and Weiqi Huang and Wei Liang and Baoxiong Jia and Siyuan Huang},
year = {2026},
eprint = {2610.35075},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35075},
}
Fine-tunes humanoid motion-tracking policies online on the real robot: a proprioceptive neural dynamics model learned from real data supplies gradients, no motion capture needed. Shown on Unitree G1.在真机上在线微调人形机器人动作跟踪策略：用真实数据训练的本体感知神经动力学模型提供可反传梯度，无需动捕系统。在 Unitree G1 上验证对多种扰动的快速适应。
@misc{chen2026humanoidndm,
title = {{HumanoidNDM: Closing the Training-Deployment Gap of Motion Tracking via Humanoid Neural Dynamics Model}},
author = {Chao Chen and Zhenrong Wang and Jilong Wang and Yusen Feng and Jiangran Lyu and Li Yi and He Wang},
year = {2026},
eprint = {2610.34126},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.34126},
}
Smooth analytic contact gradients only in the backward pass, hard-contact forward rollout unchanged: a quadruped trotting policy trains in under 2 min on a desktop and transfers to hardware without fine-tuning.只在反向传播中换用平滑的解析摩擦接触梯度，前向仍是硬接触仿真：四足小跑策略在台式机上不到 2 分钟训练完成，无需微调即可迁移到真机；在接触隐式轨迹优化中也能找到更丰富的接触解。
@misc{kim2026diffcontact,
title = {{DiffContact: Differentiable Contact Dynamics for Policy Learning and Contact-Implicit Trajectory Optimization}},
author = {Gijeong Kim and Seungbin Oh and Dongyun Kang and Hae-Won Park},
year = {2026},
eprint = {2610.35852},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35852},
}
Whole-body humanoid teleoperation from a compact target (root height and velocity + 4 keypoint poses, current frame only): ~20 ms headset-to-policy latency; a Unitree G1 catches a tossed ball.只用紧凑参考目标（根高度、根速度与 4 个稀疏关键点的当前帧位姿）实现人形机器人全身遥操作：头显到策略输出延迟约 20 ms，Unitree G1 能接住抛来的球并完成移动操作。
@misc{cao2026humit,
title = {{HuMiT: Low-Latency Whole-Body Humanoid Teleoperation via Compact Reference Tracking}},
author = {Liu Cao and Botian Xu and Ruoqu Chen and Jiajun Liu and Mingzhi Pei and Zihao Wang and Zhongyu Li and Mengdi Xu},
year = {2026},
eprint = {2610.29903},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.29903},
}
@misc{yi2026swap,
title = {{SWAP: Test-Time Steerable Humanoid Whole-Body Collision Avoidance}},
author = {Zeji Yi and Guanqi He and Xiaofeng Guo and Guannan Qu and Pieter Abbeel and Yan Duan and Karen Liu and Guanya Shi},
year = {2026},
eprint = {2610.25711},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25711},
}
FIBER plans in the latent space of a frozen behavioral foundation model: skills are projected into it, one perceptive flow planner is trained offline. Reports 42.0% mean sim success (latent replay: 11.6%); real G1.FIBER 以冻结的行为基础模型（BFM）潜在空间作为统一规划表示，把各任务专用策略投影进该空间，再离线训练单一感知流匹配规划器；仿真中搬箱、开门、抓取平均成功率 42.0%（潜在回放基线 11.6%），并在真实 Unitree G1 上验证。
@misc{lin2026action,
title = {{Action-Aligned Latent Planning for Versatile Humanoid Loco-Manipulation}},
author = {Yutang Lin and Guanya Shi and C. Karen Liu},
year = {2026},
eprint = {2610.24675},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.24675},
}
Unitree G1 walks at variable height (1.27–0.50 m) and crawls (0.50–0.20 m), with a learned transition policy between modes and a LiDAR planner that sets height targets for confined-space traversal on hardware.让 Unitree G1 在 1.27–0.50 米高度范围内变高步行、在 0.50–0.20 米范围内爬行，并用学习到的过渡策略在两种模式间切换；结合 LiDAR 规划器给出高度目标，在真机上穿越低矮受限空间。
@misc{xu2026reach,
title = {{REACH: An RL-based Hierarchical Humanoid Crawling Framework for Adaptable Confined Space Traversal}},
author = {Shengzhang Xu and Yuning Guo and Weiyi Hu and Yifeng Xu and Zhongyu Li and He Kong},
year = {2026},
eprint = {2610.36546},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36546},
}
Keeps a humanoid behavior foundation model frozen and trains a small lower-body residual adapter for stairs, gaps and platforms, distilled to onboard depth; tested on four trackers and on a Unitree G1 with SONIC.冻结人形行为基础模型，仅训练一个作用于下肢目标的轻量残差适配器来应对台阶、沟壑和平台，并从特权地形信息蒸馏到机载深度；在四种跟踪器上验证，并在 Unitree G1（SONIC）上真机部署。
@misc{li2026graft,
title = {{GRAFT: Geometry-Aware Residual Adaptation for Foundation Trackers}},
author = {Liang Li and Guoyao Zhang and Xingyu Chen and Sikai Wu and Qianyang Wu and Yuhang Lin and Renjing Xu and Mingliang Zhou and Diyun Xiang},
year = {2026},
eprint = {2610.36507},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36507},
}
A stochastic motion tokenizer gives a high-level generator a compact interface to a whole-body tracking policy; reports better G1 motion reconstruction than FSQ/VQ-VAE and real dance/fight/walk sequences.用随机运动分词器把行为标签或文本生成的动作压缩成离散表示，直接作为全身跟踪策略的参考；报告在 Unitree G1 动作重建上优于 FSQ 与 VQ-VAE，并在真机上执行舞蹈、格斗、行走及 35 秒的格斗转舞蹈序列。
RISE learns a vocabulary of recovery movements plus a visual diffusion planner whose predicted intent is steered at test time by task costs (heading, speed, path), no retraining; sim and real Unitree G1.RISE 学习一组可组合的起身动作词表，并用视觉扩散规划器预测未来身体状态作为动作意图；推理时用任务代价（朝向、起身速度、路径）引导该意图，无需重新训练即可选择不同起身方式；仿真实验并在 Unitree G1 真机演示。
SMR distils physics experts that track robot-aware targets into a strictly causal Transformer, retargeting human motion to humanoids frame by frame. Reports better closed-loop results than kinematic baselines.SMR 先让物理专家在仿真中跟踪由人体动作转换来的机器人任务空间目标，再用其执行轨迹训练严格因果的 Transformer，实现逐帧流式的人形动作重定向；报告在相同全身控制器下闭环表现优于运动学与物理精修基线。
@misc{gu2026smr,
title = {{SMR: Learning streaming motion retargeting for humanoids from human semantic}},
author = {Chun Gu and Juntong Zhang and Jingyu Li and Li Zhang},
year = {2026},
eprint = {2610.29169},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.29169},
}
One frozen humanoid policy driven by prompts (timed body-part targets, velocity, style); skills are plain code, refined by coding agents. On a Unitree G1 it walks, recovers, picks up objects and sits.可“提示”的人形全身执行层：冻结的生成式基础策略接收部分目标（定时的身体部位目标、速度指令、风格）生成全身动作；技能是普通代码，可由编码智能体分析动作数据后自动改进，无需按任务重训。在 Unitree G1 上一个检查点即可行走、恢复、遥操作抓取和坐下。
@misc{tessler2026metatron,
title = {{METATRON: A Promptable Whole-Body Execution Layer for Humanoids}},
author = {Chen Tessler and Yifeng Jiang and Alec Otto and Yi Shi and Umar Iqbal and Gal Chechik and Xue Bin Peng},
year = {2026},
eprint = {2610.38027},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.38027},
}
Open-weights whole-body motion tracker for the small Booster K1, adapting SONIC and trained on one shared GPU. Tracks most held-out clips in MuJoCo; on hardware, upper-body and slow motions work, stepping less so.将 SONIC 通用动作跟踪框架移植到小型人形 Booster K1，在一张共享 GPU 上从头训练并开放权重与代码。MuJoCo 中几乎能跟踪所有留出动作片段；真机上上半身和慢速动作表现良好，迈步与转身仍不稳定，作者坦诚指出对动作延迟敏感。
@misc{kolesnichenko2026sonic,
title = {{SONIC-K1: Porting a Universal Humanoid Motion Tracker to a Small Humanoid on a Single GPU}},
author = {Kyrylo Kolesnichenko and Jong-Hoon Kim},
year = {2026},
eprint = {2610.39327},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.39327},
}
When a quadruped manipulator keeps failing, a coding agent writes an RL problem to learn the missing whole-body skill, adds it to a library and retries; transfers to a real robot (18/20 rack retrieval).四足机械臂反复失败时，由编码智能体判断缺少哪种全身控制能力，自动编写强化学习问题（奖励、课程、终止条件）训练新控制器并加入技能库再重试；习得扶墙抬升等技能，仿真五项超出初始能力的任务成功率 98.4–100%，无需真机训练即迁移到实体机器人（货架取物 18/20）。
@misc{liu2026elevate,
title = {{ELEVATE: Agentic Controller Acquisition from Task Failure for Quadrupedal Manipulation}},
author = {Shiqi Liu and Radu Corcodel and Ding Zhao},
year = {2026},
eprint = {2610.38388},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.38388},
}
Movens-5M: 5M execution-verified humanoid motions from web videos, aligned with text, RGB, speech and music. Movens-Zero trains one flow-matching action expert on seven tasks for real-time closed-loop control.Movens-5M 从互联网视频构建 500 万条经执行验证的人形动作（超 1 万小时），对齐文本、RGB、语音和音乐。Movens-Zero 用一个流匹配动作专家联合训练七种生成任务，支持实时闭环执行；另有四赛道基准 Movens-Intent。
@misc{bu2026movens,
title = {{Movens-Zero: Scaling Humanoid Motion Foundation Models with Large-Scale Human Videos}},
author = {Wendong Bu and Tinghong Ye and Qizhou Wang and Jiacheng Li and Zhiqi Ge and Yuze Lin and Zhongqi Yue and Yi Su and Jingyu Wang and Kai Shen and Jingdong Liu and Chengyuan Deng and Xiaoxuan Chen and Xin Tan and Jiaming Ji and Wenqiao Zhang and Chenghong Jiang and Tuoyu Li and Siliang Tang and Jun Xiao and Tat-Seng Chua and Yueting Zhuang and Xingxing Wang and Juncheng Li},
year = {2026},
eprint = {2610.41877},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41877},
}
Humanoid tracker as a one-step diffusion model: one forward pass per 50 Hz step, test-time steering via cost gradients, failure discovery from plan dispersion without simulation. Deployed on a Unitree G1.用一步扩散模型做人形全身跟踪：每个 50 Hz 控制步只需一次前向；用可微代价梯度在测试时操控冻结的控制器，无需重训；用规划分散度在不做物理仿真的情况下发现跟踪失败。已部署到 Unitree G1。
@misc{huang2026diffwbc,
title = {{DiffWBC: Unlocking Steerable General Humanoid Motion Tracking via One-Step Diffusion Whole-Body Control}},
author = {Yifei Huang and Yazhao Yang and Weijia Li and Zhiheng Ma and Dian Shen},
year = {2026},
eprint = {2610.41354},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41354},
}
Fall recovery beneath overhead obstacles from proprioception alone: rewards reducing robot-obstacle overlap, no prescribed path. Reports 95 of 150 recoveries on a Unitree G1 across seven real obstacle scenes.人形在头顶障碍物下跌倒后的恢复，仅用本体感知：奖励减少机器人与障碍物的重叠，而不预设逃脱路径，由单一策略学会何时清障、建立支撑并站起。报告 Unitree G1 在七个真实场景中 150 次尝试成功 95 次。
@misc{lu2026roof,
title = {{ROOF: Learning Humanoid Fall Recovery under Overhead Constraints}},
author = {Yidan Lu and Jiayuan Yu and Zongrui Cai and Jingwen Sun and Jinrui Han and Jing Wang and Ping Luo and Peng Lu},
year = {2026},
eprint = {2610.42097},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.42097},
}
Learns a reduced-order reference generator, then trains a whole-body tracker on its online rollouts, so one pair of policies covers many gaits. On a Unitree G1: -2.0 to 2.8 m/s, hopping and continuous transitions.先学习降阶模型参考生成器，再用其在线输出训练全身跟踪策略，一对策略即可覆盖多种步态。Unitree G1 实机实现 -2.0 到 2.8 m/s 行走、单腿跳与连续切换。
@misc{ghansah2026forge,
title = {{FORGE: Feasible Online Reduced-Order Reference Generation and Whole-Body Execution for Agile Humanoid Locomotion}},
author = {Adrian B. Ghansah and Yujin An and Jeeseop Kim and Soon-Jo Chung},
year = {2026},
eprint = {2610.37974},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37974},
}
Closes the loop between a generative motion model and a humanoid tracker, which usually run open-loop and drift. Deviation-decayed recovery and history re-anchoring keep the feedback from causing jitter or falls.把通常开环运行、会累积漂移的生成式运动模型与人形跟踪器闭环耦合。真实状态反馈易致抖动甚至摔倒，因此加入偏差衰减恢复、历史重锚定与延迟补偿。
@misc{han2026hipc,
title = {{HIPC: Humanoid Iterative Planning and Control via Goal-Conditioned Motion Closure}},
author = {Yi Han and Xingyu Chen and Zixuan Lang and Qianyang Wu and Yuhang Lin and Ji Ma and Liang Li and Dongting Li and Xinpeng Liu and Ruoli Dai and Hesheng Wang and Renjing Xu and Lei Han},
year = {2026},
eprint = {2610.41334},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41334},
}
CoRL 2026 Oral. A fully virtual pipeline for humanoid loco-manipulation data: grounds video world models in 3D assets and simulator-ready scenes, with no teleoperation or physical setups.CoRL 2026 Oral。全虚拟的人形移动操作数据管线：把视频世界模型落到 3D 资产与可仿真场景中，无需遥操作或搭建实体环境。
@misc{xie2026grail,
title = {{GRAIL: Generating Humanoid Loco-Manipulation from World Models}},
author = {Tianyi Xie and Haotian Zhang and Jinhyung Park and Zi Wang and Bowen Wen and Jiefeng Li and Xueting Li and Qingwei Ben and Haoyang Weng and Yufei Ye and David Minor and Tingwu Wang and Chenfanfu Jiang and Sanja Fidler and Jan Kautz and Linxi Fan and Yuke Zhu and Zhengyi Luo and Umar Iqbal and Ye Yuan},
year = {2026},
eprint = {2606.05160},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.05160},
}
CoRL 2026 Oral. Three-stage curriculum for humanoid soccer shooting: a single human-kick reference serves as a scaffold, bridging stable motion tracking and goal-driven RL that struggles to discover kicks.CoRL 2026 Oral。人形足球射门的三阶段课程：以单条人类踢球参考为脚手架，兼顾动作跟踪的稳定性与任务奖励 RL 的目标适应能力。
@misc{zhong2026robonaldo,
title = {{RoboNaldo: Accurate, Stable and Powerful Humanoid Soccer Shooting via Motion-Guided Curriculum Reinforcement Learning}},
author = {Yichao Zhong and Yidan Lu and Yuhang Lu and Tianyang Tang and Haoguang Mai and Yixuan Pan and Tianyu Li and Li Chen and Jingbo Wang and Zhongyu Li and Peng Lu and Hongyang Li},
year = {2026},
eprint = {2606.11092},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.11092},
}
CoRL 2026 Spotlight. Language-conditioned humanoid traversal of cluttered rooms that treats navigation as whole-body adaptation of arms, torso and gait rather than 2D path planning.CoRL 2026 Spotlight。语言条件下的人形机器人杂乱环境穿行：不把导航当二维路径规划，而是手臂、躯干与步态的全身协同调整。
@misc{li2026tango,
title = {{TANGO: Humanoid Navigation in Cluttered Environments with a Whole-Body Vision-Language-Action Model}},
author = {Anqi Li and Yuxin Chen and Zhaobo Li and Zhuo Cao and Junli Ren and Masayoshi Tomizuka and Dhruv Shah},
year = {2026},
eprint = {2609.09158},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.09158},
}
CoRL 2026 Spotlight. A one-variable-at-a-time empirical study of what a whole-body native VLA needs, mapping language and pixels directly to all humanoid degrees of freedom.CoRL 2026 Spotlight。逐变量对照的实证研究：探索直接把语言与图像映射到人形全部自由度的全身原生 VLA 需要哪些条件。
@misc{hu2026openhlm,
title = {{OpenHLM: An Empirical Recipe for Whole-Body Humanoid Loco-Manipulation}},
author = {Yingdong Hu and Haodong Zhu and Boyuan Zheng and Yihang Hu and Tong Zhang and Zunhao Chen and Junming Zhao and Ruiqian Nai and Yang Gao},
year = {2026},
eprint = {2606.22174},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.22174},
}
CoRL 2026 Spotlight. A humanoid system for climbing diverse ladders and manipulating while on them; a two-stage pipeline learns climbing experts through hybrid motion tracking.CoRL 2026 Spotlight。让人形机器人攀爬多种梯子并在梯上操作；两阶段流程用混合动作跟踪学习多个攀爬专家策略。
@misc{zhao2026ladderman,
title = {{LadderMan: Learning Humanoid Perceptive Ladder Climbing}},
author = {Siheng Zhao and Yuanhang Zhang and Ziqi Lu and Pieter Abbeel and Rocky Duan and Koushil Sreenath and Yue Wang and C. Karen Liu and Guanya Shi},
year = {2026},
eprint = {2606.05873},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.05873},
}
A compact planner sends body-point targets to a frozen pretrained whole-body controller; the seven minutes excludes controller pretraining. Evaluated in simulation.紧凑规划器向冻结的预训练全身控制器发送身体点目标；所称七分钟不含控制器预训练。评估在仿真中进行。
@misc{ma2026holoboxing,
title = {{HoloBoxing: A Simple Recipe for Humanoid Boxing with 7 Minutes of Single-GPU Training}},
author = {Xihan Ma and Bo Zhang and Boxiong Li and Yuhan Bao and Ziyue Liu and Yucheng Wang},
year = {2026},
eprint = {2610.32909},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32909},
}
Scores footstep instability from factor-graph residuals in legged state estimation and projects it into earlier images to supervise a dense visual predictor.用足式状态估计因子图的残差给落足不稳定性打分，并投影到之前的图像上，监督稠密视觉预测器。
@misc{fortin2026tacos,
title = {{TACOS: Terrain-Aware COntact Supervision for Visual Footstep Planning}},
author = {Jean-Michel Fortin and William Larrivée-Hardy and Gabriel Jeanson and Olivier Gamache and Daniele De Martini and François Pomerleau and Philippe Giguère},
year = {2026},
eprint = {2610.37806},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37806},
}
Trained in simulation with no real demos. Controller-in-the-loop filtering rejects 73% of kinematically valid grasps that would destabilize the robot; reports 87.5% real-world success on a multi-level shelf task.完全在仿真中训练，无真实示范。把控制器放进回路筛选数据，剔除 73% 运动学可行但会使机器人失稳的抓取；作者报告多层货架任务真实成功率 87.5%。
@misc{peng2026hug,
title = {{HuG: Learning Humanoid Whole-Body Affordance-Aware Grasping with Active Perception}},
author = {Xuanbin Peng and Lai Wei and Yuze Zhai and Qingyuan Yu and Keyi Wang and Jianglong Ye and Xiaolong Wang},
year = {2026},
eprint = {2611.01659},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.01659},
}
Picks at runtime which low-level policy to run and with which goal, using a learned energy per policy. New policies join by training their own energy model, with no retraining of others. Deployed on a real humanoid.运行时为每个底层策略学习能量函数，据此选择执行哪个策略、给它什么目标。新增策略只需训练自己的能量模型，无需重训其他策略或中央调度器；已部署到真实人形的移动操作任务。
@misc{ye2026energy,
title = {{Energy-Based Policy Orchestration}},
author = {Weirui Ye and Yen-Jen Wang and Elvis Hsieh and Zhiyang Dou and Shibo Zhao and Pieter Abbeel and Yilun Du and Leslie Pack Kaelbling and Tomás Lozano-Pérez},
year = {2026},
eprint = {2611.02546},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.02546},
}
Dynamic routing makes policy-update divergence misestimated in mixture-of-experts policies, so MAPO optimizes router and experts in turn. Reports tracking 182 motions PPO with MoE fails on, against one the other way.动态路由会让混合专家策略的更新散度估计失准，MAPO 因此交替优化路由与专家。报告可跟踪 182 个 PPO+MoE 失败的动作，反过来只有 1 个。
@misc{lim2026mapo,
title = {{MAPO: Mixture of Experts with Alternating Policy Optimization for Universal Humanoid Motion Tracking}},
author = {Seonghyeon Lim and Hyeonwoo Lee and Mincheol Kim and Kimin Lee and Hyun Myung},
year = {2026},
eprint = {2611.02613},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.02613},
}
@misc{qi2026humanoid,
title = {{Humanoid-GPT: Scaling Data and Structure for Zero-Shot Motion Tracking}},
author = {Zekun Qi and Xuchuan Chen and Dairu Liu and Chenghuai Lin and Yunrui Lian and Sikai Liang and Zhikai Zhang and Yu Guan and Jilong Wang and Wenyao Zhang and Xinqiang Yu and He Wang and Li Yi},
year = {2026},
eprint = {2606.03985},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.03985},
}
Turns diverse human videos into deployable humanoid loco-manipulation skills without task-specific rewards, despite occlusion and retargeting errors in the priors.把多样的人类视频转化为可部署的人形移动操作技能，无需任务专用奖励，能应对先验中的遮挡与重定向误差。
@misc{wu2026sugar,
title = {{SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework}},
author = {Tianshu Wu and Xiangqi Kong and Yue Chen and Qize Yu and Hang Ye and Jia Li and Yizhou Wang and Hao Dong},
year = {2026},
eprint = {2605.20373},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.20373},
}
Trains the low-level humanoid controller by motion imitation with the object physically in the scene, then a high-level policy on top; on a Unitree G1 it carries varied boxes and recovers from drops.让人形底层控制器在场景中真实带着物体做动作模仿训练，再在其上训练高层策略；在宇树 G1 上可搬运不同箱子，并能从掉落中恢复。
@misc{sleiman2026hlo,
title = {{HLO-AMP: Hierarchical Loco-Manipulation from Object-Aware Motion Priors}},
author = {Jean-Pierre Sleiman and Bilal Hammoud and He Li and Jonathan Foster and Sylvain Bertrand and Jessica K. Hodgins and Farbod Farshidian},
year = {2026},
eprint = {2611.05113},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05113},
}
A latent world-action model scores proposed humanoid command sequences before running them and replans each step; reports 85.7% on seven real tasks against 70.0% for GR00T N1.7.用潜空间世界动作模型在执行前为人形指令序列打分，并逐步重规划；报告七项真机任务成功率 85.7%，GR00T N1.7 为 70.0%。
@misc{dong2026coordinating,
title = {{Coordinating Humanoid Loco-Manipulation through Action-Conditioned Latent Foresight}},
author = {Ju Dong and Xinyu Xiao and Yiming Zhou and Tianshuo Yang and Xuerong Liu and Lei Zhang and Jian Chen and Haocheng Zhao and Yu Fu and Yujie Wei and Diwen Zheng and Alois Christian Knoll and Angela P. Schoellig and Ping Luo and Jianwei Zhang},
year = {2026},
eprint = {2611.04839},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04839},
}
Uses sample-based MPC in simulation as an automatic expert, then trains off-policy RL on sparse rewards, avoiding reward shaping; deployed on an arm-equipped Spot and a G1 humanoid.在仿真中用采样 MPC 自动生成专家数据，再以稀疏奖励训练离策略 RL，免去奖励塑形；已部署到带臂 Spot 和 G1 人形机器人。
@misc{schuck2026learning,
title = {{Learning Loco-Manipulation From SMPC Demonstrations With Sparse Offline-to-Online RL}},
author = {Martin Schuck and Maks Sorokin and Simone Manni and Duy Ta and Angela P. Schoellig and Marco Hutter and Simon Le Cleac'H and Jan Brüdigam},
year = {2026},
eprint = {2608.12063},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.12063},
}
Separates performance from safety on humanoids: the nominal policy handles the task while a residual RL policy acts as an implicit safety filter, avoiding competing reward terms.在人形机器人上分离性能与安全：名义策略负责任务，残差 RL 策略充当隐式安全过滤器，避免相互竞争的奖励项。
@misc{qu2026ressafe,
title = {{ResSafe: Learning Safety Filtering with Residual Reinforcement Learning for Humanoids}},
author = {Gechen Qu and Tong Zhang and Bike Zhang and Yen-Jen Wang and Koushil Sreenath and Claire Tomlin and Jason Jangho Choi},
year = {2026},
eprint = {2609.15988},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.15988},
}
Adds exteroceptive sensing to whole-body motion tracking, so a humanoid performs dynamic non-locomotion skills on uneven terrain.将外部感知融入全身动作跟踪，让人形机器人在不平整地形上完成高动态的非行走类技能。
An end-to-end perceptive framework mapping depth to control for humanoid hiking over diverse unstructured terrain.端到端感知框架，将深度输入映射到控制，实现人形机器人在多样非结构化地形上徒步。
@misc{zhu2026hiking,
title = {{Hiking in the Wild: A Scalable Perceptive Parkour Framework for Humanoids}},
author = {Shaoting Zhu and Ziwen Zhuang and Mengjie Zhao and Kun-Ying Lee and Hang Zhao},
year = {2026},
eprint = {2601.07718},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2601.07718},
}
@misc{wu2026perceptive,
title = {{Perceptive Humanoid Parkour: Chaining Dynamic Human Skills via Motion Matching}},
author = {Zhen Wu and Xiaoyu Huang and Lujie Yang and Yuanhang Zhang and Xi Chen and Pieter Abbeel and Rocky Duan and Angjoo Kanazawa and Carmelo Sferrazza and Guanya Shi and C. Karen Liu},
year = {2026},
eprint = {2602.15827},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.15827},
}
Targets the 'generality barrier': tracking fidelity collapsing as motion libraries grow, especially for high-dynamic motions on real hardware.针对“通用性壁垒”：动作库扩大后跟踪精度崩塌，在真机高动态动作上尤为明显。
@misc{wang2026omnixtreme,
title = {{OmniXtreme: Breaking the Generality Barrier in High-Dynamic Humanoid Control}},
author = {Yunshen Wang and Shaohang Zhu and Peiyuan Zhi and Yuhan Li and Jiaxin Li and Yong-Lu Li and Yuchen Xiao and Xingxing Wang and Baoxiong Jia and Siyuan Huang},
year = {2026},
eprint = {2602.23843},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.23843},
}
@misc{wang2026humanx,
title = {{HumanX: Toward Agile and Generalizable Humanoid Interaction Skills from Human Videos}},
author = {Yinhuai Wang and Qihan Zhao and Yuen Fui Lau and Runyi Yu and Hok Wai Tsui and Qifeng Chen and Jingbo Wang and Jiangmiao Pang and Ping Tan},
year = {2026},
eprint = {2602.02473},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.02473},
}
Motion imitation trained from mocap, noisy monocular video and animation, deployed to hardware zero-shot with less per-skill tuning.基于动捕、含噪单目视频和动画数据训练动作模仿策略，零样本部署到真机，减少逐技能调参。
@misc{sleiman2026zest,
title = {{ZEST: Zero-shot Embodied Skill Transfer for Athletic Robot Control}},
author = {Jean Pierre Sleiman and He Li and Alphonsus Adu-Bredu and Robin Deits and Arun Kumar and Kevin Bergamin and Mohak Bhardwaj and Scott Biddlestone and Nicola Burger and Matthew A. Estrada and Francesco Iacobelli and Twan Koolen and Alexander Lambert and Erica Lin and M. Eva Mungai and Zach Nobles and Shane Rozen-Levy and Yuyao Shi and Jiashun Wang and Jakob Welner and Fangzhou Yu and Mike Zhang and Alfred Rizzi and Jessica Hodgins and Sylvain Bertrand and Yeuhi Abe and Scott Kuindersma and Farbod Farshidian},
year = {2026},
eprint = {2602.00401},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.00401},
}
Adds an explicit 3-D force command per contact body to motion tracking, set online by a planner. On hardware it pushes and kicks boxes up to 20 kg where a force-blind baseline fails, and adapts to added loads.在动作跟踪中为每个接触部位加入显式三维力指令，由高层规划器在线给出。实机可推、踢最重 20 kg 的箱子（不感知力的基线会失败），并能适应任务中途增加的负载。
@misc{zhao2026ctrl,
title = {{Ctrl-F: Force as an Interface for Humanoid Loco-Manipulation}},
author = {Mengjie Zhao and Yuxin Chen and Yiyang Shao and Chuye Hong and Jiakang Chen and Koushil Sreenath and Masayoshi Tomizuka},
year = {2026},
eprint = {2611.05507},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05507},
}
PRISM turns a few real videos into hundreds of counterfactual human-object clips via video generation, rebuilds them in sim, and trains one humanoid carrying policy that handles unseen objects with no real fine-tuning.PRISM 用视频生成把少量真实视频扩增为数百段反事实人-物交互视频，在仿真中重建并训练单一人形搬运策略，无需真机微调即可泛化到未见物体。
@misc{wang2026counterfactual,
title = {{Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation}},
author = {Zihan Wang and Zhen Wu and Pieter Abbeel and Rocky Duan and Jitendra Malik and Carmelo Sferrazza and C. Karen Liu and Guanya Shi and Angjoo Kanazawa},
year = {2026},
eprint = {2609.38172},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.38172},
}
Modular humanoid grasping: large vision models handle the scene, a sim-trained end-effector tracker with IK and a learned forward model handles the arm. Reports 2.44 cm tracking error, 5.5x better than prior work.模块化人形抓取：大视觉模型负责场景理解，仿真训练的末端跟踪策略结合 IK 与学习的正运动学模型负责执行。报告末端跟踪误差 2.44 厘米，比先前最好方法低 5.5 倍。
@misc{dong2026hero,
title = {{HERO: Learning Humanoid End-Effector Control for Visual Whole-Body Open-Vocabulary Object Grasping}},
author = {Runpei Dong and Ziyan Li and Arjun Gupta and Xialin He and Saurabh Gupta},
year = {2026},
eprint = {2602.16705},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.16705},
}
Self-supervised foot-contact estimation from joint encoders and IMU only, fed as a stance belief into a Kalman filter. Reports the lowest leg-odometry error among the compared baselines.仅用关节编码器与 IMU 自监督估计足端接触，并作为支撑置信度输入卡尔曼滤波器。报告腿式里程计误差在对比方法中最低。
Humanoid that reacts to a human partner: a token model generates G1 whole-body motion conditioned on the partner, and a tracker executes it. Adds a 260-hour human-humanoid interaction benchmark.让人形机器人对人类伙伴做出回应：token 化序列模型以伙伴动作为条件生成 G1 全身动作，再由跟踪控制器执行。并提供 260 小时的人-人形交互基准。
@misc{maluleke2026social,
title = {{Social Humanoid: Socially Responsive Human-Humanoid Interaction}},
author = {Vongani H. Maluleke and James Ni and Kie Horiuchi and Lea Wilken and Jiahui Lei and Evonne Ng and Roei Herzig and Jitendra Malik and Trevor Darrell and Angjoo Kanazawa},
year = {2026},
eprint = {2611.08407},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08407},
}
Harness that lets a VLM drive a humanoid: it points at image locations for 3D positions and issues pelvis, torso, hand, walk and gaze commands. After fine-tuning, a real G1 carries a box with no real data.为 VLM 控制人形机器人提供“挽具”接口：在图像上点选即得 3D 位置，并下达骨盆、躯干、手、行走和视线指令。用教师轨迹微调后，真实 G1 无需真机数据即可搬箱子。
@misc{lu2026hitch,
title = {{HITCH: Teaching a Vision-Language Model to Operate a Humanoid through a Whole-Body Harness}},
author = {Qingzhou Lu and Chenxin Liu and Xinming Wang and Liu Cao and Yanjiang Guo and Mengdi Xu and Jianyu Chen},
year = {2026},
eprint = {2611.08039},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08039},
}
REACT learns humanoid-human contact interactions from human-human mocap: VLM-guided contact retargeting, then onboard-depth policies that also predict the occluded partner. A real G1 does all five interactions.REACT 从人-人动捕中学习人形与人的接触交互：VLM 引导接触重定向，再蒸馏为只用机载深度的策略，并预测被遮挡伙伴的状态。真实 G1 完成全部五种交互。
One command-conditioned policy for humanoid soccer skills (dribbling, trapping, shooting) with per-skill motion priors and critics; deployed on a real Noetix E1.单一指令条件策略实现人形足球多种技能（带球、停球、射门），每个技能有独立的运动先验与评论家；部署在真实 Noetix E1 上。
@misc{ye2026skillx,
title = {{SkillX: Unified Multi-Skill Policy Learning for Humanoid Soccer}},
author = {Zhangchen Ye and Enxuan Ruan and Yifei Bao and Runhan Huang and Jiankun Yang and Jiakang Jin and Yixiao Huo and Pengyuan Wang and Yinan Han and Huaxing Huang and Wenhao Cui and Yiming Li and Xiaoyu Tian},
year = {2026},
eprint = {2609.06718},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.06718},
}
Perceptive humanoid locomotion that keeps sharp terrain edges and assigns credit over time during distillation. Reports 96% stepping-stone success and a 1.5 km outdoor walk with one depth camera.感知式人形行走：保留地形的清晰边缘，并在蒸馏时做跨时间的信用分配。报告踏石通过率 96%，仅用一个深度相机完成 1.5 公里户外路线。
@misc{sun2026solo,
title = {{SOLO: Stable Omni-terrain Long-Horizon Perceptive Humanoid Locomotion}},
author = {Pihai Sun and Gang Han and Jingkai Sun and Jiahao Ma and Zeran Su and Zelin Tao and Peiran Liu and Shuai Shi and Wei Cui and Zifan Wang and Jialin Yu and Wen Zhao and Kangning Yin and Jiaxu Wang and Jiahang Cao and Lingfeng Zhang and Hao Cheng and Jian Tang and Qiang Zhang and Yijie Guo},
year = {2026},
eprint = {2608.26583},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.26583},
}
Humanoid box-rearranging challenge that chains pickup, walking and placing skills. A shared whole-body controller hardened with closed-loop skill data composes best; controller handoffs cause much of the loss.人形机器人“汉诺塔”式搬箱挑战，需要反复串联拾取、行走与放置技能。发现用技能闭环数据加固的共享全身控制器组合最稳健，控制器切换是性能损失的重要来源。
@misc{kim2026humanoid,
title = {{Humanoid Hanoi: Investigating Shared Whole-Body Control for Skill-Based Box Rearrangement}},
author = {Minku Kim and Kuan-Chia Chen and Aayam Shrestha and Li Fuxin and Stefan Lee and Alan Fern},
year = {2026},
eprint = {2602.13850},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.13850},
}
Humanoid motion tracking that follows explicit part-level contact commands as well as keypoints, so the robot really sits, wipes or pushes, and contact can be switched on or off. 10 sim motions, 5 on hardware.人形动作跟踪在关键点之外还跟踪部位级接触指令，让机器人真正坐下、擦拭或推动物体，并可按需产生或抑制接触。仿真 10 种动作，真机 5 种。
@misc{li2026contactmimic,
title = {{ContactMimic: Humanoid Object Interaction via Contact Control}},
author = {Xinyao Li and Xialin He and Runpei Dong and Saurabh Gupta},
year = {2026},
eprint = {2607.08742},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.08742},
}
Hybrid simulator (mesh foreground over a calibrated 3DGS background) that generates humanoid loco-manipulation demos without teleop. On a G1, VLAs trained only on it match or beat teleop-trained ones.混合模拟器：网格前景叠加校准过的 3D 高斯泼溅背景，无需遥操作即可生成人形移动操作演示。在 G1 上，仅用其数据训练的 VLA 达到或超过遥操作数据训练的效果。
@misc{kim2026legs,
title = {{LEGS: Fine-Tuning Teleop-Free VLAs for Humanoid Loco-manipulation in an Embodied Gaussian Splatting World}},
author = {Hojune Kim and Timothy Chen and Jiankai Sun and Lars W. Osterberg and Qianzhong Chen and Ke Wang and Mac Schwager},
year = {2026},
eprint = {2606.01458},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.01458},
}
Low-latency humanoid teleoperation: scaled velocity feedforward cuts tracking delay by ~70 ms and link-pose control skips retargeting. End-to-end latency down to 50 ms, enough for juggling and ping-pong ball balancing.低延迟人形遥操作：缩放速度前馈让跟踪延迟减少约 70 ms，直接控制末端连杆位姿省去关节重定向。端到端延迟低至 50 ms，可完成抛接与乒乓球平衡等反应型动作。
Generates 48,000 vision-language-kinematics trajectories in 3DGS-reconstructed rooms with no human input, trains a policy that outputs whole-body kinematics, and runs it on a real G1 via a tracker.在 3D 高斯泼溅重建的室内场景中无人工合成 4.8 万条“视觉-语言-运动学”轨迹，训练输出全身运动学轨迹的策略，并通过全身跟踪器在真实 G1 上运行。
@misc{wang2026vlk,
title = {{VLK: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed Scenes}},
author = {Yen-Jen Wang and Jiaman Li and Sirui Chen and Takara E. Truong and Pei Xu and Pieter Abbeel and Rocky Duan and Koushil Sreenath and Angjoo Kanazawa and Carmelo Sferrazza and Guanya Shi and Karen Liu},
year = {2026},
eprint = {2606.30645},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.30645},
}
MimicGen-style data generation for humanoids: adapts a few whole-body demos to new object poses and links them with locomotion planning, plus a 9-task benchmark. Co-training reports +20% over real data only.面向人形机器人的 MimicGen 式数据生成：把少量全身演示适配到新物体位姿，并用移动与操作规划串联，另提出 9 任务基准。协同训练比仅用真机数据提升 20%。
@misc{lin2026humanoidmimicgen,
title = {{HumanoidMimicGen: Data Generation for Loco-Manipulation via Whole-Body Planning}},
author = {Kevin Lin and Ajay Mandlekar and Caelan Reed Garrett and Nikita Chernyadev and Yu Fang and Runyu Ding and Yuqi Xie and Justin Tran and Linxi Fan and Yuke Zhu},
year = {2026},
eprint = {2605.27724},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.27724},
}
Teleoperation for a full-size humanoid carrying heavy loads: privileged motion guidance cleans noisy VR references and a payload curriculum builds robustness. A 175 cm, 65 kg robot tracks motions with up to 24 kg.面向全尺寸人形机器人的重载遥操作：特权运动引导修正带噪 VR 参考，窗口化载荷课程提升鲁棒性。175 cm、65 kg 的机器人可携带最多 24 kg 负载跟踪动作。
@misc{liu2026heft,
title = {{HEFT: Heavy-Payload Full-size Humanoid Teleoperation with Privileged Motion Guidance and Windowed Payload Curriculum}},
author = {Chenxin Liu and Qingzhou Lu and Guangxiao Yang and Xuanyang Shi and Chenghan Yang and Yanjiang Guo and Jianyu Chen},
year = {2026},
eprint = {2607.02332},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.02332},
}
Adds runtime constraints (collisions, joint limits, CoM stability) to RL humanoid tracking policies via operational-space control and CBFs at over 1 kHz, even for constraints set after training.借助操作空间控制与控制障碍函数，以 1 kHz 以上频率为强化学习人形跟踪策略施加运行时约束（碰撞、关节限位、质心稳定），训练后新增的约束也能满足。
Learns professional tennis serve and rally styles from broadcast video with a planner-plus-tracker design and speed adaptation for sim-to-real. Deployed on a G1 and a 1.7 m Dobot Atom serving without mocap.从比赛转播视频学习职业网球发球与对打风格，采用“规划器+跟踪器”分层结构并加入速度自适应缩小仿真-真机差距。部署在 G1 和 1.7 m 的 Dobot Atom 上，无需动捕即可发球。
@misc{huang2026towards,
title = {{Towards Professional Tennis Styles for Humanoid Robots with Adaptive Motion Planning and Tracking}},
author = {Tao Huang and Ruofei Liu and Xuchen Tang and Xinyin Zhang and Junli Ren and Huayi Wang and Feiyu Jia and Yukai Qi and Kangning Yin and Weishuai Zeng and Lipeng Chen and Xi Li and Ting Wu and Kailin Li and Ruoli Dai and Jingbo Wang and Lei Han and Jiangmiao Pang},
year = {2026},
eprint = {2608.20087},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.20087},
}
Real-time text-driven humanoid control: an autoregressive motion diffusion model streams short trajectories from the current command, which can change mid-motion, and a tracking policy executes them on hardware.实时文本驱动的人形机器人控制：自回归运动扩散模型根据当前（可随时修改的）文字指令持续生成短时轨迹，再由运动跟踪策略在真机上执行，可流畅切换舞蹈、跳跃等动作。
@misc{xie2026textop,
title = {{TextOp: Real-time Interactive Text-Driven Humanoid Robot Motion Generation and Control}},
author = {Weiji Xie and Jiakun Zheng and Jinrui Han and Jiyuan Shi and Weinan Zhang and Chenjia Bai and Xuelong Li},
year = {2026},
eprint = {2602.07439},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.07439},
}
Samples many motions from a text-to-motion generator and picks one that a learned verifier says the humanoid can track and that matches the text. Improves tracking and text alignment, incl. on a real G1.从文本生成运动模型采样多个候选，用学习到的动态可行性验证器与语义对齐验证器挑选人形机器人既能跟踪又符合文字的动作。仿真与真实 G1 上跟踪精度与文本一致性均提升。
@misc{cao2026texedo,
title = {{TEXEDO : Test Time Scaling for Controller-aware Language-conditioned Humanoid Motion Generation}},
author = {Jianuo Cao and Yuxin Chen and Yuzhen Song and Masayoshi Tomizuka and Chenran Li and Thomas Tian},
year = {2026},
eprint = {2606.22998},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.22998},
}
End-to-end text-conditioned humanoid control: a diffusion action prior learned from text-labelled state-action data, plus a light residual RL policy for robustness and smooth switching between commands.端到端的文本条件人形全身控制：从带文本标签的状态-动作轨迹学习扩散动作先验，再叠加轻量残差强化学习策略提升长时鲁棒性和指令切换的平滑度。
@misc{wu2026adapt,
title = {{ADAPT: Agile Diffusion Action Priors for Robust and Steerable Online Text-Driven Humanoid Control}},
author = {Yan Wu and Chenhao Li and Kaifeng Zhao and Gen Li and Marco Hutter and Siyu Tang},
year = {2026},
eprint = {2609.00677},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.00677},
}
Planner plus inverse-dynamics student distilled from a privileged policy; at deployment only the IDM is fine-tuned on about 2 minutes of target-domain rollouts. Beats in-context and end-to-end adaptation baselines.从特权策略蒸馏出“规划器+逆动力学模型”学生策略；部署时冻结规划器，只用约 2 分钟目标域数据微调逆动力学模型。优于上下文自适应与端到端适应基线。
@misc{xie2026fada,
title = {{FADA: Few-Shot Domain Adaptation via Dynamics Alignment for Humanoid Control}},
author = {Angchen Xie and Nikhil Sobanbabu and Ishayu Shikhare and Alan Wang and Max Simchowitz and Guanya Shi},
year = {2026},
eprint = {2606.28476},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.28476},
}
Humanoid gaze system: a VLM picks where to look and a Transformer-VAE trained on retargeted human recordings generates coordinated eye, neck and eyelid motion. Lowest joint error on a 2-second benchmark.人形机器人注视系统：VLM 选择注视目标，基于重定向人类录像训练的 Transformer-VAE 生成协调的眼、颈与眼睑运动。在 2 秒基准上关节误差最低。
@misc{hu2026eye,
title = {{Eye, Robot: Learning Expressive Gaze Dynamics via a VLM-Guided Transformer-VAE}},
author = {Yuhang Hu and Junlin Lyu and Yuan Fang and Xiaoying Shen and Lechen Zhang and Jiong Lin and Judah Goldfeder and Steven Tian and Hod Lipson},
year = {2026},
eprint = {2610.19635},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.19635},
}
Discovers long-horizon, contact-rich humanoid loco-manipulation from scratch: LLM-guided evolutionary search over interaction sequences plus kinodynamic trajectory optimization, then RL tracking on a real humanoid.从零发现长时域、多接触的人形机器人移动操作动作：由大模型引导的进化搜索交互序列，配合运动动力学轨迹优化，再训练强化学习跟踪策略部署到真实人形机器人。
@misc{taouil2026motiondisco,
title = {{MotionDisco: Motion Discovery for Extreme Humanoid Loco-Manipulation}},
author = {Ilyass Taouil and Michal Ciebelski and Shafeef Omar and Haizhou Zhao and Angela Dai and Aaron M. Johnson and Majid Khadiv},
year = {2026},
eprint = {2606.06139},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.06139},
}
Teleop-free humanoid interaction: a fine-tuned VLM generates human whole-body motion from an egocentric image and instruction, retargeted and run by an interaction-aware tracker on a Unitree G1.无需遥操作的人形交互：微调后的 VLM 根据第一人称图像和语言指令生成人体全身运动 token，重定向到机器人后由注重交互的跟踪策略在 Unitree G1 上执行。
@misc{yang2026zerowbc,
title = {{ZeroWBC: Learning Natural Whole-Body Humanoid Interaction from Human Egocentric Data}},
author = {Haoran Yang and Jiacheng Bao and Yucheng Xin and Haoming Song and Yuyang Tian and Bin Zhao and Dong Wang and Xuelong Li},
year = {2026},
eprint = {2603.09170},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.09170},
}
One RL policy for legs, arm and gripper that adds a tactile latent from compliant grippers to proprioception, regulating grasp force and preventing slip. Zero-shot on a Unitree Go2 with a WidowX 250 arm.单一强化学习策略同时控制腿、机械臂与夹爪，把柔顺夹爪的触觉潜表示与本体感觉结合，调节握力并防止物体滑落。零样本部署在装有 WidowX 250 机械臂的 Unitree Go2 上。
@misc{hu2026tac,
title = {{TAC-LOCO: Unified Whole-Body Control for Quadrupedal TACtile-Informed LOCO-Manipulation}},
author = {Muqun Hu and Yuhao Zhou and Kabir Ray Malik and Chi Lin and Won Suk Lee and Yu She and Yan Gu},
year = {2026},
eprint = {2607.10132},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.10132},
}
Box-carrying humanoid policy conditioned on a desired root trajectory, with reference motion used only in the reward; a persistent object estimator fuses detection with a digital twin to spot slips and re-grasp.人形机器人搬箱策略以期望的根轨迹为条件，参考动作只用于奖励；持续物体估计模块融合实时检测与数字孪生，发现滑落后自动重新抓取。
@misc{lin2026pro,
title = {{Pro-HOI: Perceptive Root-guided Humanoid-Object Interaction}},
author = {Yuhang Lin and Jiyuan Shi and Dewei Wang and Jipeng Kong and Yong Liu and Chenjia Bai and Xuelong Li},
year = {2026},
eprint = {2603.01126},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.01126},
}
Uses 'contact flow' (key body trajectories plus binary contact signals) as the interface between a skill-tracking policy and a planner that generates future contact flows; releases a MoCap humanoid-object dataset.以“接触流”（关键身体轨迹 + 二值接触信号）作为接口：底层策略跟踪统一的移动操作技能库，高层模块生成未来接触流来串联技能；并发布基于动捕的人-物交互数据集。
@misc{yu2026omnicontact,
title = {{OmniContact: Chaining Meta-Skills via Contact Flow for Generalizable Humanoid Loco-Manipulation}},
author = {Runyi Yu and Xiaoyi Lin and Ji Ma and Yinhuai Wang and Koukou Luo and Jiahao Ji and Huayi Wang and Wenjia Wang and Runhan Zhang and Ping Tan and Ting Wu and Ruoli Dai and Qifeng Chen and Lei Han},
year = {2026},
eprint = {2606.26201},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.26201},
}
Three-stage hierarchical RL for humanoid badminton: sparse human strokes are augmented into a latent skill space, and a high-level planner composes skills online from the observed shuttle.三阶段分层强化学习实现人形机器人打羽毛球：把少量人类击球动作扩充为连续的潜技能空间，高层规划器再根据观测到的羽毛球在线组合技能。
@misc{cui2026humanoid,
title = {{Humanoid Badminton: Learning Dynamic Racket Skills from Limited Human Motion Data}},
author = {Jingzhi Cui and Zhexiong Wang and Bangjie Xu and Pengyu Zhao and Youyuan Li and Zhi Su and Peng Ren and Mengdi Xu and Chao Yu and Yi Wu and Luyang Wang and Zhongyu Li},
year = {2026},
eprint = {2609.31840},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.31840},
}
Cross-embodiment humanoid whole-body controller: shared human-motion command tokens and a causal transformer trained on mixed multi-robot rollouts, with small robot-specific modules. Nine sim bodies, four real robots.跨本体的人形全身控制基础模型：共享的人体运动指令 token 与在多机器人混合数据上训练的因果 Transformer，加上轻量的机器人专属模块。覆盖 9 种仿真本体与 4 台真机。
@misc{zhang2026x,
title = {{X-WBC: A Cross-Embodiment Foundation Model for Humanoid Whole-Body Control}},
author = {Juntong Zhang and Chun Gu and Li Zhang},
year = {2026},
eprint = {2609.15213},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.15213},
}
POT-VLA keeps role-indexed 3D object records from RGB-D as tokens for the whole-body action expert and for checking whether actions succeeded. On a Unitree G1: 71/80 successes vs 39/80 for GR00T-N1.7.POT-VLA 从 RGB-D 维护按角色索引的 3D 物体记录，既作为全身动作专家的输入 token，也用于检验动作是否达成。在 Unitree G1 上成功 71/80 次，GR00T-N1.7 基线为 39/80。
@misc{ren2026closing,
title = {{Closing the Loop in Humanoid VLA: Persistent 3D Object Tokens for Verifiable Loco-Manipulation}},
author = {Peng Ren and Haoyang Ge and Jiang Zhao and Cong Huang and Yukun Shi and Pei Chi and Kai Chen},
year = {2026},
eprint = {2607.18016},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.18016},
}
Humanoid brachiation with passive wrist hooks: biomimetic keyframes make release-swing-capture reachable early in training, and a recurrent estimator gives position/contact latents. Continuous bar traversal on hardware.带被动腕钩的人形机器人臂荡移动：仿生关键帧让“松手-摆荡-抓杆”在训练早期就能被探索到，循环网络估计位置与接触潜变量。真机实现连续过杆。
@misc{xiong2026swingbot,
title = {{SwingBot: Learning Whole-Body Brachiation for Humanoid Robots}},
author = {Yujie Xiong and Peng Zhai and Taixian Hou and Quancheng Qian and Cunwang Liu and Kangmai Hu and Long Yang and Zhiyan Dong and Lihua Zhang},
year = {2026},
eprint = {2609.10283},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.10283},
}
Carries several free-rolling balls on a wheel-legged robot's back with no fences: per-frame Deep Sets respect that ball order can change every frame, which standard history encoders miss and exploit as a shortcut.在轮腿机器人背上无围栏地同时运送多个自由滚动的球：逐帧 Deep Sets 编码器考虑到球的顺序每帧都可能变化，而常规历史拼接编码器忽略这一点并把槽位当作捷径。
@misc{chen2026s,
title = {{S2M-Trek: From Single to Multi-Sphere Transport via Per-Frame Deep Sets on a Wheel-Legged Robot}},
author = {Zong Chen and Xuebin Li and Jinpeng Xiao and Shaoyang Li and Ben Liu and Min Li and Zhouping Yin and Yiqun Li},
year = {2026},
eprint = {2606.01332},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.01332},
}
Fault-tolerant whole-body control for legged manipulators: a fault estimator finds failed leg joints from proprioception, and a posture module turns arm-driven base requests into safe ones.面向腿足机械臂的容错全身控制：故障估计器从下肢本体感觉历史中识别失效关节，姿态适配模块把机械臂策略提出的底盘姿态请求转为安全可执行的指令。
@misc{zhong2026ft,
title = {{FT-WBC: Learning Fault-Tolerant Whole-Body Control for Legged Loco-Manipulation}},
author = {Yudong Zhong and Pengfei Mai and Sikai Guo and Jiahang Cao and Zhihai Bi and Qiuyue Liu and Ziyan Feng and Jinni Zhou and Jun Ma},
year = {2026},
eprint = {2606.24466},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.24466},
}
FAST: a general humanoid motion tracker with a small residual adaptation policy under orthogonality and KL constraints for new motions without forgetting, plus center-of-mass-aware terms for balance.FAST：通用人形运动跟踪控制器，在正交与 KL 约束下学习轻量残差策略来快速适配新动作且不遗忘，并加入质心相关观测与目标提升平衡。
@misc{wang2026general,
title = {{General Humanoid Whole-Body Control via Pretraining and Fast Adaptation}},
author = {Zepeng Wang and Jiangxing Wang and Shiqing Yao and Yu Zhang and Ziluo Ding and Ming Yang and Yuxuan Wang and Haobin Jiang and Chao Ma and Xiaochuan Shi and Zongqing Lu},
year = {2026},
eprint = {2602.11929},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.11929},
}
A latent motion prior distilled from human soccer MoCap lets one skill-conditioned policy move, dribble, trap and kick with human-like motion; transfers zero-shot to a Unitree G1.从人类足球动捕数据蒸馏出潜在运动先验，使单一技能条件策略能完成移动、带球、停球和射门，动作自然流畅；零样本迁移到 Unitree G1。
@misc{shao2026playing,
title = {{Playing Soccer Like Humans Do: Learning Versatile and Agile Humanoid Soccer via Latent Motion Priors}},
author = {Haoyu Shao and Yunshen Wang and Jingze Zhang and Shaohang Zhu and Yong-Lu Li and Tianzhu Zhang and Baoxiong Jia and Siyuan Huang},
year = {2026},
eprint = {2611.06564},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06564},
}
Uni-Mo generates quadruped motions with an LLM plus a video diffusion model and lifts them to 3D references for tracking policies; the Quad-Imaginarium dataset has 7,488 motions, 96.7% of 392 sampled ran on a Go2.Uni-Mo 用大语言模型和视频扩散模型生成四足动作，并提升为 3D 参考轨迹训练跟踪策略；Quad-Imaginarium 数据集含 7488 个动作，抽样的 392 个在 Go2 真机上 96.7% 成功执行。
@misc{liu2026unleashing,
title = {{Unleashing Infinite Motion: Scaling Expressive Quadrupedal Motion via Generative Video Priors}},
author = {Youzhi Liu and Li Gao and Yifei Qian and Liu Liu and Yang Cai and Wenbin Tang},
year = {2026},
eprint = {2606.28237},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.28237},
}
MorFiC finds that a shared critic averages incompatible values across robot bodies and fixes it with multiplicative morphology conditioning; trained on one robot, it transfers zero-shot to seven quadrupeds.MorFiC 指出共享价值网络会把不同机身的价值目标混为一谈，并用乘性形态条件化修正该问题；只在一台机器人上训练即可零样本迁移到七种四足机器人。
QWM conditions one latent world model on scale-invariant morphology features and trains policies in imagination; unlike a model-free policy, it transfers zero-shot to unseen quadruped morphologies.QWM 以尺度无关的形态特征作为条件训练单一潜空间世界模型，并在想象中训练策略；与无模型策略不同，它能零样本迁移到未见过的四足形态。
@misc{danesh2026morphology,
title = {{Morphology-Conditioned World Model for Cross-Embodiment Quadrupedal Locomotion}},
author = {Mohamad H. Danesh and Chenhao Li and Amin Abyaneh and Anas Houssaini and Kirsty Ellis and Glen Berseth and Marco Hutter and Hsiu-Chin Lin},
year = {2026},
eprint = {2604.08780},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.08780},
}
Splits quadruped training into rewards for the task, constraints for limits and energy minimization for gait, with no gait priors; reports 56% lower cost of transport and 96% fewer limit violations, on a Go2.将四足训练拆分为：奖励定义任务、约束保证运行限制、能耗最小化决定步态，且不使用任何步态先验；报告运输代价降低 56%、越限减少 96%，并在 Go2 上部署。
@misc{kordos2026locomposition,
title = {{LoComposition: Terrain-Adaptive Energy-Efficient Quadruped Locomotion without Gait Priors}},
author = {Loukas Kordos and Leonard T. Franz and Simon Rappenecker and Oliver Hausdoerfer and Angela P. Schoellig and Pavel Kolev and Georg Martius},
year = {2026},
eprint = {2606.15896},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.15896},
}
HumoSlope trains humanoid walking on steep slopes with a slope-aware ZMP balance prior and a gait adapter that avoids crouching; the blind policy walks outdoor grass slopes up to 32.1 degrees.HumoSlope 用基于斜坡支撑面的 ZMP 平衡先验和避免蹲伏步态的适配器训练人形机器人陡坡行走；纯本体感知策略可在户外草坡上行走，坡度最高 32.1 度。
@misc{chen2026physics,
title = {{Physics-Guided Biomechanical Gait Adaptation for Humanoid Locomotion on Extreme Sloped Terrains}},
author = {Xuanyu Chen and Mohan Liu and Dengchen Mei and Zhihao Gu and Haitian Zhang and Kaimin Mao and Haiyue Zhu and Shijun Yan and Lin Wang},
year = {2026},
eprint = {2607.07830},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.07830},
}
RPL distills terrain-specific experts into one multi-depth-camera transformer for multi-directional humanoid walking with payloads, with a ray-casting depth renderer about 5x faster than existing pipelines.RPL 将各地形专家策略蒸馏为一个多深度相机 Transformer 策略，实现可负重的人形多方向行走，并配套比现有管线快约 5 倍的光线投射深度渲染器。
@misc{zhang2026rpl,
title = {{RPL: Learning Robust Humanoid Perceptive Locomotion on Challenging Terrains}},
author = {Yuanhang Zhang and Younggyo Seo and Juyue Chen and Yifu Yuan and Koushil Sreenath and Pieter Abbeel and Carmelo Sferrazza and Karen Liu and Rocky Duan and Guanya Shi},
year = {2026},
eprint = {2602.03002},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.03002},
}
Humanoid parkour over sparse footholds from onboard depth: a saliency-guided gated memory keeps useful depth features across frames, and an alternation loss enforces left-right foot placement.基于机载深度相机的人形跑酷，跨越稀疏落脚点：显著性引导的门控记忆跨帧保留有用的深度特征，交替损失约束左右脚交替落步。
@misc{lee2026echo,
title = {{Echo in the Steps: Learning Perceptive Humanoid Parkour with Gated Memory}},
author = {Ming-Ju Lee and Zizhuo Wang and Shaoting Zhu and Haozhe Lou and Hang Zhao and Yiming Li},
year = {2026},
eprint = {2609.28960},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.28960},
}
MeshMimic reconstructs both human motion and terrain geometry from monocular video, then retargets with contact constraints, so humanoids learn motion-terrain interactions without MoCap.MeshMimic 从单目视频中同时重建人体运动与地形几何，再以接触约束进行重定向，使人形机器人无需动捕即可学习动作与地形的交互。
@misc{zhang2026meshmimic,
title = {{MeshMimic: Geometry-Aware Humanoid Motion Learning through 3D Scene Reconstruction}},
author = {Qiang Zhang and Jiahao Ma and Peiran Liu and Shuai Shi and Zeran Su and Zifan Wang and Jingkai Sun and Wei Cui and Jialin Yu and Gang Han and Wen Zhao and Pihai Sun and Kangning Yin and Jiaxu Wang and Jiahang Cao and Lingfeng Zhang and Hao Cheng and Xiaoshuai Hao and Yiding Ji and Junwei Liang and Jian Tang and Renjing Xu and Yijie Guo},
year = {2026},
eprint = {2602.15733},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.15733},
}
Perceptive BFM grounds human motion references in the robot's local terrain: synthesized terrain-conformal references train a teacher, and a student tracker adds terrain corrections only when needed.Perceptive BFM 让人类动作参考适配机器人所处的局部地形：用合成的贴合地形参考训练教师策略，学生跟踪器仅在必要时加入地形修正。
@misc{wang2026perceptive,
title = {{Perceptive Behavior Foundation Model: Adapting Human Motion Priors to Robot-Centric Terrain}},
author = {Zifan Wang and Yizhao Li and Teli Ma and Qiang Zhang and Yudong Fan and Hao Xu and Shuo Yang and Junwei Liang},
year = {2026},
eprint = {2606.08059},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.08059},
}
MARCH guides RL with a safe reference from simplified models and a control-Lyapunov reward, then distills to a vision policy; a Unitree G1 crosses sparse footholds with lateral constraints.MARCH 用简化模型生成的安全参考轨迹和控制李雅普诺夫奖励引导强化学习，再蒸馏为视觉策略；Unitree G1 可通过带侧向约束的稀疏落脚点。
@misc{crismariu2026march,
title = {{MARCH: Model-Assisted Reinforcement Learning for the Perceptive Control of Humanoids over Sparse Footholds}},
author = {Codrin Crismariu and Ryan K. Cosner},
year = {2026},
eprint = {2606.10288},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.10288},
}
CAP keeps one humanoid policy working as depth degrades: a world-model encoder learns to denoise corrupted depth, alongside a proprioceptive encoder; tested on a G1 indoors and outdoors.CAP 让单一人形策略在深度感知退化时仍能工作：世界模型编码器学习对受损深度去噪，并配合本体感知编码器；在 G1 上进行了室内外测试。
@misc{chen2026cap,
title = {{CAP: Continuously Adaptive Perception-Blind Humanoid Locomotion via Learned Denoising}},
author = {Hongjin Chen and Zijun Xu and Shihao Ma and Yi Zhao and Xilai Liu and Ke Ma and Wei Zhang and Chunyang Xie and Pengfei Li and Jieru Zhao and Wenchao Ding},
year = {2026},
eprint = {2609.11553},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.11553},
}
RoboDreamer distills a humanoid teacher into a Mamba student that sees randomly masked recent observations, so it infers missing state from history; tested in IsaacLab, MuJoCo and on a G1.RoboDreamer 将人形教师策略蒸馏为 Mamba 学生策略，训练时随机遮蔽最近观测，使其从历史推断缺失状态；在 IsaacLab、MuJoCo 和 G1 上验证。
@misc{li2026robodreamer,
title = {{RoboDreamer: Anticipatory Humanoid Locomotion with Predictive State-Space Models}},
author = {Zhe Li and Yangyang Wei and Xichen Yuan and Zhenzhe Zhang and Weihao Yuan and Shanghang Zhang and Jianfei Yang},
year = {2026},
eprint = {2609.07096},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.07096},
}
ADAPT feeds a humanoid policy an analytical whole-body disturbance observer that estimates external force/torque from robot dynamics without F/T sensors; more robust to pushes and payloads on a G1.ADAPT 为人形策略提供解析式全身扰动观测器，无需力/力矩传感器即可由机器人动力学在线估计外力与力矩；在 G1 上对推搡和负载更鲁棒。
@misc{lyu2026adapt,
title = {{ADAPT: Analytical Disturbance-Aware Policy Training for Humanoid Locomotion}},
author = {Bofan Lyu and Jindou Jia and Kuangji Zuo and Yanshuo Lu and Shijia Han and Gen Li and Boyu Ma and Jingliang Li and Geng Li and Jianfei Yang},
year = {2026},
eprint = {2606.16542},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.16542},
}
Walk Like This: give a style clip and a joystick velocity; a style-conditioned autoregressive diffusion model generates the reference and an RL controller tracks it, in simulation and on a robot.Walk Like This：给定风格参考片段和摇杆速度指令，风格条件自回归扩散模型生成运动参考，再由强化学习控制器跟踪，在仿真角色和真实机器人上均可运行。
@misc{aurand2026walk,
title = {{Walk Like This: Stylized Locomotion for Physics-Based Characters through Example-Driven Diffusion}},
author = {Joshua Aurand and Agon Serifi and Sammy Christen and Ruben Grandia and Espen Knoop and Marco Hutter and Moritz Bächer},
year = {2026},
eprint = {2610.18147},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18147},
}
Starts RL with stiff joints and anneals stiffness to the identified value as episodes lengthen, which enlarges the viable set; enables a quadruped-to-handstand transition on Spot that otherwise never succeeds.强化学习训练初期使用较高的关节刚度，随回合变长逐步退火到辨识值，从而扩大可行状态集；使原本无法学会的 Spot 由四足站立转为倒立的动作得以完成并上真机。
@misc{chakraborty2026actuator,
title = {{Actuator Dynamics Curricula for Narrow-Viability Tasks in Legged Robot Learning}},
author = {Kousheek Chakraborty and Chandan K. Rajendra and Ayham Alharbat and Abeje Y. Mersha},
year = {2026},
eprint = {2609.09492},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.09492},
}
TACL builds curricula from trajectory tasks on unstructured terrain maps: an evaluator learns difficulty for the current policy and a sampler proposes the next tasks; +18.5% on the hardest terrain vs handcrafted.TACL 直接在非结构化地形图上以轨迹任务构建课程：评估器学习当前策略的任务难度，采样器据此提出下一批任务；在最难地形上比人工课程成功率高 18.5%。
@misc{liu2026trajectory,
title = {{Trajectory-Level Automatic Curriculum Learning for Legged Locomotion on Unstructured Terrain}},
author = {Rocky Liu and Tengyu Liu and Baoxiong Jia and Fangwei Zhong and Xinyi Tong and Hongzhao Xie and Siyuan Huang},
year = {2026},
eprint = {2608.16164},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2608.16164},
}
SHOES adapts a frozen sim-trained quadruped policy to footwear by fine-tuning only small identity-initialized modulation modules on the robot, with a workspace-bounded sampler; Go2 tested with shoes.SHOES 冻结仿真训练的四足策略，仅在真机上微调以恒等初始化的轻量调制模块来适应穿鞋带来的形态变化，并用限定工作空间的采样器保证安全；在穿鞋的 Go2 上测试。
Studies locomotion policies via effective rank of the network Jacobian conditioned on gait phase; normalization and residual connections help, giving up to 2x lower joint jitter on a real Spot.以步态相位为条件分析运动策略网络雅可比的有效秩；发现归一化与残差连接更健康，在真实 Spot 上关节抖动最多降低 2 倍。
@misc{tommaselli2026mind,
title = {{Mind the Phase: Effective Rank and Representation Health in Legged Locomotion}},
author = {Felipe Tommaselli and Ricardo V. Godoy and Thiago H. Segreto and Juliano D. Negri and Marcelo Becker},
year = {2026},
eprint = {2609.06958},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.06958},
}
A full-scale humanoid sprints above 7 m/s and corners with up to 25 degrees of trunk lean, using causal attention over feedback, a heading-direction reward and skill-conditioned motion-prior discriminators.全尺寸人形机器人以超过 7 m/s 冲刺，并以最高 25 度躯干侧倾高速过弯；方法包括对反馈的因果注意力、航向奖励以及按技能条件化的运动先验判别器。
@misc{huang2026push,
title = {{Push the Limit: Learning High-Speed Sprinting and Aggressive Cornering on a Full-Scale Humanoid}},
author = {Huaxing Huang and Ruidong Li and Jinchao Han and Ziyang Dong and Wenhao Cui},
year = {2026},
eprint = {2610.14840},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.14840},
}
Splits humanoid whole-body control: inverse-dynamics feedforward plus PD give tracking torques, an RL residual adds only balance corrections; one balance policy carries across motion datasets, payloads and contact.拆分人形全身控制：逆动力学前馈加 PD 反馈提供跟踪力矩，强化学习残差只负责平衡修正；同一平衡策略可跨动作数据集、负载与物理交互泛化，仿真与真机均有验证。
@misc{wei2026dkwbc,
title = {{DKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body Control}},
author = {Dingkun Wei and Zehong Shen and Huaijin Pi and Xueyang Wang and Xing Zhu and Hujun Bao and Yujun Shen and Xiaowei Zhou},
year = {2026},
eprint = {2611.09605},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09605},
}
Follow-up to SUGAR: synthesizes robot and object reference motions from reusable task templates, then trains visual policies with one mimic reward and teacher distillation; 91.0% mean sim success on five tasks.SUGAR 的后续：用可复用的任务模板合成机器人与物体参考动作，再以统一的模仿奖励和教师蒸馏训练视觉策略；五项任务仿真平均成功率 91.0%，并完成真机迁移。
@misc{wu2026sugarv,
title = {{SUGARv2: Learning Diverse Visual Humanoid Loco-Manipulation Skills from Synthesized Interaction References}},
author = {Tianshu Wu and Tianxiang Gui and Yue Chen and Tingxu Liu and Ziye Peng and Guangjie Yang and Xiangqi Kong and Ye Shi and Jingya Wang and Hao Dong},
year = {2026},
eprint = {2611.09266},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09266},
}
HumanoidPose recovers a humanoid's full-body joint configuration from video alone — useful when onboard state is unavailable, and for turning online robot videos into training data.HumanoidPose 仅凭视频恢复人形机器人的全身关节构型：适用于拿不到机载状态的场景，也能把网上的机器人视频变成训练数据。
@misc{jiang2026vision,
title = {{Vision as Proprioception: Estimating Humanoid Robot Pose from Video}},
author = {Jiaxi Jiang and Matthias Heyrman and René Zurbrügg and Marco Hutter and Christian Holz},
year = {2026},
eprint = {2610.25288},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25288},
}
@misc{wang2026generalist,
title = {{A generalist policy for muscle-driven control across individual musculoskeletal models}},
author = {Bozhao Wang and Yinglei Zhu and Zhengqing Li and Zhiyuan Zhang and Weiming Kang and Yiheng He and XinJun Liu and Senhao Hou and Huichan Zhao and Xiaoqiang Tang},
year = {2026},
eprint = {2610.36535},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36535},
}
A policy predicts each action and how long to hold it. Reports matching high-frequency control at under half the rate on an RC car and Unitree Go1, plus simulated VLA inference.策略同时预测动作及其保持时长。报告在遥控车、Unitree Go1 和仿真 VLA 上以不到一半的控制频率达到高频控制的性能。
@misc{sukhija2025tarc,
title = {{TARC: Time-Adaptive Robotic Control}},
author = {Arnav Sukhija and Lenart Treven and Jin Cheng and Florian Dörfler and Stelian Coros and Andreas Krause},
year = {2025},
eprint = {2510.23176},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.23176},
}
Success Guided Sampling focuses massively parallel RL on task configurations at the edge of the policy's ability. With up to a million envs it solves quadruped terrain and assembly tasks; assembly transfers zero-shot.成功引导采样把大规模并行强化学习集中在策略能力边界附近的任务配置上。用最多约一百万个并行环境解决四足多地形与装配任务，装配策略零样本迁移到真机。
@misc{zhang2026balanced,
title = {{A Balanced Data Diet: Addressing the Exploration Bottleneck in Mega-Scale RL for Robot Control}},
author = {Octi Zhang and Mateo Guaman Castro and Patrick Yin and Ignacio Dagnino and Abhishek Gupta and Rosario Scalise and Byron Boots},
year = {2026},
eprint = {2610.12465},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.12465},
}
100 h of wearable-captured human loco-manipulation (mostly professional work) aligning egocentric video, whole-body and finger motion, touch and step-level text; reports 71.9% lower body error than baselines.用可穿戴设备采集的 100 小时人类移动操作数据，侧重职业场景，对齐第一视角视频、全身与手指动作、触觉和步骤级文字描述；其补全全身动作的方法报告身体关键点误差比最强基线低 71.9%。
@misc{zhou2026egoexplorer,
title = {{EgoExplorer: A Vision-Motion-Touch Observatory for Human Loco-Manipulation In Situ}},
author = {Yunsong Zhou and Xuekun Jiang and Haiqi Shen and Hangxu Liu and Chaojian Shi and Yujie Liu and Jingyuan Li and Haoran Yang and Jiafei Cao and Dongzhan Zhou and Dong Wang and Yao Mu and Chunhua Shen and Weinan Zhang},
year = {2026},
eprint = {2610.18872},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18872},
}
Runs fourteen curricula under one protocol on five quadrupeds in MuJoCo MJX: no method tops every robot and level set, and methods with easy training levels lead on in-distribution random terrain.在 MuJoCo MJX 中用统一协议在五款四足上比较 14 种课程方法：没有方法在所有机器人和地形集上都领先，而训练关卡偏简单的方法在分布内随机地形上表现靠前。
@misc{gilewicz2026is,
title = {{Is It the Curriculum or Just the Easy Start? Benchmarking Unsupervised Environment Design on Quadrupeds}},
author = {Jakub Gilewicz and Dominik Galus and Piotr Wojdas},
year = {2026},
eprint = {2610.40364},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2610.40364},
}
200+ hours from 14 people: six head cameras, wrist vision and IMU, plus 120 Hz optical full-body mocap, covering walking, terrain and carrying. Benchmarks body reconstruction, foot contact and humanoid retargeting.14 名参与者、200 多小时数据：六个头戴相机、腕部视觉与惯性传感，加 120 Hz 光学全身动捕，覆盖行走、地形穿越与搬运等。提供全身重建、足部接触估计与人形重定向三项基准。
@misc{deshmukh2026robomocap,
title = {{RoboMoCap: A Large-Scale Whole-Body Motion Capture Dataset with Egocentric Vision for Robot Learning}},
author = {Mayur Deshmukh and Aaron Tung and Junyi Li and Nikash Bhardwaj and Ademi Adeniji and Vincent Liu and Michael Cho and Jiafei Duan},
year = {2026},
eprint = {2611.10994},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10994},
}
One operator drives a Unitree Go2 + Piper arm with a leader arm and two handheld controllers, with sensorless force feedback; records a unified 11-D action reused at deployment (shown by fine-tuning MolmoAct2).单人即可同时操控 Unitree Go2 四足与 Piper 机械臂：主端机械臂加两个手柄，无需力传感器即可力反馈；ROS 2 管线记录与部署共用的 11 维统一动作，并通过微调 MolmoAct2 验证从采集到部署的全流程。
@misc{langella2026locopilot,
title = {{LocoPilot: Single-Operator Bilateral Teleoperation for Loco-Manipulation Policy Learning}},
author = {Antonio Langella and Giovanni Minelli and Giovanni B. Dessy and Lorenzo Amatucci and Giulio Turrisi and Victor Barasuol and Claudio Semini},
year = {2026},
eprint = {2610.27998},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27998},
}
A humanoid arm whose force, speed and endurance targets come from human upper-limb measurements; reports a 20 kg peak payload (2.25x its own mass) and 20.6 m/s endpoint speed.人形手臂，其力量、速度和耐力指标取自人类上肢测量数据；报告峰值负载 20 公斤（自重的 2.25 倍），末端速度 20.6 米每秒。
@misc{kim2026raipal,
title = {{RAIPAL: A Powerful and Agile Humanoid Arm for Human Tasks}},
author = {Donggeon Kim and Hyungho Chris Choi and Donghyuk Choi and Jeongsoo Park and Choongin Lee and Chan Park and Jemin Hwangbo},
year = {2026},
eprint = {2611.04334},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04334},
}
Being-0: a humanoid agent that coordinates a foundation model, a vision-language model and modular skills to limit compounding errors on long tasks.Being-0：协调基础模型、视觉语言模型与模块化技能的人形智能体，以减少长任务中的误差累积。
@misc{yuan2025being,
title = {{Being-0: A Humanoid Robotic Agent with Vision-Language Models and Modular Skills}},
author = {Haoqi Yuan and Yu Bai and Yuhui Fu and Bohan Zhou and Yicheng Feng and Xinrun Xu and Yi Zhan and Börje F. Karlsson and Zongqing Lu},
year = {2025},
eprint = {2503.12533},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2503.12533},
}
@misc{xue2025leverb,
title = {{LeVERB: Humanoid Whole-Body Control with Latent Vision-Language Instruction}},
author = {Haoru Xue and Xiaoyu Huang and Dantong Niu and Qiayuan Liao and Thomas Kragerud and Jan Tommy Gravdahl and Xue Bin Peng and Guanya Shi and Trevor Darrell and Koushil Sreenath and Shankar Sastry},
year = {2025},
eprint = {2506.13751},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.13751},
}
A high-density tactile sensor across a quadruped's back lets it carry unsecured cylinders over long distances, a task that usually needs mounting hardware.在四足机器人背部铺设高密度触觉传感器，无需固定装置即可长距离运输未固定的圆柱物体。
@misc{lin2025locotouch,
title = {{LocoTouch: Learning Dynamic Quadrupedal Transport with Tactile Sensing}},
author = {Changyi Lin and Yuxin Ray Song and Boda Huo and Mingyang Yu and Yikai Wang and Shiqi Liu and Yuxiang Yang and Wenhao Yu and Tingnan Zhang and Jie Tan and Yiyue Luo and Ding Zhao},
year = {2025},
eprint = {2505.23175},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.23175},
}
Latent VLA that learns from cheap action-free egocentric video, paired with an RL controller for advancing, turning and squatting; reports a 21.3% improvement over a baseline on the AgiBot X2.潜空间 VLA 从低成本、无动作标注的第一视角视频学习，再配合面向前进、转向、下蹲的强化学习控制器；在智元 X2 上报告比基线提升 21.3%。
@misc{jiang2025wholebodyvla,
title = {{WholeBodyVLA: Towards Unified Latent VLA for Whole-Body Loco-Manipulation Control}},
author = {Haoran Jiang and Jin Chen and Qingwen Bu and Li Chen and Modi Shi and Yanjie Zhang and Delong Li and Chuanzhe Suo and Chuang Wang and Zhihui Peng and Hongyang Li},
year = {2025},
eprint = {2512.11047},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2512.11047},
}
One policy learns full-body interaction with varied objects from hours of imperfect MoCap: subject-specific teachers refine the data first, then scale up.单一策略从数小时不完美动捕数据学习全身人-物交互：先用个体教师策略精修数据，再扩展规模。
@misc{xu2025intermimic,
title = {{InterMimic: Towards Universal Whole-Body Control for Physics-Based Human-Object Interactions}},
author = {Sirui Xu and Hung Yu Ling and Yu-Xiong Wang and Liang-Yan Gui},
year = {2025},
eprint = {2502.20390},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2502.20390},
}
Trains one unified policy to track diverse human motions on a real humanoid, instead of one policy per skill, using adaptive sampling and a motion mixture-of-experts.训练单一统一策略，让真实人形机器人跟踪多样的人体动作，而非每个技能一个策略；核心是自适应采样和动作混合专家结构。
@misc{chen2025gmt,
title = {{GMT: General Motion Tracking for Humanoid Whole-Body Control}},
author = {Zixuan Chen and Mazeyu Ji and Xuxin Cheng and Xuanbin Peng and Xue Bin Peng and Xiaolong Wang},
year = {2025},
eprint = {2506.14770},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.14770},
}
@misc{ze2025twist,
title = {{TWIST: Teleoperated Whole-Body Imitation System}},
author = {Yanjie Ze and Zixuan Chen and João Pedro Araújo and Zi-ang Cao and Xue Bin Peng and Jiajun Wu and C. Karen Liu},
year = {2025},
eprint = {2505.02833},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.02833},
}
HugWBC: one controller covering running, jumping and hopping, with adjustable gait parameters such as frequency and foot height.HugWBC：用一个控制器覆盖跑、跳、单脚跳等动作，并可调节步频、抬脚高度等步态参数。
@misc{xue2025unified,
title = {{A Unified and General Humanoid Whole-Body Controller for Versatile Locomotion}},
author = {Yufei Xue and Wentao Dong and Minghuan Liu and Weinan Zhang and Jiangmiao Pang},
year = {2025},
eprint = {2502.03206},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.03206},
}
AMO joins sim-to-real RL with trajectory optimization for adaptive whole-body control, enabling large-workspace tasks such as picking objects off the ground.AMO 将 sim-to-real 强化学习与轨迹优化结合，实现自适应全身控制，支持从地面拾物等大工作空间任务。
@misc{li2025amo,
title = {{AMO: Adaptive Motion Optimization for Hyper-Dexterous Humanoid Whole-Body Control}},
author = {Jialong Li and Xuxin Cheng and Tianshu Huang and Shiqi Yang and Ri-Zhao Qiu and Xiaolong Wang},
year = {2025},
eprint = {2505.03738},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.03738},
}
A promptable humanoid foundation model: a shared latent space embeds motions, goals and rewards, so one policy can be prompted for different tasks on a real robot.可提示的人形基础模型：用共享隐空间嵌入动作、目标和奖励，同一策略可在真机上按提示执行不同任务。
@misc{li2025bfm,
title = {{BFM-Zero: A Promptable Behavioral Foundation Model for Humanoid Control Using Unsupervised Reinforcement Learning}},
author = {Yitang Li and Zhengyi Luo and Tonghe Zhang and Cunxi Dai and Anssi Kanervisto and Andrea Tirinzoni and Haoyang Weng and Kris Kitani and Mateusz Guzek and Ahmed Touati and Alessandro Lazaric and Matteo Pirotta and Guanya Shi},
year = {2025},
eprint = {2511.04131},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.04131},
}
Sim-to-real visual loco-manipulation with a task-agnostic keypoint tracker below a policy that generates motion, so egocentric vision replaces external motion capture.视觉移动操作 sim-to-real 框架：底层是任务无关的关键点跟踪器，上层策略生成动作，以第一视角视觉取代外部动捕。
@misc{yin2025visualmimic,
title = {{VisualMimic: Visual Humanoid Loco-Manipulation via Motion Tracking and Generation}},
author = {Shaofeng Yin and Yanjie Ze and Hong-Xing Yu and C. Karen Liu and Jiajun Wu},
year = {2025},
eprint = {2509.20322},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.20322},
}
Two-stage humanoid loco-manipulation: a general motion-tracking policy trained on human data, then a residual policy adds object-aware precision.两阶段人形移动操作：先用人体数据训练通用动作跟踪策略，再用残差策略补上感知物体的精确控制。
@misc{zhao2025resmimic,
title = {{ResMimic: From General Motion Tracking to Humanoid Whole-body Loco-Manipulation via Residual Learning}},
author = {Siheng Zhao and Yanjie Ze and Yue Wang and C. Karen Liu and Pieter Abbeel and Guanya Shi and Rocky Duan},
year = {2025},
eprint = {2510.05070},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.05070},
}
Dual-agent RL for force-adaptive loco-manipulation: a lower-body agent keeps walking stable while an upper-body agent tracks the end effector under external forces.面向力自适应移动操作的双智能体强化学习：下半身智能体保持行走稳定，上半身智能体在外力下跟踪末端位置。
@misc{zhang2025falcon,
title = {{FALCON: Learning Force-Adaptive Humanoid Loco-Manipulation}},
author = {Yuanhang Zhang and Yifu Yuan and Prajwal Gurunath and Ishita Gupta and Shayegan Omidshafiei and Ali-akbar Agha-mohammadi and Marcell Vazquez-Chanlatte and Liam Pedersen and Tairan He and Guanya Shi},
year = {2025},
eprint = {2505.06776},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.06776},
}
Tackles balance-heavy skills such as one-leg stands and high kicks, where motion tracking usually fails because of reference errors. Oral at CoRL 2025.攻克单脚站立、高踢腿等强平衡技能，这类动作常因参考动作误差导致跟踪失败。CoRL 2025 Oral。
@misc{zhang2025hub,
title = {{HuB: Learning Extreme Humanoid Balance}},
author = {Tong Zhang and Boyuan Zheng and Ruiqian Nai and Yingdong Hu and Yen-Jen Wang and Geng Chen and Fanqi Lin and Jiongye Li and Chuye Hong and Koushil Sreenath and Yang Gao},
year = {2025},
eprint = {2505.07294},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.07294},
}
Agile humanoid walking on sparse footholds, using a sampling-based foothold reward to overcome how rarely good footsteps occur during learning.让人形机器人在稀疏落脚点上敏捷行走，用基于采样的落脚奖励解决学习中有效落脚过于稀少的问题。
@misc{wang2025beamdojo,
title = {{BeamDojo: Learning Agile Humanoid Locomotion on Sparse Footholds}},
author = {Huayi Wang and Zirui Wang and Junli Ren and Qingwei Ben and Tao Huang and Weinan Zhang and Jiangmiao Pang},
year = {2025},
eprint = {2502.10363},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.10363},
}
Blends reusable skills for humanoid whole-body loco-manipulation, avoiding the task-specific tuning most controllers need.通过融合可复用技能实现人形全身移动操作，避免多数控制器所需的按任务调参。
@misc{kuang2025skillblender,
title = {{SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending}},
author = {Yuxuan Kuang and Haoran Geng and Amine Elhafsi and Tan-Dzung Do and Pieter Abbeel and Jitendra Malik and Marco Pavone and Yue Wang},
year = {2025},
eprint = {2506.09366},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.09366},
}
A diffusion prior trained on human motion guides RL in simulation, so policies find whole-body solutions for tasks such as opening a drawer that direct RL misses.用人体动作训练的扩散先验引导仿真中的 RL，使策略找到开抽屉等任务的全身解法，而直接 RL 难以发现。
@misc{kalaria2025dreamcontrol,
title = {{DreamControl: Human-Inspired Whole-Body Humanoid Control for Scene Interaction via Guided Diffusion}},
author = {Dvij Kalaria and Sudarshan S Harithas and Pushkal Katara and Sangkyung Kwak and Sarthak Bhagat and Shankar Sastry and Srinath Sridhar and Sai Vemprala and Ashish Kapoor and Jonathan Chung-Kuan Huang},
year = {2025},
eprint = {2509.14353},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.14353},
}
PhysHSI: a real-world humanoid-scene interaction system for tasks such as carrying objects and sitting on chairs, keeping motion natural and scene perception robust.PhysHSI：面向搬物、坐椅等任务的真实人形-场景交互系统，兼顾动作自然与场景感知鲁棒。
@misc{wang2025physhsi,
title = {{PhysHSI: Towards a Real-World Generalizable and Natural Humanoid-Scene Interaction System}},
author = {Huayi Wang and Wentao Zhang and Runyi Yu and Tao Huang and Junli Ren and Feiyu Jia and Zirui Wang and Xiaojie Niu and Xiao Chen and Jiahe Chen and Qifeng Chen and Jingbo Wang and Jiangmiao Pang},
year = {2025},
eprint = {2510.11072},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.11072},
}
HDMI learns whole-body humanoid-object interaction from monocular RGB videos, by extracting robot and object trajectories and co-tracking them with RL.HDMI 从单目 RGB 视频学习人形全身与物体的交互：提取并重定向人与物体轨迹，再用 RL 同时跟踪二者。
@misc{weng2025hdmi,
title = {{HDMI: Learning Interactive Humanoid Whole-Body Control from Human Videos}},
author = {Haoyang Weng and Yitang Li and Nikhil Sobanbabu and Zihan Wang and Zhengyi Luo and Tairan He and Deva Ramanan and Guanya Shi},
year = {2025},
eprint = {2509.16757},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.16757},
}
Humanoid table tennis with a model-based planner and an RL whole-body controller; reports up to 106 consecutive shots with a human opponent.人形乒乓球系统，结合基于模型的规划器和 RL 全身控制器；报告与人对打最多连续 106 拍。
@misc{su2025hitter,
title = {{HITTER: A HumanoId Table TEnnis Robot via Hierarchical Planning and Learning}},
author = {Zhi Su and Bike Zhang and Nima Rahmanian and Yuman Gao and Qiayuan Liao and Caitlin Regan and Koushil Sreenath and S. Shankar Sastry},
year = {2025},
eprint = {2508.21043},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2508.21043},
}
Humanoid shadowing that allows contact with any body part, such as sitting, rolling and getting up, instead of only feet and hands.允许任意身体部位接触环境的人形动作跟随，可完成坐、滚、起身等动作，而不只限于手脚接触。
HoST learns to stand up from diverse postures with RL, without predefined ground-specific motions, and deploys on real robots.HoST 用强化学习从多种姿态起身，无需预定义的特定地面动作，并部署到真机。
@misc{huang2025learning,
title = {{Learning Humanoid Standing-up Control across Diverse Postures}},
author = {Tao Huang and Junli Ren and Huayi Wang and Zirui Wang and Qingwei Ben and Muning Wen and Xiao Chen and Jianan Li and Jiangmiao Pang},
year = {2025},
eprint = {2502.08378},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.08378},
}
A learning framework for getting up from varied fallen configurations on varied terrains, where hand-designed controllers struggle.针对不同跌倒姿态和地形学习起身控制器，而这类场景很难手工设计控制器。
R2S2 tunes primitive skills for sim-to-real, ensembles them into a latent space, and trains a planner over it for large-space humanoid reaching.R2S2 先调好可迁移到真机的基础技能，再融合成隐空间，并在其上训练规划器，实现人形大范围够取。
@misc{zhang2025unleashing,
title = {{Unleashing Humanoid Reaching Potential via Real-world-Ready Skill Space}},
author = {Zhikai Zhang and Chao Chen and Han Xue and Jilong Wang and Sikai Liang and Yun Liu and Zongzhang Zhang and He Wang and Li Yi},
year = {2025},
eprint = {2505.10918},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.10918},
}
Learns compliant whole-body control from example motions, so a humanoid yields to unexpected contact instead of stiffly correcting back to the reference.从示例动作学习柔顺全身控制，使人形机器人在意外接触时顺应，而不是僵硬地回到参考动作。
@misc{margolis2025softmimic,
title = {{SoftMimic: Learning Compliant Whole-body Control from Examples}},
author = {Gabriel B. Margolis and Michelle Wang and Nolan Fey and Pulkit Agrawal},
year = {2025},
eprint = {2510.17792},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.17792},
}
Integrates impedance control into whole-body tracking so a humanoid's upper body stays compliant in contact with people and objects.将阻抗控制融入全身跟踪，使人形上半身在接触人和物体时保持柔顺。
@misc{lu2025gentlehumanoid,
title = {{GentleHumanoid: Learning Upper-body Compliance for Contact-rich Human and Object Interaction}},
author = {Qingzhou Lu and Yao Feng and Baiyu Shi and Michael Piseno and Zhenan Bao and C. Karen Liu},
year = {2025},
eprint = {2511.04679},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.04679},
}
CHIP is a plug-and-play module adding controllable end-effector stiffness to agile tracking policies, for forceful tasks such as wiping and pushing carts.CHIP 是即插即用模块，为敏捷跟踪策略加入可控的末端刚度，用于擦拭、推车等需要施力的任务。
@misc{chen2025chip,
title = {{CHIP: Adaptive Compliance for Humanoid Control through Hindsight Perturbation}},
author = {Sirui Chen and Zi-ang Cao and Zhengyi Luo and Fernando Castañeda and Chenran Li and Tingwu Wang and Ye Yuan and Linxi "Jim" Fan and C. Karen Liu and Yuke Zhu},
year = {2025},
eprint = {2512.14689},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2512.14689},
}
Survey of behavior foundation models, pretrained generalist whole-body controllers that avoid costly retraining for each new task. Has a companion paper list.综述行为基础模型：预训练的通用全身控制器，避免每个新任务都重新训练。附配套论文列表。
@misc{yuan2025survey,
title = {{A Survey of Behavior Foundation Model: Next-Generation Whole-Body Control System of Humanoid Robots}},
author = {Mingqi Yuan and Tao Yu and Wenqi Ge and Xiuyong Yao and Huijiang Wang and Jiayu Chen and Bo Li and Wei Zhang and Wenjun Zeng and Hua Chen and Xin Jin},
year = {2025},
eprint = {2506.20487},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.20487},
}
@misc{gu2025humanoid,
title = {{Humanoid Locomotion and Manipulation: Current Progress and Challenges in Control, Planning, and Learning}},
author = {Zhaoyuan Gu and Junheng Li and Wenlan Shen and Wenhao Yu and Zhaoming Xie and Stephen McCrory and Xianyi Cheng and Abdulaziz Shamsah and Robert Griffin and C. Karen Liu and Abderrahmane Kheddar and Xue Bin Peng and Yuke Zhu and Guanya Shi and Quan Nguyen and Gordon Cheng and Huijun Gao and Ye Zhao},
year = {2025},
eprint = {2501.02116},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2501.02116},
}
Humanoid-VLA combines language understanding, egocentric perception and motion control toward universal humanoid control.Humanoid-VLA 融合语言理解、第一视角感知和运动控制，面向通用人形控制。
@misc{ding2025humanoid,
title = {{Humanoid-VLA: Towards Universal Humanoid Control with Visual Integration}},
author = {Pengxiang Ding and Jianfei Ma and Xinyang Tong and Binghong Zou and Xinxin Luo and Yiguo Fan and Ting Wang and Hongchao Lu and Panzhong Mo and Jinxin Liu and Yuefan Wang and Huaicheng Zhou and Wenshuo Feng and Jiacheng Liu and Siteng Huang and Donglin Wang},
year = {2025},
eprint = {2502.14795},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.14795},
}
RHINO learns real-time humanoid-human-object interaction from human demonstrations, reacting to human signals instead of fixed task stages.RHINO 从人类示教学习实时人形-人-物交互，根据人的信号反应，而非固定任务阶段。
@misc{chen2025rhino,
title = {{RHINO: Learning Real-Time Humanoid-Human-Object Interaction from Human Demonstrations}},
author = {Jingxiao Chen and Xinyao Li and Jiahang Cao and Zhengbang Zhu and Wentao Dong and Minghuan Liu and Ying Wen and Yong Yu and Liqing Zhang and Weinan Zhang},
year = {2025},
eprint = {2502.13134},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.13134},
}
NIL learns motor skills across humanoids, quadrupeds and animals from pre-trained video diffusion models, with no demo data.NIL 借助预训练视频扩散模型学习人形、四足等形态的运动技能，无需示教数据。
@misc{albaba2025nil,
title = {{NIL: No-data Imitation Learning by Leveraging Pre-trained Video Diffusion Models}},
author = {Mert Albaba and Chenhao Li and Markos Diomataris and Omid Taheri and Andreas Krause and Michael Black},
year = {2025},
eprint = {2503.10626},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2503.10626},
}
Fine-tunes video world models on a robot to dream new demos, labels them with latent actions or an inverse dynamics model, and trains policies on them; a humanoid learns 20+ new behaviors from one-scene teleop data.把视频世界模型微调到目标机器人上“做梦”生成新示范，再用隐动作或逆动力学模型补上动作标签用于训练策略；人形机器人仅凭单一场景的遥操作数据学会 20 多种新行为。
@misc{jang2025dreamgen,
title = {{DreamGen: Unlocking Generalization in Robot Learning through Video World Models}},
author = {Joel Jang and Seonghyeon Ye and Zongyu Lin and Jiannan Xiang and Johan Bjorck and Yu Fang and Fengyuan Hu and Spencer Huang and Kaushil Kundalia and Yen-Chen Lin and Loic Magne and Ajay Mandlekar and Avnish Narayan and You Liang Tan and Guanzhi Wang and Jing Wang and Qi Wang and Yinzhen Xu and Xiaohui Zeng and Kaiyuan Zheng and Ruijie Zheng and Ming-Yu Liu and Luke Zettlemoyer and Dieter Fox and Jan Kautz and Scott Reed and Yuke Zhu and Linxi Fan},
year = {2025},
eprint = {2505.12705},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.12705},
}
Tests 'embodiment scaling laws' on about 1,000 generated robots: does training on more bodies improve generalization to unseen ones?在约 1000 个生成的机器人形态上检验“形态缩放定律”：用更多形态训练能否提升对未见形态的泛化？
@misc{ai2025towards,
title = {{Towards Embodiment Scaling Laws in Robot Locomotion}},
author = {Bo Ai and Liu Dai and Nico Bohlinger and Dichen Li and Tongzhou Mu and Zhanxin Wu and K. Fay and Henrik I. Christensen and Jan Peters and Hao Su},
year = {2025},
eprint = {2505.05753},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.05753},
}
Drops 3D Gaussian Splatting into vectorized physics simulators as a renderer, reporting over 100,000 steps per second on consumer GPUs for learning from pixels.把 3D 高斯溅射作为渲染器接入矢量化物理仿真，报告在消费级 GPU 上超过每秒 10 万步，用于从像素学习。
@misc{escontrela2025gaussgym,
title = {{GaussGym: An open-source real-to-sim framework for learning locomotion from pixels}},
author = {Alejandro Escontrela and Justin Kerr and Arthur Allshire and Jonas Frey and Rocky Duan and Carmelo Sferrazza and Pieter Abbeel},
year = {2025},
eprint = {2510.15352},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.15352},
}
Portable, mocap-free humanoid teleoperation and data collection built around a VR headset, aimed at making whole-body demonstrations scale.基于 VR 头显的便携式免动捕人形遥操作与数据采集系统，目标是让全身示教数据能够规模化采集。
@misc{ze2025twist,
title = {{TWIST2: Scalable, Portable, and Holistic Humanoid Data Collection System}},
author = {Yanjie Ze and Siheng Zhao and Weizhuo Wang and Angjoo Kanazawa and Rocky Duan and Pieter Abbeel and Guanya Shi and Jiajun Wu and C. Karen Liu},
year = {2025},
eprint = {2511.02832},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.02832},
}
A small, 3D-printed open-source humanoid with a matching simulation model, built so that ML policies and teleoperated data transfer to the real robot.3D 打印的小型开源人形机器人，配套仿真模型，便于机器学习策略与遥操作数据迁移到真机。
@misc{shi2025toddlerbot,
title = {{ToddlerBot: Open-Source ML-Compatible Humanoid Platform for Loco-Manipulation}},
author = {Haochen Shi and Weizhuo Wang and Shuran Song and C. Karen Liu},
year = {2025},
eprint = {2502.00893},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.00893},
}
Open-source whole-body humanoid teleoperation — an arm exoskeleton plus motion-sensing gloves, with a foot pedal steering an RL walking controller; the cockpit hardware costs about US$500.开源的人形机器人全身遥操作系统：手臂外骨骼加动作捕捉手套，脚踏板驱动强化学习行走控制器；整套座舱硬件约 500 美元。
@misc{ben2025homie,
title = {{HOMIE: Humanoid Loco-Manipulation with Isomorphic Exoskeleton Cockpit}},
author = {Qingwei Ben and Feiyu Jia and Jia Zeng and Junting Dong and Dahua Lin and Jiangmiao Pang},
year = {2025},
eprint = {2502.13013},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.13013},
}
Open-source biped built from e-commerce parts by sheet-metal welding, avoiding the size and fragility limits of 3D-printed designs.用电商零件和钣金焊接搭建的开源双足机器人，避开 3D 打印方案在尺寸和强度上的限制。
@misc{kawaharazuka2025mevita,
title = {{MEVITA: Open-Source Bipedal Robot Assembled from E-Commerce Components via Sheet Metal Welding}},
author = {Kento Kawaharazuka and Shogo Sawaguchi and Ayumu Iwata and Keita Yoneda and Temma Suzuki and Kei Okada},
year = {2025},
eprint = {2508.17684},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2508.17684},
}
Imitates upper-body human motion capture while keeping robust legs, giving humanoids expressive gestures and dance.上半身模仿人类动捕、下肢保持鲁棒行走，让人形机器人做出富有表现力的动作和舞蹈。
@misc{cheng2024expressive,
title = {{Expressive Whole-Body Control for Humanoid Robots}},
author = {Xuxin Cheng and Yandong Ji and Junming Chen and Ruihan Yang and Ge Yang and Xiaolong Wang},
year = {2024},
eprint = {2402.16796},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2402.16796},
}
H2O — an RL tracking policy lets an RGB camera stream a person's whole-body motion onto a humanoid in real time.H2O：强化学习跟踪策略借助 RGB 相机，把人的全身动作实时映射到人形机器人。
@misc{he2024learning,
title = {{Learning Human-to-Humanoid Real-Time Whole-Body Teleoperation}},
author = {Tairan He and Zhengyi Luo and Wenli Xiao and Chong Zhang and Kris Kitani and Changliu Liu and Guanya Shi},
year = {2024},
eprint = {2403.04436},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2403.04436},
}
@misc{fu2024humanplus,
title = {{HumanPlus: Humanoid Shadowing and Imitation from Humans}},
author = {Zipeng Fu and Qingqing Zhao and Qi Wu and Gordon Wetzstein and Chelsea Finn},
year = {2024},
eprint = {2406.10454},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2406.10454},
}
One RL whole-body controller that a humanoid can follow from a VR headset, verbal commands or an RGB camera, and that also runs learned autonomous policies.一个强化学习全身控制器：人形机器人可跟随 VR 头显、语音指令或 RGB 相机的输入，也可执行学习得到的自主策略。
@misc{he2024omnih,
title = {{OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning}},
author = {Tairan He and Zhengyi Luo and Xialin He and Wenli Xiao and Chong Zhang and Weinan Zhang and Kris Kitani and Changliu Liu and Guanya Shi},
year = {2024},
eprint = {2406.08858},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2406.08858},
}
Distills one policy that supports several humanoid control modes (root velocity, joint angles, keypoint tracking) instead of training a separate controller per mode.蒸馏出单一策略，支持多种人形控制模式（根部速度、关节角、关键点跟踪），不必为每种模式各训练一个控制器。
@misc{he2024hover,
title = {{HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots}},
author = {Tairan He and Wenli Xiao and Toru Lin and Zhengyi Luo and Zhenjia Xu and Zhenyu Jiang and Jan Kautz and Changliu Liu and Guanya Shi and Xiaolong Wang and Linxi Fan and Yuke Zhu},
year = {2024},
eprint = {2410.21229},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.21229},
}
Learns a vision-based whole-body parkour policy for the Unitree H1 without motion references, covering skills such as jumping onto platforms and leaping over gaps, and transfers it to the real robot.无需动作参考，为 Unitree H1 学习基于视觉的全身跑酷策略，涵盖跳上平台、跨越沟壑等技能，并迁移到真实机器人。
ABS pairs an agile RL locomotion policy with a learned reach-avoid value network and a recovery policy that takes over near obstacles, so a quadruped can run fast without crashing.ABS 将敏捷的强化学习运动策略与学习的可达-规避价值网络及恢复策略结合，接近障碍时由恢复策略接管，让四足机器人高速奔跑而不碰撞。
@misc{he2024agile,
title = {{Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion}},
author = {Tairan He and Chong Zhang and Wenli Xiao and Guanqi He and Changliu Liu and Guanya Shi},
year = {2024},
eprint = {2401.17583},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2401.17583},
}
Imposes a Lipschitz constraint on the learned policy as a general route to smooth motion, replacing hand-tuned low-pass filters and smoothness rewards.对学习到的策略施加 Lipschitz 约束，作为获得平滑动作的通用方法，取代需手动调参的低通滤波和平滑奖励。
@misc{chen2024learning,
title = {{Learning Smooth Humanoid Locomotion through Lipschitz-Constrained Policies}},
author = {Zixuan Chen and Xialin He and Yen-Jen Wang and Qiayuan Liao and Yanjie Ze and Zhongyu Li and S. Shankar Sastry and Jiajun Wu and Koushil Sreenath and Saurabh Gupta and Xue Bin Peng},
year = {2024},
eprint = {2410.11825},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.11825},
}
@misc{ji2024exbody,
title = {{ExBody2: Advanced Expressive Humanoid Whole-Body Control}},
author = {Mazeyu Ji and Xuanbin Peng and Fangchen Liu and Jialong Li and Ge Yang and Xuxin Cheng and Xiaolong Wang},
year = {2024},
eprint = {2412.13196},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.13196},
}
Generates humanoid whole-body motion from language, starting from human motion priors and refining with LLM commonsense reasoning. Oral at CoRL 2024.从语言描述生成人形全身动作：以人体动作先验初始化，再用 LLM 的常识推理加以修正。CoRL 2024 Oral。
@misc{jiang2024harmon,
title = {{Harmon: Whole-Body Motion Generation of Humanoid Robots from Language Descriptions}},
author = {Zhenyu Jiang and Yuqi Xie and Jinhan Li and Ye Yuan and Yifeng Zhu and Yuke Zhu},
year = {2024},
eprint = {2410.12773},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.12773},
}
Keeps IK and retargeting for precise upper-body control, and uses RL with Predictive Motion Priors for robust legs.上半身保留 IK 与重定向以求精确，腿部用结合预测运动先验的 RL 保证鲁棒。
@misc{lu2024mobile,
title = {{Mobile-TeleVision: Predictive Motion Priors for Humanoid Whole-Body Control}},
author = {Chenhao Lu and Xuxin Cheng and Jialong Li and Shiqi Yang and Mazeyu Ji and Chengjing Yuan and Ge Yang and Sha Yi and Xiaolong Wang},
year = {2024},
eprint = {2412.07773},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.07773},
}
Learns humanoid tasks as sequences of contacts, reducing the task-specific tuning and state machines RL usually needs.把人形任务建模为接触序列来学习，减少 RL 通常需要的按任务调参和状态机设计。
@misc{zhang2024wococo,
title = {{WoCoCo: Learning Whole-Body Humanoid Control with Sequential Contacts}},
author = {Chong Zhang and Wenli Xiao and Tairan He and Guanya Shi},
year = {2024},
eprint = {2406.06005},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2406.06005},
}
@misc{bohlinger2024one,
title = {{One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion}},
author = {Nico Bohlinger and Grzegorz Czechmanowski and Maciej Krupka and Piotr Kicki and Krzysztof Walas and Jan Peters and Davide Tateo},
year = {2024},
eprint = {2409.06366},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2409.06366},
}
Hierarchical world model for visual whole-body humanoid control: a high-level agent commands a low-level one, with no reward design or skill primitives.用于视觉全身人形控制的分层世界模型：高层智能体指挥低层智能体，不依赖奖励设计或技能基元。
@misc{hansen2024hierarchical,
title = {{Hierarchical World Models as Visual Whole-Body Humanoid Controllers}},
author = {Nicklas Hansen and Jyothir S and Vlad Sobal and Yann LeCun and Xiaolong Wang and Hao Su},
year = {2024},
eprint = {2405.18418},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2405.18418},
}
A VLA outputs mid-level navigation commands in language that an RL locomotion policy executes, so legged robots follow natural-language routes; adds Isaac Lab VLN benchmarks.VLA 先用语言输出中层导航指令，再由强化学习运动策略执行，让腿式机器人按自然语言指令导航；并提供 Isaac Lab 上的 VLN 基准。
@misc{cheng2024navila,
title = {{NaVILA: Legged Robot Vision-Language-Action Model for Navigation}},
author = {An-Chieh Cheng and Yandong Ji and Zhaojing Yang and Zaitian Gongye and Xueyan Zou and Jan Kautz and Erdem Bıyık and Hongxu Yin and Sifei Liu and Xiaolong Wang},
year = {2024},
eprint = {2412.04453},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.04453},
}
27 simulated tasks (12 locomotion, 15 whole-body manipulation) on a Unitree H1 with two dexterous hands; reports that DreamerV3, TD-MPC2, SAC and PPO fail most of them, especially long-horizon whole-body ones.27 个仿真任务（12 个运动、15 个全身操作），机器人为带两只灵巧手的 Unitree H1；报告 DreamerV3、TD-MPC2、SAC、PPO 在大多数任务上未达标，长时程全身协调任务尤甚。
@misc{sferrazza2024humanoidbench,
title = {{HumanoidBench: Simulated Humanoid Benchmark for Whole-Body Locomotion and Manipulation}},
author = {Carmelo Sferrazza and Dun-Ming Huang and Xingyu Lin and Youngwoon Lee and Pieter Abbeel},
year = {2024},
eprint = {2403.10506},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2403.10506},
}
@misc{liao2024berkeley,
title = {{Berkeley Humanoid: A Research Platform for Learning-based Control}},
author = {Qiayuan Liao and Bike Zhang and Xuanyu Huang and Xiaoyu Huang and Zhongyu Li and Koushil Sreenath},
year = {2024},
eprint = {2407.21781},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2407.21781},
}
Robot-free UMI gripper demos train the manipulation policy; a sim-trained whole-body controller tracks its end-effector trajectories on a quadruped. Reports >70% success on all tasks and zero-shot reuse of an arm policy.用手持 UMI 夹爪采集无需机器人的示教来训练操作策略，再由仿真训练的全身控制器在四足机器人上跟踪其输出的末端轨迹；报告所有任务成功率超过 70%，并能零样本复用原为固定机械臂训练的策略。
@misc{ha2024umi,
title = {{UMI-on-Legs: Making Manipulation Policies Mobile with Manipulation-Centric Whole-body Controllers}},
author = {Huy Ha and Yihuai Gao and Zipeng Fu and Jie Tan and Shuran Song},
year = {2024},
eprint = {2407.10353},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2407.10353},
}
Open-source 10-DoF humanoid for locomotion research; an RL policy using passive dynamics cuts cost of transport by up to 50% in sim and 31% on hardware.面向运动研究的开源 10 自由度人形；利用被动动力学的强化学习策略将运输成本在仿真中最多降低 50%，实机降低 31%。
@misc{xia2024duke,
title = {{The Duke Humanoid: Design and Control For Energy Efficient Bipedal Locomotion Using Passive Dynamics}},
author = {Boxi Xia and Bokuan Li and Jacob Lee and Michael Scutari and Boyuan Chen},
year = {2024},
eprint = {2409.19795},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2409.19795},
}
A blind quadruped policy that infers terrain properties implicitly from proprioception history with a context-aided estimator, enabling robust walking on stairs and rough ground without cameras.盲视四足策略：借助上下文辅助估计器从本体感觉历史中隐式推断地形特性，无需相机即可稳健地走楼梯和崎岖地面。
A learned navigation policy sequences pre-trained skills (walk, jump, climb, crouch) using a learned obstacle reconstruction; reports up to 2 m/s on hardware (Science Robotics).分层的全学习系统：导航策略结合学习到的障碍物重建来调度预训练技能（行走、跳跃、攀爬、下蹲）；报告实机速度可达 2 米/秒。发表于 Science Robotics。
TRILL collects VR demonstrations and uses a whole-body controller to turn task-space commands into joint torques, for data-efficient loco-manipulation imitation.TRILL 用 VR 采集演示，由全身控制器把任务空间指令转为关节力矩，实现数据高效的人形移动操作模仿。
@misc{seo2023deep,
title = {{Deep Imitation Learning for Humanoid Loco-manipulation through Human Teleoperation}},
author = {Mingyo Seo and Steve Han and Kyutae Sim and Seung Hyeon Bang and Carlos Gonzalez and Luis Sentis and Yuke Zhu},
year = {2023},
eprint = {2309.01952},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2309.01952},
}
@misc{wang2023prompt,
title = {{Prompt a Robot to Walk with Large Language Models}},
author = {Yen-Jen Wang and Bike Zhang and Jianyu Chen and Koushil Sreenath},
year = {2023},
eprint = {2309.09969},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2309.09969},
}
ANYmal fuses proprioception with noisy height-map perception through a learned belief encoder, and is shown hiking an alpine trail and other rough terrain.ANYmal 通过学习的信念编码器融合本体感知与带噪声的高程图感知，在高山徒步路线等复杂地形上完成行走。
@misc{miki2022learning,
title = {{Learning robust perceptive locomotion for quadrupedal robots in the wild}},
author = {Takahiro Miki and Joonho Lee and Jemin Hwangbo and Lorenz Wellhausen and Vladlen Koltun and Marco Hutter},
year = {2022},
eprint = {2201.08117},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2201.08117},
}
@misc{kumar2021rma,
title = {{RMA: Rapid Motor Adaptation for Legged Robots}},
author = {Ashish Kumar and Zipeng Fu and Deepak Pathak and Jitendra Malik},
year = {2021},
eprint = {2107.04034},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2107.04034},
}
A discriminator trained on unstructured mocap clips acts as a style reward, so RL learns tasks with natural motion without hand-picked reference clips; widely reused as a gait prior for legged robots.用在无结构动捕片段上训练的判别器作为"风格奖励"，强化学习无需逐段指定参考动作即可学到自然的动作；后来被广泛用作足式机器人的步态先验。
@misc{peng2021amp,
title = {{AMP: Adversarial Motion Priors for Stylized Physics-Based Character Control}},
author = {Xue Bin Peng and Ze Ma and Pieter Abbeel and Sergey Levine and Angjoo Kanazawa},
year = {2021},
eprint = {2104.02180},
archivePrefix = {arXiv},
primaryClass = {cs.GR},
url = {https://arxiv.org/abs/2104.02180},
}
@misc{lee2020learning,
title = {{Learning Quadrupedal Locomotion over Challenging Terrain}},
author = {Joonho Lee and Jemin Hwangbo and Lorenz Wellhausen and Vladlen Koltun and Marco Hutter},
year = {2020},
eprint = {2010.11251},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2010.11251},
}
@misc{hwangbo2019learning,
title = {{Learning Agile and Dynamic Motor Skills for Legged Robots}},
author = {Jemin Hwangbo and Joonho Lee and Alexey Dosovitskiy and Dario Bellicoso and Vassilios Tsounis and Vladlen Koltun and Marco Hutter},
year = {2019},
eprint = {1901.08652},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/1901.08652},
}
Debadrata Sarkar, Upamanyu Ray, Soumen Sen, Aman Arora
Abstract摘要
Wearable robots depend on a software chain that links body-worn sensing to gait interpretation, assistance generation, and actuation. Personalization further requires each measured response to remain associated with the assistance parameters that generated it. This paper presents ExoStack, a modular ROS 2 framework for implementing that chain across embedded hubs and a host computer. Its Embedded Sensing and Actuation (ESA) library configures ESP32 boards as microROS sensing and actuation hubs. ExoCore supplies multi-hub synchronization, gait-event detection, profile generation, motion …
@misc{sarkar2026exostack,
title = {{ExoStack: A Modular and Reconfigurable Software Framework for Wearable Robotics Using ROS 2}},
author = {Debadrata Sarkar and Upamanyu Ray and Soumen Sen and Aman Arora},
year = {2026},
eprint = {2611.12306},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12306},
}
Xueyang Wang, Zehong Shen, Dingkun Wei, Huaijin Pi et al.
Abstract摘要
We study humanoid navigation across multiple destinations under long-horizon language instructions. This task is challenging because robots must maintain progress across successive goals while continuously producing natural, stable whole-body motion through confined spaces. However, task decomposition in existing methods does not by itself ensure progress throughout execution, while waypoint-based navigation does not guarantee natural, route-faithful whole-body motion. Our key idea is to organize navigation into three complementary roles operating at different time scales: a Brain for task …
@misc{wang2026lhnav,
title = {{LHNav: Long-Horizon Humanoid Navigation via Multirate Coordination and Physics-Grounded Action Data}},
author = {Xueyang Wang and Zehong Shen and Dingkun Wei and Huaijin Pi and Changpeng Wang and Xing Zhu and Hujun Bao and Yujun Shen and others},
year = {2026},
eprint = {2611.12136},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12136},
}
Bingzhang Wang, Nathan Zhao, Jialeng Ni, Jiajin Zheng et al.
Abstract摘要
Vision-language-action (VLA) models pair a vision-language backbone with an action head, but in closed-loop control the backbone sits on the control-critical path: while it computes, the robot keeps moving, so each executed action is conditioned on an increasingly stale observation. A larger backbone therefore directly blocks reactivity, especially in humanoid dexterous manipulation. We present DASH, a Decoupled Asynchronous System for Humanoids that removes this coupling. DASH splits the policy into a slow vision-language model (System 2) and a fast flow-matching action head (System 1) that …
@misc{wang2026decoupled,
title = {{Decoupled Asynchronous System for Humanoids (DASH): Latency-Agnostic Scaling of Vision-Language-Action Models for Humanoid Manipulation}},
author = {Bingzhang Wang and Nathan Zhao and Jialeng Ni and Jiajin Zheng and Yin Zhang and Kehua Chen and Jian Yao},
year = {2026},
eprint = {2611.11854},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11854},
}
Ziwon Yoon, Seok Hwan Kim, Yiyu Chen, Quan Nguyen et al.
Abstract摘要
Humanoid robots have achieved significant progress in locomotion and manipulation, enabling traversal over rough terrain and richer interaction with the environment, such as moving obstacles. However, most traditional navigation methods assume non-interactive environments and focus only on traversal. While Navigation Among Movable Obstacles (NAMO) approaches relax this assumption, prior efforts have largely focused on flat environments and reactive interaction, rather than predicting object movability and planning ahead. These limitations become critical in outdoor rough-terrain scenarios, …
@misc{yoon2026hint,
title = {{HINT-VLM: Humanoid Interactive Navigation over Rough Terrain with Embodiment-Grounded Vision-Language Model}},
author = {Ziwon Yoon and Seok Hwan Kim and Yiyu Chen and Quan Nguyen and Lu Gan and Ye Zhao},
year = {2026},
eprint = {2611.11665},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11665},
}
Yunheng Wang, Jiaxi Zhang, Wei Lu, Yuetong Fang et al.
Abstract摘要
Humanoid loco-manipulation requires continuous coordination of high-dimensional whole-body motion to realize task objectives. While vision-language-action models and world action models transfer semantic and visual-dynamics priors to robot control, these priors do not necessarily capture the structure of robot motion. We introduce Kinematic Action Models (KAMs), a paradigm that treats whole-body motion as a direct pretraining target and uses a pretrained motion generator as the primary action-generation backbone. KAMs ground learned temporal and cross-body coordination priors through task …
@misc{wang2026kimodo,
title = {{Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation}},
author = {Yunheng Wang and Jiaxi Zhang and Wei Lu and Yuetong Fang and Taowen Wang and Yixiao Feng and Shuning Zhang and Junzhe Xu and others},
year = {2026},
eprint = {2611.11554},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11554},
}
We present a pipeline for transferring Awa Odori, a traditional Japanese festival dance, from monocular video to a commercial Unitree G1 EDU humanoid. The pipeline combines human mesh recovery, kinematic retargeting, reinforcement learning for motion tracking in simulation, and onboard deployment via ONNX. Training ran on a single RTX 5070 GPU. An eleven-version deployment ladder evaluated on hardware examines three interventions. We detect and repair silent corruption of deployment artifacts, model state-estimator drift during training, and edit the reference motion to support stable …
@misc{shehata2026from,
title = {{From Festival Video to the Stage: A Sim-to-Real Ladder for Long-Horizon Expressive Whole-Body Dance on a Commercial Humanoid}},
author = {Hesham Shehata},
year = {2026},
eprint = {2611.11127},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11127},
}
Mayur Deshmukh, Aaron Tung, Junyi Li, Nikash Bhardwaj et al.
Abstract摘要
Humanoid robots must coordinate locomotion and manipulation to navigate complex environments and perform everyday tasks. Learning such behaviours from human demonstrations requires accurate whole-body motion paired with egocentric vision, particularly for lower-body movements and foot placements that are frequently invisible from a first-person perspective. However, existing egocentric datasets often lack high-fidelity optical motion capture or systematic coverage of diverse loco-manipulation behaviours. To address this gap, we introduce RoboMoCap, a large-scale multimodal dataset combining …
@misc{deshmukh2026robomocap,
title = {{RoboMoCap: A Large-Scale Whole-Body Motion Capture Dataset with Egocentric Vision for Robot Learning}},
author = {Mayur Deshmukh and Aaron Tung and Junyi Li and Nikash Bhardwaj and Ademi Adeniji and Vincent Liu and Michael Cho and Jiafei Duan},
year = {2026},
eprint = {2611.10994},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10994},
}
Kevin Park, Izat Rasyad, Dvij Kalaria, Shankar Sastry
Abstract摘要
Training vision-language-action (VLA) policies typically depends on substantial robot demonstration data, making human teleoperation a costly and cumbersome bottleneck, particularly for whole-body humanoid interaction. Reinforcement learning in simulation offers an alternative for scaling this data, but designing task-specific rewards remains challenging. To address this, we propose DreamVLA, which uses generated motion trajectories to guide reinforcement learning (RL) in simulation, converting them into reliable task execution trajectories. These trajectories train VLA models in simulation …
@misc{park2026dreamvla,
title = {{DreamVLA: From Generated Human Motions to Humanoid Vision-Language-Action Policies}},
author = {Kevin Park and Izat Rasyad and Dvij Kalaria and Shankar Sastry},
year = {2026},
eprint = {2611.10872},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10872},
}
Athanasios Kyriakopoulos, Shubham S. Kumbhar, Kunal S. Narkhede, Abhijeet M. Kulkarni et al.
Abstract摘要
A fundamental aspect of humanoid footstep planning is to jointly guarantee obstacle avoidance and locomotion stability. This paper presents a capturability-aware sequential Model Predictive Control (MPC) framework for Linear Inverted Pendulum (LIP)-based humanoid footstep planning in the presence of obstacles. The proposed method combines obstacle avoidance with capturability constraints through safe corridors constructed as sequences of intersecting rectangles, yielding a sequential MPC framework with recursive feasibility guarantees for the LIP-based planning layer. The planner autonomously …
@misc{kyriakopoulos2026capturability,
title = {{Capturability-Aware Footstep Planning in Cluttered Environments: A Recursively Feasible Sequential MPC on Axis-Aligned Safe Corridors}},
author = {Athanasios Kyriakopoulos and Shubham S. Kumbhar and Kunal S. Narkhede and Abhijeet M. Kulkarni and Ioannis Poulakakis},
year = {2026},
eprint = {2611.10169},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10169},
}
Yutian Chen, Shibo Zhao, Rocky Duan, Karen Liu et al.
Abstract摘要
Learning humanoid state estimation and language-grounded motion understanding from robot recordings ties training coverage to the robot's motion repertoire. Human motion datasets offer broader coverage, yet transferring this knowledge requires overcoming differences in body geometry, sensing, and physical disturbances. To enable this transfer, we propose a unified inertial encoder whose shared inertial interface uses kinematic IMU synthesis to represent human and humanoid motion at corresponding body sites. Within this common input space, joint training on human and kinematically retargeted …
@misc{chen2026unified,
title = {{A Unified Inertial Encoder for Human and Humanoid Motion}},
author = {Yutian Chen and Shibo Zhao and Rocky Duan and Karen Liu and Carmelo Sferrazza and Guanya Shi and Yuheng Qiu and Koushil Sreenath},
year = {2026},
eprint = {2611.10137},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10137},
}
This paper presents the design and preliminary evaluation of KR A1, a humanoid-robot interaction prototype that integrates personalized voice synthesis, contextual dialogue processing, and emotion-aware response generation for potential mental-healthcare applications. The system architecture combines a locally hosted large language model (Qwen2.5-3B-Instruct, served via Ollama) for natural-language understanding and response generation, OpenVoice v2 for zero-shot speaker-adaptive speech synthesis, and a two-tier emotion-processing pipeline comprising a keyword heuristic and LLM-based …
Adarsh Salagame, Henry Noyes, Haroon Hublikar, Alireza Ramezani
Abstract摘要
Snake robots can leverage their hyper-redundant morphology for both locomotion and manipulation through distributed body contact but whole-body loco-manipulation is particularly challenging since the entire body serves simultaneously as propulsive mechanism and manipulation tool Prior work on wheeled legged and snake robot platforms has decoupled these functions through dedicated appendages body segmentation or grippers We present closed-loop loco-manipulation with our snake robot COBRA without this distinction such that any part of the body may make contact We develop a contact-implicit …
@misc{salagame2026autonomous,
title = {{Autonomous Whole-body Loco-manipulation Using a Snake Robot}},
author = {Adarsh Salagame and Henry Noyes and Haroon Hublikar and Alireza Ramezani},
year = {2026},
eprint = {2611.09777},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09777},
}
Dingkun Wei, Zehong Shen, Huaijin Pi, Xueyang Wang et al.
Abstract摘要
Humanoid whole-body control requires a robot to track diverse reference motions while maintaining balance under gravity, torque limits, contact switching, and external disturbances. Existing reinforcement-learning-based trackers typically learn control outputs that jointly address motion tracking and balance. However, similar reference poses can correspond to substantially different dynamic demands, requiring a unified policy to learn motion-dependent control requirements alongside balance corrections. This can complicate learning and lead to high-frequency fluctuations in control outputs. We …
@misc{wei2026dkwbc,
title = {{DKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body Control}},
author = {Dingkun Wei and Zehong Shen and Huaijin Pi and Xueyang Wang and Xing Zhu and Hujun Bao and Yujun Shen and Xiaowei Zhou},
year = {2026},
eprint = {2611.09605},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09605},
}
Tianshu Wu, Tianxiang Gui, Yue Chen, Tingxu Liu et al.
Abstract摘要
Scaling up visual humanoid loco-manipulation skill acquisition requires reducing the human effort needed as tasks and spatial configurations expand, particularly in demonstration collection and task-specific training design. We present SUGARv2, a unified framework for learning diverse visual humanoid loco-manipulation skills through interaction reference synthesis and a shared training pipeline. To obtain diverse references at low cost, we adapt reusable task templates to different spatial configurations, synthesizing coordinated robot and object motions. To reduce task-specific reward design …
@misc{wu2026sugarv,
title = {{SUGARv2: Learning Diverse Visual Humanoid Loco-Manipulation Skills from Synthesized Interaction References}},
author = {Tianshu Wu and Tianxiang Gui and Yue Chen and Tingxu Liu and Ziye Peng and Guangjie Yang and Xiangqi Kong and Ye Shi and others},
year = {2026},
eprint = {2611.09266},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09266},
}
Learning locomotion directly on a humanoid robot is a promising direction to avoid the sim-to-real gap. However, locomotion rewards in Reinforcement Learning (RL) typically track the base velocity, which cannot be measured onboard without an external motion capture system or well-calibrated state estimator. This is why we introduce \method{}, which learns humanoid locomotion from minutes of samples with a reward that only requires the onboard Inertial Measurement Unit (IMU). Instead of tracking the base velocity, STRIDE rewards the robot for following a commanded acceleration profile. To …
ZhenYu Zou, Xing Liu, ShiChuang Wang, YunLong Dong et al.
Abstract摘要
Humanoid robots are expected to take over strenuous material handling in manufacturing, logistics, and healthcare, where operating existing wheeled carts would let them move heavy loads without modifying the workplace. However, humanoid cart pushing remains challenging because the bimanual grasp closes a force-coupled loop in which the contact wrenches that drive the cart also disturb the robot's balance, and payload changes alter the coupled dynamics. We propose ForPC, a force-aware whole-body control framework that regulates cart velocity through two-stage reinforcement learning. A …
@misc{zou2026forpc,
title = {{ForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing}},
author = {ZhenYu Zou and Xing Liu and ShiChuang Wang and YunLong Dong and LiNan Deng and Feng Hua and Jun Wan and Cheng Cheng},
year = {2026},
eprint = {2611.08664},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08664},
}
Vongani H. Maluleke, James Ni, Kie Horiuchi, Lea Wilken et al.
Abstract摘要
Humanoid robots are typically trained as solitary agents, leaving them unable to coordinate physically with human partners who naturally anticipate and adjust to one another. To bridge this gap, we present Social Humanoid, a hierarchical system that generates and executes socially responsive whole-body motion on a 29-DoF Unitree G1 conditioned on a human partner's motion. Its high-level policy performs flexible sequence modeling of human and humanoid motion through a shared tokenized representation, generating reactive humanoid motion and anticipating the partner's motion when they are …
@misc{maluleke2026social,
title = {{Social Humanoid: Socially Responsive Human-Humanoid Interaction}},
author = {Vongani H. Maluleke and James Ni and Kie Horiuchi and Lea Wilken and Jiahui Lei and Evonne Ng and Roei Herzig and Jitendra Malik and others},
year = {2026},
eprint = {2611.08407},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08407},
}
Youhao Lyu, Chenghao Sun, Zhenxuan Ding, Shuting Zhang et al.
Abstract摘要
For perceptive humanoid locomotion, the relevance of terrain geometry depends on the robot's body state and how both evolve under its actions. Existing work has advanced terrain-aware sensory fusion and cross-modal prediction, improving how visual and proprioceptive information is combined and related. However, feature aggregation alone does not specify which cross-modal relationships should be learned, while prediction between simultaneous observations does not explicitly constrain their action-dependent evolution. We present Spatiotemporal Training with Explicit Prediction (STEP), a …
@misc{lyu2026step,
title = {{STEP: Spatiotemporal Training with Explicit Prediction for Perceptive Humanoid Locomotion}},
author = {Youhao Lyu and Chenghao Sun and Zhenxuan Ding and Shuting Zhang and Ziyang Dong and Wenhao Cui},
year = {2026},
eprint = {2611.08097},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08097},
}
Chenghao Sun, Youhao Lyu, Zizheng Li, Shuting Zhang et al.
Abstract摘要
We present a progressive imitation framework that learns terrain-appropriate humanoid parkour skills and subsequently adapts their hand--knee support. Human demonstrations provide traversal behaviors, but their contact patterns need not suit a robot's proportions or the available support surfaces. Our framework changes the role of demonstrations across three stages: privileged imitation produces physically executable robot references; terrain-indexed adversarial motion priors guide skill acquisition; and geometric support guidance with a reduced motion-prior weight enables contact adaptation. …
@misc{sun2026terrain,
title = {{Terrain-Indexed Motion Priors and Geometric Support for Extreme Humanoid Parkour}},
author = {Chenghao Sun and Youhao Lyu and Zizheng Li and Shuting Zhang and Ziyang Dong and Wenhao Cui},
year = {2026},
eprint = {2611.08056},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08056},
}
Qingzhou Lu, Chenxin Liu, Xinming Wang, Liu Cao et al.
Abstract摘要
Enabling vision-language models (VLMs) to control humanoid robots requires both reliable whole-body interfaces and knowledge of how to use the body that carries over across tasks. We present HITCH (Harness Interfaces To Command Humanoids), a harness in which perception returns the 3D position of any point the VLM selects in its image, and execution turns commands for the pelvis, torso, hands, walking, and gaze into whole-body motion with a learned tracker. We find that both measuring the scene and executing commands accurately raise task success. We then fine-tune a VLM on trajectories in …
@misc{lu2026hitch,
title = {{HITCH: Teaching a Vision-Language Model to Operate a Humanoid through a Whole-Body Harness}},
author = {Qingzhou Lu and Chenxin Liu and Xinming Wang and Liu Cao and Yanjiang Guo and Mengdi Xu and Jianyu Chen},
year = {2026},
eprint = {2611.08039},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08039},
}
Nikita Kuzmin, Georgii Demianchuk, Mariya Lezina, Yara Mahmoud et al.
Abstract摘要
Social touch interactions, such as handshakes, fist bumps, and hugs, require humanoid robots to infer human intent, generate contact-rich motions, and maintain safe physical contact. We present AffectiveVLA, an intent-gated Vision-Language-Action (VLA) pipeline deployed on a Unitree G1 humanoid with dexterous hands. A lightweight RGB-based model recognizes interaction intent from two onboard cameras and converts it into a language prompt for the VLA policy. Low-confidence predictions suppress motion execution, reducing unintended activations. During physical contact, interaction-specific …
@misc{kuzmin2026affectivevla,
title = {{AffectiveVLA: Safe and Intent-Aware VLA Control for Human-Humanoid Social Interaction}},
author = {Nikita Kuzmin and Georgii Demianchuk and Mariya Lezina and Yara Mahmoud and Iana Zhura and Miguel Altamirano Cabrera and Dzmitry Tsetserukou},
year = {2026},
eprint = {2611.07712},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07712},
}
Humanoid locomotion policies trained without an explicit smoothness objective tend to produce jerky, high-energy gaits that stress actuators. Vision-language models (VLMs) offer an appealing reward source: given a rendered rollout, a VLM returns an aesthetic quality score on [0,10] without task-specific feature engineering. We find, however, that aesthetic preference and physical smoothness point in opposite directions: across four production VLMs, clips with higher joint acceleration receive higher aesthetic scores (mean Pearson r=+0.66). We study PhySA (Physical Smoothness and Aesthetics), …
@misc{fu2026physa,
title = {{PhySA: Decoupling Aesthetic and Physical Rewards for Humanoid Locomotion}},
author = {Yunwan Fu and Sicheng He and Jiageng Mao and Yue Wang},
year = {2026},
eprint = {2611.07679},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07679},
}
Wheeled humanoid robots carry both arms on one actuated waist, usually without torque sensors, so the accuracy of both end-effectors depends on a single waist model. After a first stage that estimates the tracker and target extrinsics, we identify the waist and both arms in one problem from the target positions on both flanges, measured by an optical tracker. The kinematic model is complete, with waist-base, parallel-axis, and branch parameters; arm compliances are driven by measured torques and waist compliances by gravity torques of the robot's mass model; and Gaussian priors with …
@misc{zha2026whole,
title = {{Whole-Body Elasto-Geometrical Calibration of Wheeled Humanoid Robots Using a Shared-Waist Dual-Chain Model}},
author = {Pengxin Zha and Yi Lu and Haojun Ma and Jun Zhu and Minghao Qiu and Xiaoyan Wang and Wei Shan and Maoqing Yao},
year = {2026},
eprint = {2611.07560},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07560},
}
This paper presents the kinematic modeling and singularity analysis of a three-degree-of-freedom, double four-bar closed-loop quadruped leg mechanism. This leg is essential for transitioning between the two operational modes; therefore, it is examined in both joint and task spaces, leading to the development of its Jacobian matrix, which is vital for designing the gaits required for terrestrial mode and mid-air aerial reconfiguration (for aerial mode). Various gaits are devised for the system's motion in the MATLAB simulation, and their practical implementation on the single-leg test rig is …
@misc{raturi2026kinematic,
title = {{Kinematic analysis of a double-closed-loop four-bar chain-based leg for a multimodal quadruped-pentarotor robot}},
author = {Prashant Raturi and Santhakumar Mohan},
year = {2026},
eprint = {2611.07470},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07470},
}
This paper presents a layered control framework that recasts high-dimensional system dynamics as a composition of dynamically interconnected subsystems, with dynamic dual-arm manipulation on the humanoid serving as a representative application. The humanoid is decomposed into a single rigid body (SRB) representing the torso and floating-base dynamic models representing the articulated arms and lower-body. The rigid connections between arm and torso and the closed kinematic chain formed between the end-effectors during cooperative manipulation are represented as holonomic constraints. The …
Siyuan Hu, Abulikemu Abuduweili, Yue Meng, Shipeng Liu et al.
Abstract摘要
Humanoid robots have the potential to automate a wide range of physical tasks. Progress toward generalist humanoid autonomy, however, remains constrained by a data bottleneck: mapping pixels and free-form language to low-level joint commands requires paired supervision, and no corpus of such data exists at sufficiently large scale. Some approaches rely on human teleoperation and wearable capture devices, which require at least an hour of human time to produce an hour of data. We present HALO, a humanoid autonomy system powered by a pipeline that synthesizes vision-language-motion data across …
@misc{hu2026perceptive,
title = {{Perceptive Humanoid Motion from Simulation}},
author = {Siyuan Hu and Abulikemu Abuduweili and Yue Meng and Shipeng Liu and Quentin Leboutet and Chen Chen and Matthias M\"uller and Vladlen Koltun},
year = {2026},
eprint = {2611.06908},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06908},
}
Sourav Selvaraj, Juan Pablo Espinosa Chessal, Mahdi Agheli
Abstract摘要
Wiping, sanding, and inspection robots press a sliding tool on a surface with controlled force. Controlling that force needs the surface normal, which is unknown on real-world, unmapped surfaces. Surface-normal estimators can help estimate the surface normal, but they either need a friction model or assume that the instantaneous tool velocity lies along the surface; however, walking violates this assumption. We propose SCRUB, a contact-based surface-frame estimator, which lets an armless quadruped slide a body-mounted tool along near-vertical surfaces in the absence of known, axis-aligned, or …
@misc{selvaraj2026scrub,
title = {{SCRUB: Contact-Based Surface-Frame Estimation for Force-Motion Control with an Armless Quadruped}},
author = {Sourav Selvaraj and Juan Pablo Espinosa Chessal and Mahdi Agheli},
year = {2026},
eprint = {2611.06649},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06649},
}
Haoyu Shao, Yunshen Wang, Jingze Zhang, Shaohang Zhu et al.
Abstract摘要
Humanoid soccer requires a robot to seamlessly combine agile locomotion, ball manipulation, and impact-rich contact dynamics across a continuous, dynamic behavioral spectrum. Yet existing approaches either cover only a limited set of skills, or forego motion priors entirely and produce unnatural behaviors, leaving the combination of a unified skill set and human-like motion quality unaddressed. We present soccerbot, a hierarchical reinforcement learning framework that resolves both limitations through a structured latent motion prior distilled from large-scale human soccer motion capture …
@misc{shao2026playing,
title = {{Playing Soccer Like Humans Do: Learning Versatile and Agile Humanoid Soccer via Latent Motion Priors}},
author = {Haoyu Shao and Yunshen Wang and Jingze Zhang and Shaohang Zhu and Yong-Lu Li and Tianzhu Zhang and Baoxiong Jia and Siyuan Huang},
year = {2026},
eprint = {2611.06564},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06564},
}
Xuemin Hu, Kaixin Li, Qingqing Xia, Yuxin Wang et al.
Abstract摘要
Coordinated whole-body control of humanoid robots remains challenging due to high-dimensional coupled dynamics and substantial physical heterogeneity across joint degrees of freedom (JDoFs). Existing multi-agent reinforcement learning methods either employ coarse agent partitions or directly apply generic heterogeneous-agent algorithms, while hierarchical controllers commonly optimize high-level planning and low-level execution independently, leading to uncoordinated motion. In this paper, we propose H^2MARL, a hierarchical heterogeneous multi-agent reinforcement learning framework with two …
@misc{hu2026h,
title = {{H^2MARL: Hierarchical Heterogeneous Multi-agent Reinforcement Learning for Humanoid Robot Control}},
author = {Xuemin Hu and Kaixin Li and Qingqing Xia and Yuxin Wang and Wei Xie and Bo Tang and Long Chen},
year = {2026},
eprint = {2611.06523},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06523},
}
Roberto Mauceri, Stefano Dafarra, Giulio Romualdi, Marco Gabiccini et al.
Abstract摘要
This letter presents a trajectory planner for humanoid robots to perform dynamic motions on complex, parameterizable terrains using a contact-implicit whole-body optimization framework. The planner generates feasible movements without predefined contact sequences, considering the robot's full dynamics and kinematics. To plan bipedal locomotion on steps, ramps, and other terrains, we propose a mathematical model to approximate ground geometry with a smooth surface. The rigid interaction between the robot and the ground is modeled via novel contact constraints, extending the domain of …
Mengjie Zhao, Yuxin Chen, Yiyang Shao, Chuye Hong et al.
Abstract摘要
Humanoid motion tracking has scaled to general policies that reproduce diverse human motion, but the resulting behavior is in free-space: the tracker knows where the body should be, not how it should interact with the environment. Other work makes the object part of the objective, yet abstracts that interaction to contact or geometry, which leaves the demanding cases under-explored. We argue that force is a natural command channel between motion imitation and object interaction, one that specifies not just where to make contact but how hard. We present Ctrl-F, a hierarchical motion-force …
@misc{zhao2026ctrl,
title = {{Ctrl-F: Force as an Interface for Humanoid Loco-Manipulation}},
author = {Mengjie Zhao and Yuxin Chen and Yiyang Shao and Chuye Hong and Jiakang Chen and Koushil Sreenath and Masayoshi Tomizuka},
year = {2026},
eprint = {2611.05507},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05507},
}
Jean-Pierre Sleiman, Bilal Hammoud, He Li, Jonathan Foster et al.
Abstract摘要
Hierarchical pipelines for humanoid loco-manipulation typically pair a task-agnostic whole-body controller (WBC) with a downstream task policy. Such WBCs are often trained through manual reward shaping, which yields visibly machine-like motions; more consequentially, they are trained without the manipulated object in the scene. Object dynamics are instead approximated through randomized end-effector force disturbances, a proxy that fails to reproduce the distributed multi-contact wrench a real payload applies across the arms and torso, with the gap becoming critical near the platform's …
@misc{sleiman2026hlo,
title = {{HLO-AMP: Hierarchical Loco-Manipulation from Object-Aware Motion Priors}},
author = {Jean-Pierre Sleiman and Bilal Hammoud and He Li and Jonathan Foster and Sylvain Bertrand and Jessica K. Hodgins and Farbod Farshidian},
year = {2026},
eprint = {2611.05113},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05113},
}
Abhijeet M. Kulkarni, Saimouli Katragadda, Yuze Du, Zijun Sha et al.
Abstract摘要
By harnessing the inherent commonsense capabilities of pretrained Large Language Models (LLMs) and Vision-Language Models (VLMs), current Vision-Language Action(VLA) models for legged robots excel at interpreting language-based navigation commands but lack the physical grounding and explicit safety assurance required for real-world operation. Therefore, this paper aims to cage Legged-Vision-Language Navigation (VLN) through a safety architecture that provides physical grounding and ensures navigation safety. The key idea is not to abandon the powerful capabilities of end-to-end VLA planning, …
@misc{kulkarni2026caging,
title = {{Caging Vision-Language Navigation for Legged Robots via Barrier-Certified Safety Architecture}},
author = {Abhijeet M. Kulkarni and Saimouli Katragadda and Yuze Du and Zijun Sha and Guoquan Huang},
year = {2026},
eprint = {2611.04972},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04972},
}
Ju Dong, Xinyu Xiao, Yiming Zhou, Tianshuo Yang et al.
Abstract摘要
Humanoid loco-manipulation requires coordinated body motion and hand interaction over time, and a plausible-looking command sequence can still drop a held object or leave an interaction incomplete. We present action-conditioned latent foresight, a latent world-action model (WAM) that evaluates proposed command sequences before execution. An action generator proposes joint motion-hand command sequences from vision, language, and the robot state. An action-conditioned endpoint predictor estimates the visual features each sequence prefix reaches and measures progress toward a …
@misc{dong2026coordinating,
title = {{Coordinating Humanoid Loco-Manipulation through Action-Conditioned Latent Foresight}},
author = {Ju Dong and Xinyu Xiao and Yiming Zhou and Tianshuo Yang and Xuerong Liu and Lei Zhang and Jian Chen and Haocheng Zhao and others},
year = {2026},
eprint = {2611.04839},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04839},
}
Humanoid robots operating in cluttered environments must react to unexpected contacts before the first impact peak, which arrives within tens of milliseconds. Learned whole-body policies running at 50 Hz introduce a policy-cycle delay and may suppress a protective withdrawal as a tracking error. To address these challenges, we propose ARC (Ascending Reflex Control), a bottom-up framework that integrates instantaneous tactile reflexes with learned whole-body control. ARC consists of two components: an instantaneous reflex pathway and bottom-up reflex integration. The reflex pathway detects …
@misc{wang2026reflexarc,
title = {{ReflexArc: Bottom-Up Integration of Instantaneous Tactile Reflexes into Whole-Body Humanoid Control}},
author = {Mingyu Wang and Si Li and Ziqing Chen and He Wang and Li Yi},
year = {2026},
eprint = {2611.04424},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04424},
}
Donggeon Kim, Hyungho Chris Choi, Donghyuk Choi, Jeongsoo Park et al.
Abstract摘要
This paper presents humanoid manipulator hardware designed specifically for human-level performance throughout its broad domain of tasks, from heavy payloads to highly dynamic motions. Unlike conventional designs targeting uniform workspace capability, the proposed approach derives force, speed, and endurance targets from published measurements of task-relevant human upper-limb performance. These targets guide the actuator, mechanism, and thermal design toward torque transparency and sustained operation. The platform integrates force-oriented upper-arm components, speed-oriented forearm …
@misc{kim2026raipal,
title = {{RAIPAL: A Powerful and Agile Humanoid Arm for Human Tasks}},
author = {Donggeon Kim and Hyungho Chris Choi and Donghyuk Choi and Jeongsoo Park and Choongin Lee and Chan Park and Jemin Hwangbo},
year = {2026},
eprint = {2611.04334},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04334},
}
Han Zhang, Sicheng Lu, Aleks Santari, Jeffrey K. Jopling et al.
Abstract摘要
Operating rooms (ORs) increasingly contain ambient sensing infrastructure to enable AI-based optimization of workflows and utilization. We pursue an alternative path: using ambient AI to guide robot action and optimize utilization. Deploying humanoids in the OR without ambient AI is severely limited by cumbersome de novo orientation and planning due to partial onboard observations. We posit that this can be achieved by coupling an ambient AI-based agentic orchestrator with the humanoid robotic actor, providing shared perception and execution feedback for humanoid autonomy in the OR. ActOR …
@misc{zhang2026actor,
title = {{ActOR: Ambient AI-Based Agentic Orchestration for Humanoid Autonomy in Operating Rooms}},
author = {Han Zhang and Sicheng Lu and Aleks Santari and Jeffrey K. Jopling and Masaru Ishii and Mathias Unberath},
year = {2026},
eprint = {2611.04255},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04255},
}
Humanoid locomotion requires deciding where to step and how to distribute contact forces. However, a disturbance can invalidate the nominal plan, requiring coordinated revision under constraints that switch with every footfall. To solve this online, the proposed approach employs the single-rigid-body model, which retains attitude dynamics and the full contact wrench. An iterative linear-quadratic-regulator MPC (iLQR-MPC) then jointly optimizes contact forces and footstep locations in a single receding-horizon problem, without a separate step planner. Specifically, contact and motion …
@misc{li2026online,
title = {{Online iLQR-MPC with Whole-Body Control for Robust Humanoid Locomotion}},
author = {Xiang Li and Linlin Ou and Ruihan Wu and Xinyi Yu},
year = {2026},
eprint = {2611.04204},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04204},
}
Tony Tao, Jason Jingzhou Liu, Ruslan Salakhutdinov, Deepak Pathak
Abstract摘要
General-purpose mobile manipulation requires a combination of capable robot hardware, scalable demonstration collection, expressive whole-body control, and effective policy learning. Yet these components are often developed independently, creating mismatches that limit the capabilities of the complete system. We present Pixel, a co-designed full-stack system for learning general-purpose mobile manipulation. Pixel combines a low-cost wheeled humanoid with human-aligned kinematics and compliant grippers, multiple teleoperation and robot-free human data-collection interfaces, a unified …
@misc{tao2026pixel,
title = {{Pixel: A Co-Designed Full-Stack Robot System for Learning General-Purpose Mobile Manipulation}},
author = {Tony Tao and Jason Jingzhou Liu and Ruslan Salakhutdinov and Deepak Pathak},
year = {2026},
eprint = {2611.03648},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03648},
}
Juho Kalliokoski, Evan G. Center, Steven M. LaValle, Timo Ojala et al.
Abstract摘要
Virtual reality experiments involving navigation and simulated robot motion often require large, realistic environments and numerous distinct trajectories to avoid repetitive navigation while providing sufficient variation for controlled experiments. Designing such scenarios manually is labor-intensive, particularly when trajectories must satisfy predefined geometric and kinematic constraints while remaining diverse and comparable across experimental conditions. This paper presents TRAILS, an automated framework for generating complete virtual navigation scenarios consisting of an environment …
@misc{kalliokoski2026trails,
title = {{TRAILS: Trajectory Generation for Automated Indoor Locomotion Studies}},
author = {Juho Kalliokoski and Evan G. Center and Steven M. LaValle and Timo Ojala and Basak Sakcak},
year = {2026},
eprint = {2611.03254},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03254},
}