The University of Hong Kong · embodied AI & robot learning香港大学 · 具身智能与机器人学习
9 curated papers, code, datasets and hardware from The University of Hong Kong (9 with code), with one-line takeaways.9 个来自香港大学的精选论文、代码、数据集与硬件（9 个附代码），均配一句话要点。
Labs are matched from each entry's affiliation field by simple name rules (data/labs.yaml); joint work appears under every lab named. Only curated entries are listed, because daily arXiv papers carry no affiliations.机构由每个条目的机构字段按名称规则匹配（data/labs.yaml），合作工作会出现在所有署名机构下。仅列出精选条目，因为每日 arXiv 论文不含机构信息。
A sparse-voxel radiance map built from LiDAR and camera data, aimed at robot logs — few viewpoints, imperfect poses and large scenes — while keeping geometry accurate and GPU memory in check.用激光雷达与相机数据构建稀疏体素辐射场地图，针对机器人数据视角少、位姿有误差、场景大的问题，在保证几何精度的同时控制显存占用。
@misc{liu2026sparv,
title = {{SpaRV: Sparse Radiance Voxels for Scalable Geometrically Consistent LiDAR-Visual Mapping and Rendering}},
author = {Jianheng Liu and Yunfei Wan and Fangming Cheng and Xiangyu Hong and Zuhao Zou and Haotian Li and Longji Yin and Chunran Zheng and Jiarong Lin and Fu Zhang},
year = {2026},
eprint = {2610.25665},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25665},
}
Distills a pretrained VLA teacher into a policy without action pretraining via on-policy supervision plus teacher-run recoveries, triggered by a reward model (AutoDAgger); reports up to 4.1x over BC on 3 real tasks.把预训练具身基础模型（教师）蒸馏到未做动作预训练的策略：在学生访问的状态上做在线监督，并由奖励模型监控进度、自动触发教师接管采集纠错轨迹（AutoDAgger，无需人工看护）；报告在 3 个真机任务上比行为克隆最高提升 4.1 倍。
@misc{lu2026tilde,
title = {{TILDE: Teacher-in-the-Loop Distillation from Embodied Foundation Models}},
author = {Yuxiang Lu and Yantong Liu and Jinghua Hou and Xueli Sun and Ruihua Han and Zhe Liu and Hengshuang Zhao},
year = {2026},
eprint = {2610.25376},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25376},
}
Online RL for VLAs with GRPO and only binary success rewards on OpenVLA-OFT, plus exploration tweaks; reports 99.1% on LIBERO and 70.4% on RoboTwin 1.0 (vs 39.8% for the SFT baseline).在 OpenVLA-OFT 上用 GRPO 与仅"成功/失败"二值奖励做在线强化学习，并加入若干促进探索的改动；论文报告 LIBERO 成功率 99.1%、RoboTwin 1.0 为 70.4%（SFT 基线 39.8%）。
@misc{li2025simplevla,
title = {{SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning}},
author = {Haozhan Li and Yuxin Zuo and Jiale Yu and Yuhao Zhang and Zhaohui Yang and Kaiyan Zhang and Xuekai Zhu and Yuchen Zhang and Tianxing Chen and Ganqu Cui and Dehui Wang and Dingxiang Luo and Yuchen Fan and Youbang Sun and Jia Zeng and Jiangmiao Pang and Shanghang Zhang and Yu Wang and Yao Mu and Bowen Zhou and Ning Ding},
year = {2025},
eprint = {2509.09674},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.09674},
}
Learns task-centric latent actions from videos (in DINO feature space, filtered by language) so one policy can be pretrained across robots and human video; reports beating OpenVLA with far less pretraining compute.在 DINO 特征空间中、借助语言指令过滤，从视频里学习“任务相关”的隐动作，使策略可跨机器人与人类视频预训练；报告以少得多的预训练算力超过 OpenVLA。
@misc{bu2025univla,
title = {{UniVLA: Learning to Act Anywhere with Task-centric Latent Actions}},
author = {Qingwen Bu and Yanting Yang and Jisong Cai and Shenyuan Gao and Guanghui Ren and Maoqing Yao and Ping Luo and Hongyang Li},
year = {2025},
eprint = {2505.06111},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.06111},
}
Apple Vision Pro teleoperation of two dexterous robot hands with real-time retargeting, collision and singularity avoidance, and low-cost vibration haptics for the operator.用 Apple Vision Pro 实时遥操作双灵巧手，带实时重定向、碰撞与奇异点规避，并为操作者提供低成本振动触觉反馈。
@misc{ding2024bunny,
title = {{Bunny-VisionPro: Real-Time Bimanual Dexterous Teleoperation for Imitation Learning}},
author = {Runyu Ding and Yuzhe Qin and Jiyue Zhu and Chengzhe Jia and Shiqi Yang and Ruihan Yang and Xiaojuan Qi and Xiaolong Wang},
year = {2024},
eprint = {2407.03162},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2407.03162},
}
Trains a large teacher only on synthetic images with precise depth, then pseudo-labels many real images to train smaller students; gives sharper, more robust monocular depth than V1.先只用深度精确的合成图像训练大教师模型，再给大量真实图像打伪标签来训练小模型；单目深度比 V1 更锐利、更鲁棒。
@misc{yang2024depth,
title = {{Depth Anything V2}},
author = {Lihe Yang and Bingyi Kang and Zilong Huang and Zhen Zhao and Xiaogang Xu and Jiashi Feng and Hengshuang Zhao},
year = {2024},
eprint = {2406.09414},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2406.09414},
}
Monocular depth model trained on tens of millions of pseudo-labelled images; a common off-the-shelf depth prior for robots.在数千万张伪标注图像上训练的单目深度模型，常被机器人系统直接拿来作为深度先验。
@misc{yang2024depth,
title = {{Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data}},
author = {Lihe Yang and Bingyi Kang and Zilong Huang and Xiaogang Xu and Jiashi Feng and Hengshuang Zhao},
year = {2024},
eprint = {2401.10891},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2401.10891},
}
FAST-LIVO2 — fuses LiDAR, camera and IMU in one filter with direct (feature-free) updates for both sensors, sharing a single voxel map where image patches are attached to LiDAR points.FAST-LIVO2：在同一滤波器中融合激光雷达、相机与 IMU，两种传感器都采用直接法（无需提取特征）更新，并共用一张体素地图，图像块挂载在激光点上。
@misc{zheng2024fast,
title = {{FAST-LIVO2: Fast, Direct LiDAR-Inertial-Visual Odometry}},
author = {Chunran Zheng and Wei Xu and Zuhao Zou and Tong Hua and Chongjian Yuan and Dongjiao He and Bingyang Zhou and Zheng Liu and Jiarong Lin and Fangcheng Zhu and Yunfan Ren and Rong Wang and Fanle Meng and Fu Zhang},
year = {2024},
eprint = {2408.14035},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2408.14035},
}
Registers raw LiDAR points directly to the map with an incremental k-d tree; fast enough for onboard drone and robot computers.用增量 k-d 树把原始激光点直接配准到地图，速度足以在无人机和机器人机载计算机上运行。
@misc{xu2021fast,
title = {{FAST-LIO2: Fast Direct LiDAR-inertial Odometry}},
author = {Wei Xu and Yixi Cai and Dongjiao He and Jiarong Lin and Fu Zhang},
year = {2021},
eprint = {2107.06829},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2107.06829},
}