Hongjie (Tony) Fang 方泓杰

I am a third-year Ph.D. student @ Computer Science in Wu Wenjun Honorable Class, Shanghai Jiao Tong University (SJTU) & Shanghai Artificial Intelligence Laboratory, advised by Prof. Cewu Lu. Previously, I got my B. Eng. degree @ Computer Science and Engineering, and B. Ec. degree @ Finance from SJTU in 2022.

My research interests mainly lie on robotic fields, specifically, robotic manipulation, robot learning and grasping. I am currently a member of SJTU Machine Vision and Intelligence Group (MVIG). My goal is to enable robots to perform various tasks in the real world, improving the quality of human life.

Email  /  GScholar  /  Github  /  X  /  WeChat  /  CV

[Mar. 2025] AirExo-2 is released! Check our website for more details.
[Jan. 2025] Two papers (S2I and CAGE) are accepted by ICRA 2025.
[Nov. 2024] FoAR is released! Check your website for more details.
[Jun. 2024] RISE is accepted by IROS 2024.
[Jan. 2024] Four papers (AirExo, RH20T, Open X-Embodiment and AnyGrasp) are accepted by ICRA 2024.
[Oct. 2023] Open X-Embodiment is released! Proud of this wonderful collaboration in robotics community!
[Sept. 2023] AirExo is released! Check our website for more details.
[Jun. 2023] One paper is accepted by IROS 2023.
[Apr. 2023] AnyGrasp is accepted by T-RO.
[Feb, 2023] One paper is accepted by CVPR 2023.
[Feb, 2023] TransCG is accepted by ICRA 2023 as RA-L submission. See you in London!
[Jun, 2022] TransCG is accepted by RA-L.
[Aug, 2021] One paper is accepted by ICCV 2021.

Representative papers are highlighted. * denotes equal contribution.

Develop AirExo-2, an updated low-cost exoskeleton system for large-scale in-the-wild demonstration collection. By transforming the collected in-the-wild demonstrations into pseudo-robot demonstrations, our system addresses key challenges in utilizing in-the-wild demonstrations for downstream imitation learning in the real world. Propose RISE-2, a generalizable imitation policy that integrates 2D and 3D perceptions, outperforming previous imitation learning policies in both in-domain and out-of-domain tasks, even with limited demonstrations. By leveraging in-the-wild demonstrations collected and transformed by the AirExo-2 system, without the need for additional robot demonstrations, RISE-2 achieves comparable or superior performance to policies trained with teleoperated data, highlighting the potential of AirExo-2 for scalable and generalizable imitation learning.

Introduce AnyDexGrasp, an efficient approach for learning dexterous grasping with minimal data, advancing robotic manipulation capabilities across different robotic hands. Our results show a grasp success rate of 75-95% across three different robotic hands in real-world cluttered environments with over 150 novel objects, improving to 80-98% with increased training objects.

Propose FoAR, a force-aware reactive policy that combines high-frequency force/torque sensing with visual inputs to enhance the performance in contact-rich manipulation. Built upon the RISE policy, FoAR incorporates a multimodal feature fusion mechanism guided by a future contact predictor, enabling dynamic adjustment of force/torque data usage between non-contact and contact phases. Its reactive control strategy also allows FoAR to accomplish contact-rich tasks accurately through simple position control.

Propose MBA, a novel module that employs two cascaded diffusion processes for object motion generation and robot action generation under object motion guidance. Designed as a plug-and-play component, MBA can be flexibly integrated into existing robotic manipulation policies with diffusion action heads. Extensive experiments in both simulated and real-world environments demonstrate that our approach substantially improves the performance of existing policies across a wide range of manipulation tasks.

Introduce CAGE, a data-efficient generalizable robotic manipulation policy. With less than 50 demonstrations in the mono-distributed training environment, CAGE can effectively generalize to similar and unseen environments with different levels of distribution shifts (background, object and camera view changes), outperforming previous state-of-the-art policies. This work makes a step forward in developing data-efficient, scalable, and generalizable robotic manipulation policies.

Introduce S2I, a framework that selects and optimizes mixed-quality demonstration data at the segment level, while ensuring plug-and-play compatibility with existing robotic manipulation policies. With only 3 expert demonstrations for reference, S2I can improve the performance of various downstream policies when trained with mixed-quality demonstrations.

Propose RISE, an end-to-end baseline for real-world robot imitation learning, which predicts continuous actions directly from single-view point clouds. Trained with 50 demonstrations for each real-world task, RISE surpasses currently representative 2D and 3D policies by a large margin, showcasing significant advantages in both accuracy and efficiency.

Introduce the Open X-Embodiment Dataset, the largest robot learning dataset to date with 1M+ real robot trajectories, spanning 22 robot embodiments. Train large, transformer-based policies on the dataset (RT-1-X, RT-2-X) and show that co-training with our diverse dataset substantially improves performance.

Develop AirExo, a low-cost, adaptable, and portable dual-arm exoskeleton, for joint-level teleoperation and demonstration collection. Further leverage AirExo for learning with cheap demonstrations in the wild to improve sample efficiency and robustness of the policy.

Collect a dataset comprising over 110k contact-rich robot manipulation sequences across diverse skills, contexts, robots, and camera viewpoints, all collected in the real world. Each sequence in the dataset includes visual, force, audio, and action information, along with a corresponding human demonstration video. Put significant efforts in calibrating all the sensors and ensures a high-quality dataset.

Propose an approach for effective and robust flexible handover, which enables the robot to grasp moving objects with flexible motion trajectories with a high success rate. The key innovation of our approach is the generation of real-time robust grasp trajectories. Designs a future grasp prediction algorithm to enhance the system's adaptability to dynamic handover scenes.

Focus on semantic consistency instead of temporal smoothness of the predicted grasp poses during reactive grasping. Solve the reactive grasping problem in a target-referenced setting by tracking through generated grasp spaces.

Propose a powerful AnyGrasp model for general grasping, including static scenes and dynamic scenes. AnyGrasp can generate accurate, full-DoF, dense and temporally-smooth grasp poses efficiently, and it works robustly against large depth sensing noise.

Propose TransCG, a large-scale real-world dataset for transparent object depth completion, along with a depth completion method DFNet based on the TransCG dataset.

Propose graspness, a quality based on geometry cues that distinguishes graspable area in cluttered scenes, which can be measured by a look-ahead searching method. Propose a graspness model to approximate the graspness value for quickly detect grasps in practice.

Provides an easy and unified interface for robots, grippers, sensors and pedals.

Proposes that we can learn "jargons" like "ResNet" and "YOLO" from academic paper citation information, and such citation information can be regarded as the searching results of the corresponding "jargon". For example, when searching "ResNet", the engine should return the "Deep Residual Learning for Image Recognition", instead of the papers that contains word "ResNet" in their titles, as current scholar search engines commonly return.

Reviewer: ICRA 2023/2024/2025, IROS 2023/2024, ICLR 2025, RA-L.

Some of My Notes ---> Notes

The website is built upon this template.