Zhengtong Xu

Zhengtong Xu 徐政通

I am Zhengtong Xu, and I go by Tong. I'm a fourth-year PhD candidate in robot learning at Purdue Logo Purdue University, advised by Professor Yu She.

I'm also a part-time student researcher at Meta Logo Meta Reality Labs, where I work on dexterous manipulation policy learning. Before that, I did research internships at Meta Reality Labs and MERL Logo MERL.

I was a recipient of the 2025 Magoon Graduate Student Research Excellence Award at Purdue University (awarded to only 25 PhD students across the entire Purdue College of Engineering).

I received my Bachelor's degree in mechanical engineering at HUST Logo Huazhong University of Science and Technology.

G. Scholar / Twitter / Github / LinkedIn

News

[Jan. 2026]I received the Bilsland Dissertation Fellowship ($62,513.22) from Purdue University to support my final year of PhD studies.
[Oct. 2025]One paper accepted by IEEE T-RO.
[Apr. 2025]I received the Magoon Graduate Student Research Excellence Award at Purdue University (awarded to only 25 PhD students across the entire College of Engineering).
[Nov. 2024]I have passed my PhD preliminary exam and officially become a PhD candidate.
[Sep. 2024]One paper accepted by IEEE T-RO.

Research

* indicates equal contribution.

My research focuses on developing scalable neural-symbolic multimodal learning frameworks to enable robots to perform everyday manipulation tasks with human-level proficiency and dexterity. Specifically, I investigate:

Scalable robot learning paradigms based on vision, tactile, 3D, and their multimodal fusion.
Neural-symbolic approaches for building generalizable, interpretable, and robust robot learning systems, including differentiable optimization and agentic planning.

	DiffOG: Differentiable Policy Trajectory Optimization with Generalizability Zhengtong Xu, Zichen Miao, Qiang Qiu, Zhe Zhang, Yu She IEEE Transactions on Robotics (T-RO), 2025 website / arXiv / video / code(soon) / bibtex @misc{xu2025diffog, title={{DiffOG}: Differentiable Policy Trajectory Optimization with Generalizability}, author={Zhengtong Xu and Zichen Miao and Qiang Qiu and Zhe Zhang and Yu She}, year={2025}, eprint={2504.13807}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2504.13807}, } Close DiffOG introduces a transformer-based differentiable trajectory optimization framework for action refinement in imitation learning.
	LeTac-MPC: Learning Model Predictive Control for Tactile-reactive Grasping Zhengtong Xu, Yu She IEEE Transactions on Robotics (T-RO), 2024 arXiv / video / code / bibtex @article{xu2024letac, author={Xu, Zhengtong and She, Yu}, journal={IEEE Transactions on Robotics}, title={{LeTac-MPC}: Learning Model Predictive Control for Tactile-Reactive Grasping}, year={2024}, volume={40}, number={}, pages={4376-4395}, doi={10.1109/TRO.2024.3463470} } Close A generalizable end-to-end tactile-reactive grasping controller with differentiable MPC, combining learning and model-based approaches.
	UNIC: Learning Unified Multimodal Extrinsic Contact Estimation Zhengtong Xu, Yuki Shirai IEEE International Conference on Robotics and Automation (ICRA), 2026 arXiv / video / code(soon) / bibtex @misc{xu2026unic, title={UNIC: Learning Unified Multimodal Extrinsic Contact Estimation}, author={Zhengtong Xu and Yuki Shirai}, year={2026}, eprint={2601.04356}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2601.04356}, } Close A unified multimodal framework for extrinsic contact estimation that operates without any prior knowledge or camera calibration.
	Canonical Policy: Learning Canonical 3D Representation for Equivariant Policy Zhiyuan Zhang, Zhengtong Xu, Jai Nanda Lakamsani, Yu She IEEE Transactions on Robotics (T-RO), conditionally accepted, 2026 website / arXiv / code / bibtex @misc{zhang2025canonical, title={Canonical Policy: Learning Canonical 3D Representation for Equivariant Policy}, author={Zhiyuan Zhang and Zhengtong Xu and Jai Nanda Lakamsani and Yu She}, year={2025}, eprint={2505.18474}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2505.18474}, } Close Canonical Policy enables equivariant observation-to-action mappings by grouping both in-distribution and out-of-distribution point clouds to a canonical 3D representation.
	ManiFeel: Benchmarking and Understanding Visuotactile Manipulation Policy Learning Quan Khanh Luu, Pokuang Zhou, Zhengtong Xu, Zhiyuan Zhang, Qiang Qiu, Yu She Under Review, 2025 New England Manipulation Symposium (Oral)*, 2025 website / arXiv / code(soon) / bibtex @misc{luu2025manifeel, title={ManiFeel: Benchmarking and Understanding Visuotactile Manipulation Policy Learning}, author={Quan Khanh Luu and Pokuang Zhou and Zhengtong Xu and Zhiyuan Zhang and Qiang Qiu and Yu She}, year={2025}, eprint={2505.18472}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2505.18472}, } Close ManiFeel is a reproducible and scalable simulation benchmark for studying supervised visuotactile policy learning.
	UniT: Data Efficient Tactile Representation with Generalization to Unseen Objects Zhengtong Xu, Raghava Uppuluri, Xinwei Zhang, Cael Fitch, Philip Glen Crandall, Wan Shou, Dongyi Wang, Yu She IEEE Robotics and Automation Letters (RA-L), 2025 website / arXiv / video / code / bibtex @misc{xu2025unit, title={{UniT}: Data Efficient Tactile Representation with Generalization to Unseen Objects}, author={Zhengtong Xu and Raghava Uppuluri and Xinwei Zhang and Cael Fitch and Philip Glen Crandall and Wan Shou and Dongyi Wang and Yu She}, year={2025}, eprint={2408.06481}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2408.06481}, } Close Learn a tactile representation with generalizability only by a single simple object.
	VILP: Imitation Learning with Latent Video Planning Zhengtong Xu, Qiang Qiu, Yu She IEEE Robotics and Automation Letters (RA-L), 2025 arXiv / video / code / bibtex @misc{xu2025vilp, title={VILP: Imitation Learning with Latent Video Planning}, author={Zhengtong Xu and Qiang Qiu and Yu She}, year={2025}, eprint={2502.01784}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2502.01784}, } Close VILP integrates the video generation model into policies, enabling the representation of multi-modal action distributions while reducing reliance on extensive high-quality robot action data.
	Safe Human-Robot Collaboration with Risk-tunable Control Barrier Functions Vipul K. Sharma, Pokuang Zhou, Zhengtong Xu, Yu She, S. Sivaranjani IEEE/ASME Transactions on Mechatronics (TMECH)*, 2025 paper / video / bibtex @ARTICLE{xu2025rtcbf, author={Sharma, Vipul K. and Zhou, Pokuang and Xu, Zhengtong and She, Yu and Sivaranjani, S.}, journal={IEEE/ASME Transactions on Mechatronics}, title={Safe Human–Robot Collaboration With Risk Tunable Control Barrier Functions}, year={2025}, doi={10.1109/TMECH.2025.3572047}} Close We address safety in human-robot collaboration with uncertain human positions by formulating a chance-constrained problem using uncertain control barrier functions.
	LeTO: Learning Constrained Visuomotor Policy with Differentiable Trajectory Optimization Zhengtong Xu, Yu She IEEE Transactions on Automation Science and Engineering (T-ASE), 2024 arXiv / video / code / bibtex @article{xu2024leto, title={LeTO: Learning Constrained Visuomotor Policy with Differentiable Trajectory Optimization}, author={Xu, Zhengtong and She, Yu}, journal={arXiv preprint arXiv:2401.17500}, year={2024} } Close LeTO is a "gray box" method which marries optimization-based safety and interpretability with representational abilities of neural networks.
	VisTac: Toward a Unified Multimodal Sensing Finger for Robotic Manipulation Sheeraz Athar^, Gaurav Patel^, Zhengtong Xu, Qiang Qiu, Yu She IEEE Sensors Journal, 2023 paper / video / bibtex @article{athar2023vistac, title={Vistac towards a unified multi-modal sensing finger for robotic manipulation}, author={Athar, Sheeraz and Patel, Gaurav and Xu, Zhengtong and Qiu, Qiang and She, Yu}, journal={IEEE Sensors Journal}, year={2023}, publisher={IEEE} } Close VisTac seamlessly combines high-resolution tactile and visual perception in a single unified device.

Awards

Bilsland Dissertation Fellowship, Purdue University, 2026
(valued at $62,513.22, supporting final year of PhD studies)
Magoon Graduate Student Research Excellence Award, Purdue University, 2025
(awarded to only 25 PhD students across the entire Purdue College of Engineering)
Dr. Theodore J. and Isabel M. Williams Fellowship, Purdue University, 2022
National Scholarship, Ministry of Education of China, 2017

Reviewer Service

Conference on Robot Learning (CoRL)
IEEE Robotics and Automation Letters (RA-L)
IEEE Transactions on Robotics (T-RO)
IEEE International Conference on Robotics and Automation (ICRA)

Teaching

Vertically Integrated Projects (VIP)-GE Robotics and Autonomous Systems, Grad Mentor, Spring 2024/Fall 2023/Summer 2023
IE 474-Industrial Control Systems, Teaching Assistant, Fall 2022

Website template from Jon Barron's website.