Changyi Lin

I am a second-year PhD student at CMU Safe AI Lab, advised by Prof. Ding Zhao. Previously, I was a research assistant at Tsinghua Embodied AI Lab, where I gained my initial research experience under the guidance of Prof. Huazhe Xu. I obtained my bachelor degree in ME&CS from Huazhong University of Science and Technology (HUST).

Email | Google Scholar | Github | Twitter | LinkedIn

Research

I am passionate about robotics, with a research focus on whole-body loco-manipulation control, visual-tactile sensing, intelligent hardware design, and reinforcement learning. My goal is to develop robotic systems that can fully perceive, deeply understand, and safely interact with the physical world.

	LocoTouch: Learning Dexterous Quadrupedal Transport with Tactile Sensing Changyi Lin, Yuxin Ray Song, Boda Huo, Mingyang Yu, Yikai Wang, Shiqi Liu, Yuxiang Yang, Wenhao Yu, Tingnan Zhang, Jie Tan, Yiyue Luo, and Ding Zhao Preprint 2025 paper \| website \| video \| code (coming) \| X \| media(IEEE Spectrum) Tactile sensing unlocks a new level of contact-rich interaction for legged robots! Equipped with a high-density distributed tactile sensor, our quadrupedal robot can transport everyday objects without mounting or strapping. The transport policy achieves zero-shot sim-to-real transfer, featuring two task-agnostic components: high-fidelity tactile simulation and robust, symmetric, frequency-adaptive gaits.
	Human2LocoMan: Learning Versatile Quadrupedal Manipulation with Human Pretraining Yaru Niu, Yunzhe Zhang, Mingyang Yu, Changyi Lin, Chenhao Li, Yikai Wang, Yuxiang Yang, Wenhao Yu, Tingnan Zhang, Zhenzhen Li, Jonathan Francis, Bingqing Chen, Jie Tan, and Ding Zhao RSS 2025 paper \| website \| video \| code \| X We enable LocoMan to autonomously perform various household tasks in multiple operation modes. We train the policy more efficiently through pretraining on human demonstrations and fine-tuning with teleoperated robot data.
	QuietPaw: Learning Quadrupedal Locomotion with Versatile Noise Preference Alignment Yuyou Zhang, Yihang Yao, Shiqi Liu, Yaru Niu, Changyi Lin, Yuxiang Yang, Wenhao Yu, Tingnan Zhang, Jie Tan, and Ding Zhao Preprint 2025 paper By incorporating conditional policy learning with noise constraints, we enable quadrupedal robots to adaptively adjust their locomotion behaviors based on diverse noise preferences using a single policy.
	DTactive: A Vision-Based Tactile Sensor with Active Surface Jikai Xu, Lei Wu, Changyi Lin, Ding Zhao, and Huazhe Xu Preprint 2024 paper \| website \| hardware \| X We extend 9DTact with an active surface, enabling efficient continuous in-hand rotation with precise angular pose estimation across diverse objects.
	Agile Continuous Jumping in Discontinuous Terrains Yuxiang Yang, Guanya Shi, Changyi Lin, Xiangyun Meng, Rosario Scalise, Mateo Guaman Castro, Wenhao Yu, Tingnan Zhang, Ding Zhao, Jie Tan, Byron Boots ICRA 2025 paper \| website \| video \| code \| X By augmenting the GPU-accelerated RL-QP hierarchical framework (CAJun) with perception and precise system identification, we enable quadrupedal robots to perform continuous, terrain-adaptive jumping over challenging environments.
	LocoMan: Advancing Versatile Quadrupedal Dexterity with Lightweight Loco-Manipulators Changyi Lin, Xingyu Liu, Yuxiang Yang, Yaru Niu, Wenhao Yu, Tingnan Zhang, Jie Tan, Byron Boots, and Ding Zhao IROS 2024, Featured Cover on IEEE Spectrum paper \| website \| video \| hardware & code \| X \| media(IEEE Spectrum Cover, TechXplore, CMU Engineering News) Limbs are not just for walking, they can be arms too! We reimagine quadrupedal limbs as 6-DoF arms by equipping them with custom-designed manipulators that mimic huma wrists. Powered by a unified whole-body controller, LocoMan can perform a wide range of dexterous manipulation tasks.
	ArrayBot: Reinforcement Learning for Generalizable Distributed Manipulation through Touch Zhengrong Xue, Han Zhang, Jingwen Cheng, Zhengmao He, Yuanchen Ju, Changyi Lin, Gu Zhang, and Huazhe Xu ICRA 2024 paper \| website \| video \| code \| X ArrayBot has 16x16 vertically sliding pillars, each equipped with a tactile sensor. By reshaping actions in the frequency domain, we train a policy that enables ArrayBot to translate a variety of everyday objects with tactile sensing.
	9DTact: A Compact Vision-Based Tactile Sensor for Accurate 3D Shape Reconstruction and Generalizable 6D Force Estimation Changyi Lin, Han Zhang, Jikai Xu, Lei Wu, and Huazhe Xu RAL & ICRA 2024 paper \| website \| video \| hardware & code \| dataset \| X \| tutorial-en \| tutorial-cn 9D = 3D Shape Reconstruction + 6D Force Estimation! Only white light! No markers! One-shot calibration! Easy fabrication! Open source! So many features to explore! Why not check out the website?
	DTact: A Vision-Based Tactile Sensor that Measures High-Resolution 3D Geometry Directly from Darkness Changyi Lin, Ziqi Lin, Shaoxiong Wang, and Huazhe Xu ICRA 2023 paper \| website \| video \| hardware & code \| X \| media(Technology) The deeper you press, the darker you see! We leverage the reflection property of translucent silicone to reconstruct 3D contact geometry using only white light. We also demonstrate its robustness and feasibility for non-planar contact surfaces.

Projects

	Generalizable BeamWalking for Legged Robots with Reinforcement Learning Course Project for Intro to Robot Learning(16-831 by Deepak Pathak) Techniques: Reinforcement Learning, Curriculum Learning, Computer Vision
	RoboMaster Robotics Competition Engineer Robot 2019, Dart System 2020, SLAM System 2021 Techniques: Mechanical Design, Control, SLAM, Computer Vision

Talks

[2024/05/01] Invited Talk at CMU Manipulation Seminar

[2024/04/26] Invited Talk at CMU Locomotion Seminar

Academic Services

Conference Reviewer: CoRL, ICRA, IROS, Humanoids
Journal Reviewer: RAL, TMECH

Template from Jon Barron's website.