Zhongyu Li

About Me

Hi there! I am Ph.D. student in Controls and Robotics at University of California, Berkeley, working with Prof. Koushil Sreenath. Before that, I spent a wonderful year doing research with Prof. Ralph Hollis at Carnegie Mellon University. I obtained my B.Eng. of Mechatronics at Zhejiang University in 2019.

My goal is to create safe, intelligent, and agile legged robots for better human companions. My current research is primarily centered on tackling the control and planning of legged robots using a combination of model-based optimal control and model-free reinforcement learning. I am also interested to expand the boundaries of legged robots beyond just locomotion.

I am always happy to chat or collaborate with people with different backgrounds. If you are interested in my work, please feel free to reach out!

Google Scholar: CLICK HERE
Linkedin: CLICK HERE

Twitter: CLICK HERE
Email: CLICK HERE

Research Interests

I am interested in creating new generations of dynamic robots, such as humanoids and other bio-inspired robots, to make them more intelligent, agile, robust, and safe, and ultimately, useful. My research lies in the combination of control, optimization, motion planning, reinforcement learning, imitation learning, and multi-agent interaction.

Research Highlights

My work has enabled a bipedal robot Cassie to perform robust, agile, and highly dynamic maneuvers such as walking, jumping, and running, and to safely and autonomously navigate in unknown and cluttered environments. I have also expanded the capacities of quadrupedal robots beyond locomotion by performing various loco-manipulation tasks, such as functioning as guide dogs, soccer ball shooters and goalkeepers, collaborative agents, and more.

Honors and Awards

Selected as one of the Rising Stars in Mechanical Engineering (in the US) in 2023
Selected as one of 30 cohort of RSS Pioneers (cross the globe) in 2023
William S. Floyd, Jr. Graduate Student Fellowship at UC Berkeley, 2022
Best RoboCup Paper Finalist in IROS 2022
Graduate Division Block Grant Award in Mechanical Engineering at UC Berkeley, 2021
Best Service Robot Paper Finalist in ICRA 2021
Best Entertainment and Amusement Paper Finalist in IROS 2020
IROS Student and Developing Countries (SDC) Travel Award, 2019
Best Undergraduate Thesis Award at Zhejiang University, 2019

Invited Talks

Can We Bridge Model-based Control and Model-free RL on Legged Robots?

09/2022, GRASP SFI, University of Pennsylvania
09/2022, Mila – Quebec AI Institute
11/2022, Beijing Academy of Artificial Intelligence (BAAI)

Towards Safe, Robust, and Dynamic Legged Robots and Beyond

07/2023, Southern University of Science and Technology (SUSTech)
08/2023, Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University
08/2023, Huzhou Research Institute, Zhejiang University
02/2024, Seminar in the group of Prof. Guanya Shi, Carnegie Mellon University
03/2024, Seminar in the group of Prof. Ding Zhao, Carnegie Mellon University
04/2024, Berkeley Control Seminar, UC Berkeley
09/2024, AI Institute, Shanghai Jiao Tong University
10/2024, Seminar in the group of Prof. Sophia Shao, UC Berkeley
12/2024, Humanoid Intelligence Seminar, UC Berkeley

Towards General-Purpose Robots from the Perspectives of Legged Robots

06/2024, Guest Lecture in Deep Learning Course at IIIS, Tsinghua University

Publications

^*Equal Contribution, †Project Lead, Highlights

Peer-Reviewed Journals

	Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal Locomotion Control Z. Li, X. B. Peng, P. Abbeel, S. Levine, G. Berseth, K. Sreenath International Journal of Robotics Research (IJRR), 2024 Paper / Video / Media: MIT Technology Review A general RL framework for bipedal robot to perform various locomotion skills from standing to walking, jumping, and running.
	Autonomous Navigation of Bipedal Robots in Height-Constrained Environments Z. Li, J. Zeng, S. Chen and K. Sreenath International Journal of Robotics Research (IJRR), 2023 Paper / Video / Media: Video Friday A bipedal robot now can autonomously and safely navigate unknown and congested environments.
	Collaborative Navigation and Manipulation of a Cable-towed Load by Multiple Quadrupedal Robots C. Yang^, G. N. Sue^, Z. Li^, L. Yang, H. Shen, Y. Chi, A. Rai, J. Zeng, K. Sreenath IEEE Robotics and Automation Letters (RA-L), 2022 Paper / Video / Media: Video Friday A legged robot team that can tow a heavy load by cables*.

Journals in Review

Traversability-Aware Legged Navigation by Learning from Real-World Visual Data
H. Zhang, Z. Li†, X. Zeng, L. Smith, K. Stachowicz, D. Shah, L. Yue, Z. Song, W. Xia, S. Levine, K. Sreenath, Y. H. Liu
In Submission to IEEE Transactions on Robotics (T-RO), 2024
Paper / Video

Learn from real-world RGBD visual data to navigate with lower traversability cost for quadrupedal robots.

Peer-Reviewed Conference Papers

	DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Offline Datasets X. Huang^, Y. Chi^, R. Wang^, Z. Li†, X. B. Peng, S. Shao, B. Nikolic, K. Sreenath Conference on Robot Learning (CoRL)*, 2024 Paper / Website Diffusing multimodal legged locomotion skills purely from offline datasets with a 6.8M transformer DDPM (YES it runs onboard at 50Hz!).
	Learning Visual Quadrupedal Loco-Manipulation from Demonstrations Z. He, K. Lei, Y. Ze, K. Sreenath, Z. Li, H. Xu International Conference on Intelligent Robots and Systems (IROS), 2024 Paper / Website Imitation learning for end-effector planning + reinforcement learning whole-body contol for loco-manipulation tasks.
	Leveraging Symmetry in RL-based Legged Locomotion Control Z. Su^, X. Huang^, D. Ordoñez-Apraez, Y. Li, Z. Li, Q. Liao, G. Turrisi, M. Pontil, C. Semini, Y. Wu, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2024 Paper / Website The efficacy of two approaches to incorporate symmetry in RL for locomotion control.
	HiLMa-Res: A General Hierarchical Framework via Residual RL for Combining Quadrupedal Locomotion and Manipulation X. Huang, Q. Liao, Y. Ni, Z. Li†, L. Smith, S. Levine, X. B. Peng, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2024 Paper / Video A hierarchical RL framework to tackle various loco-manipulation tasks while performing continuous quadrupedal locomotion skills.
	Robust and Versatile Bipedal Jumping Control through Reinforcement Learning Z. Li, X. B. Peng, P. Abbeel, S. Levine, G. Berseth, K. Sreenath Robotics: Science and Systems (RSS), 2023 Paper / Video / Media: Video Friday A bipedal robot learns to perform various aggressive jumping maneuvers.
	i2LQR: Iterative LQR for Iterative Tasks in Dynamic Environments Y. Zeng^, S. He^, H. H. Nguyen, Z. Li, K. Sreenath, J. Zeng Conference on Decision and Control (CDC), 2023 Paper Iterative LQR for iterative tasks, analogous to learning MPC :)
	Walking in Narrow Spaces: Safety-critical Locomotion Control for Quadrupedal Robots with Duality-based Optimization Q. Liao, Z. Li†, A. Thirugnanam, J. Zeng, and K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2023 Paper / Video / Code A safety-critical legged locomotion controller to travel through narrow spaces.
	Creating a Dynamic Quadrupedal Robotic Goalkeeper with Reinforcement Learning X. Huang^, Z. Li^, Y. Xiang, Y. Ni, Y. Chi, Y. Li, L. Yang, X. B. Peng, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2023 Paper / Video / Media: IEEE Specturm • Tech Xplore • TechCrunch • DailyMail • DailyMail A agile quadrupedal goalkeeper that can intercept fast-moving ball.
	GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots G. Feng^, H. Zhang^, Z. Li†, X. B. Peng†, B. Basireddy, L. Yue, Z. Song, L. Yang, Y. Liu, K. Sreenath, S. Levine Conference on Robot Learning (CoRL), 2022 Paper / Video / Code An opensourced generalized locomotion controller for a vast number of quadrupedal robots.
	Hierarchical Reinforcement Learning for Precise Soccer Shooting Skills using Quadrupedal Robots Y. Ji^, Z. Li^, Y. Sun, X. B. Peng, S. Levine, G. Berseth, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2022 *Best RoboCup Paper Finalist, IROS 2022* Paper / Video / Media: Video Friday • Tech Xplore A quadrupedal robot that can precisely shoot a soft ball to the given target.
	Adapting Rapid Motor Adaptation for Bipedal Robots A. Kumar^, Z. Li^, J. Zeng, D. Pathak, K. Sreenath, J. Malik International Conference on Intelligent Robots and Systems (IROS), 2022 Paper / Video An adaptive bipedal locomotion controller by RL.
	Teaching Robots to Span the Space of Functional Expressive Motion A. Sripathy, A. Bobu, Z. Li, K. Sreenath, D. S. Brown, A. D. Dragan International Conference on Intelligent Robots and Systems (IROS), 2022 Paper / Video Humans can teach robot how to behave with emotions!
	Bridging Model-based Safety and Model-free Reinforcement Learning through System Identification of Low Dimensional Linear Models Z. Li, J. Zeng, A. Thirugnanam, K. Sreenath Robotics: Science and Systems (RSS), 2022 Paper / Video / Seminar A bipedal robot controlled by its RL policy is a linear system!
	Bayesian Optimization Meets Hybrid Zero Dynamics: Safe Parameter Learning for Bipedal Locomotion Control L. Yang^, Z. Li^, J. Zeng, K. Sreenath International Conference on Robotics and Automation (ICRA), 2022 Paper / Video Auto-tuning a HZD-based bipedal locomotion controller by Bayesian Optimization.
	Autonomous navigation for quadrupedal robots with optimized jumping through constrained obstacles S. Gilroy^, D. Lau^, L. Yang^, E. Izaguirre, K. Biermayer, A. Xiao, M. Sun, A. Agrawal, J. Zeng, Z. Li†, and K. Sreenath Conference on Automation Science and Engineering (CASE)*, 2021 Paper / Video / Media: Video Friday Quadrupedal robots can traverse more complex environments by leveraging a big jump.
	Enhancing Feasibility and Safety of Nonlinear Model Predictive Control with Discrete-Time Control Barrier Functions J. Zeng^, Z. Li^, and K. Sreenath Conference on Decision and Control (CDC), 2021 Paper Incorporating Control Barrier Functions (CBFs) with Model Predictive Control (MPC).
	Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots Z. Li, X. Cheng, X. Peng, P. Abbeel, S. Levine, G. Berseth and K. Sreenath International Conference on Robotics and Automation (ICRA), 2021 Paper / Video / Seminar / Media: MIT Technology Review • Tech Xplore • Inverse • MathWorks • heise (German) • DeepTech (Chinese) A bipedal robot learns to walk, turn, and squat, with significant robustness.
	Robotic Guide Dog: Leading a Human with Leash-Guided Hybrid Physical Interaction A. Xiao^, W. Tong^, L. Yang^, J. Zeng, Z. Li†, and K. Sreenath International Conference on Robotics and Automation (ICRA), 2021 Best Service Robot Paper Finalist, ICRA 2021* Paper / Video / Seminar / Media: New Scientist • Daily Mail • Tech Xplore • Daily Californian • Independent • Futurism • China Daily • DeepTech (Chinese) Quadrupedal robots can serve as guide dogs!
	Safety-Critical Control with Optimal-decay Control Barrier Functions with Guaranteed Point-wise Feasibility J. Zeng, B. Zhang, Z. Li, and K. Sreenath American Control Conference (ACC), 2021 Paper How can we ensure feasibility using Control Barrier Functions (CBFs) with bounded inputs?
	Animated Cassie: A Dynamic Relatable Robotic Character Z. Li, C. Cummings and K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2020 *Best Entertainment and Amusement Paper Finalist, IROS 2020* Paper / Video / Seminar / Media: Video Friday A bipedal robot can behave emotively to be a better human companion!
	Toward A Ballbot for Physically Leading People: A Human-Centered Approach Z. Li, and R. Hollis International Conference on Intelligent Robots and Systems (IROS), 2019 Paper / Video / Media: Video Friday A ballbot that can lead a blind-folded person to navigate environments. This is how my story begins.

Conference Papers in Review

	Long-horizon Locomotion and Manipulation on a Quadrupedal Robot with Large Language Models Y. Ouyang, J. Li, Y. Li, Z. Li, C. Yu, K. Sreenath, Y. Wu In Submission to International Conference on Robotics and Automation (ICRA), 2025 Paper LLM agents for long-horizon task and motion planning for loco-manipulation tasks using quadrupeds
	CurricuLLM: Automatic Task Curricula Design for Learning Complex Robot Skills using Large Language Models K. Ryu, Q. Liao, Z. Li, K. Sreenath, N. Mehr In Submission to International Conference on Robotics and Automation (ICRA), 2025 Paper / Code LLM agents for curricula design for RL.
	Berkeley Humanoid: A Research Platform for Learning-based Control Q. Liao, B. Zhang, X. Huang, X. Huang, Z. Li, K. Sreenath In Submission to International Conference on Robotics and Automation (ICRA), 2025 Paper / Website / Media: IEEE Specturm An in-house-built and cool humanoid (bipedal) robot.
	Learning Smooth Humanoid Locomotion through Lipschitz-Constrained Policies Z. Chen^, X. He^, Y. J. Wang^, Q. Liao, Y. Ze, Z. Li, S. Sastry, J. Wu, K. Sreenath, S. Gupta, X. B. Peng In Submission to International Conference on Robotics and Automation (ICRA)*, 2025 Paper / Website Using Lipschitz constraint to have smooth RL policy.
	Interactive Navigation with Adaptive Non-prehensile Mobile Manipulation C. Dai^, X. Liu^, K. Sreenath, Z. Li†, R. Hollis In Submission to International Conference on Robotics and Automation (ICRA), 2025 Paper / Website An adpative control framework for interactive navigation tasks.

Teaching

UC Berkeley	[DEWA]: Optimization & Machine Learning with Applications to Energy Systems	Graudate Level	Class Size:~30	2020-2024
UC Berkeley	[E7]: Introduction to Computer Programming for Scientists and Engineers	Undergraudate Level	Class Size:~200	Fall 2020
Zhejiang University	Voluntary teacher in an underrepresented primary school	K-12	Class Size:~160	Part time in 2014-2015

Academic Services

Journal Reviewer:
The International Journal of Robotics Research (IJRR), Transactions on Robotics (T-RO), Transactions on Mechatronics (T-Mech), Transactions on Pattern Analysis and Machine Intelligence (TPAMI), Robotics and Automation Letters (RA-L), Robotics & Automation Magazine (RAM), Transactions on Cognitive and Developmental Systems (TCDS), Frontiers in Neurorobotics, Transactions on Industrial Electronics, Journal of Zhejiang University (JZUSA).
Conference Reviewer:
Robotics: Science and Systems (RSS), International Conference on Robotics and Automation (ICRA), International Conference on Intelligent Robots and Systems (IROS), International Conference on Humanoid Robots (Humanoids), International Conference on Automation Science and Engineering (CASE), The Conference on Robot Learning (CoRL), Conference on Decision and Control (CDC), Control Conference Africa (CCA).

	DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Offline Datasets X. Huang^, Y. Chi^, R. Wang^, Z. Li†, X. B. Peng, S. Shao, B. Nikolic, K. Sreenath Conference on Robot Learning (CoRL)*, 2024 Paper / Website Diffusing multimodal legged locomotion skills purely from offline datasets with a 6.8M transformer DDPM (YES it runs onboard at 50Hz!).
	Learning Visual Quadrupedal Loco-Manipulation from Demonstrations Z. He, K. Lei, Y. Ze, K. Sreenath, Z. Li, H. Xu International Conference on Intelligent Robots and Systems (IROS), 2024 Paper / Website Imitation learning for end-effector planning + reinforcement learning whole-body contol for loco-manipulation tasks.
	Leveraging Symmetry in RL-based Legged Locomotion Control Z. Su^, X. Huang^, D. Ordoñez-Apraez, Y. Li, Z. Li, Q. Liao, G. Turrisi, M. Pontil, C. Semini, Y. Wu, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2024 Paper / Website The efficacy of two approaches to incorporate symmetry in RL for locomotion control.
	HiLMa-Res: A General Hierarchical Framework via Residual RL for Combining Quadrupedal Locomotion and Manipulation X. Huang, Q. Liao, Y. Ni, Z. Li†, L. Smith, S. Levine, X. B. Peng, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2024 Paper / Video A hierarchical RL framework to tackle various loco-manipulation tasks while performing continuous quadrupedal locomotion skills.
	Robust and Versatile Bipedal Jumping Control through Reinforcement Learning Z. Li, X. B. Peng, P. Abbeel, S. Levine, G. Berseth, K. Sreenath Robotics: Science and Systems (RSS), 2023 Paper / Video / Media: Video Friday A bipedal robot learns to perform various aggressive jumping maneuvers.
	i2LQR: Iterative LQR for Iterative Tasks in Dynamic Environments Y. Zeng^, S. He^, H. H. Nguyen, Z. Li, K. Sreenath, J. Zeng Conference on Decision and Control (CDC), 2023 Paper Iterative LQR for iterative tasks, analogous to learning MPC :)
	Walking in Narrow Spaces: Safety-critical Locomotion Control for Quadrupedal Robots with Duality-based Optimization Q. Liao, Z. Li†, A. Thirugnanam, J. Zeng, and K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2023 Paper / Video / Code A safety-critical legged locomotion controller to travel through narrow spaces.
	Creating a Dynamic Quadrupedal Robotic Goalkeeper with Reinforcement Learning X. Huang^, Z. Li^, Y. Xiang, Y. Ni, Y. Chi, Y. Li, L. Yang, X. B. Peng, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2023 Paper / Video / Media: IEEE Specturm • Tech Xplore • TechCrunch • DailyMail • DailyMail A agile quadrupedal goalkeeper that can intercept fast-moving ball.
	GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots G. Feng^, H. Zhang^, Z. Li†, X. B. Peng†, B. Basireddy, L. Yue, Z. Song, L. Yang, Y. Liu, K. Sreenath, S. Levine Conference on Robot Learning (CoRL), 2022 Paper / Video / Code An opensourced generalized locomotion controller for a vast number of quadrupedal robots.
	Hierarchical Reinforcement Learning for Precise Soccer Shooting Skills using Quadrupedal Robots Y. Ji^, Z. Li^, Y. Sun, X. B. Peng, S. Levine, G. Berseth, K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2022 *Best RoboCup Paper Finalist, IROS 2022* Paper / Video / Media: Video Friday • Tech Xplore A quadrupedal robot that can precisely shoot a soft ball to the given target.
	Adapting Rapid Motor Adaptation for Bipedal Robots A. Kumar^, Z. Li^, J. Zeng, D. Pathak, K. Sreenath, J. Malik International Conference on Intelligent Robots and Systems (IROS), 2022 Paper / Video An adaptive bipedal locomotion controller by RL.
	Teaching Robots to Span the Space of Functional Expressive Motion A. Sripathy, A. Bobu, Z. Li, K. Sreenath, D. S. Brown, A. D. Dragan International Conference on Intelligent Robots and Systems (IROS), 2022 Paper / Video Humans can teach robot how to behave with emotions!
	Bridging Model-based Safety and Model-free Reinforcement Learning through System Identification of Low Dimensional Linear Models Z. Li, J. Zeng, A. Thirugnanam, K. Sreenath Robotics: Science and Systems (RSS), 2022 Paper / Video / Seminar A bipedal robot controlled by its RL policy is a linear system!
	Bayesian Optimization Meets Hybrid Zero Dynamics: Safe Parameter Learning for Bipedal Locomotion Control L. Yang^, Z. Li^, J. Zeng, K. Sreenath International Conference on Robotics and Automation (ICRA), 2022 Paper / Video Auto-tuning a HZD-based bipedal locomotion controller by Bayesian Optimization.
	Autonomous navigation for quadrupedal robots with optimized jumping through constrained obstacles S. Gilroy^, D. Lau^, L. Yang^, E. Izaguirre, K. Biermayer, A. Xiao, M. Sun, A. Agrawal, J. Zeng, Z. Li†, and K. Sreenath Conference on Automation Science and Engineering (CASE)*, 2021 Paper / Video / Media: Video Friday Quadrupedal robots can traverse more complex environments by leveraging a big jump.
	Enhancing Feasibility and Safety of Nonlinear Model Predictive Control with Discrete-Time Control Barrier Functions J. Zeng^, Z. Li^, and K. Sreenath Conference on Decision and Control (CDC), 2021 Paper Incorporating Control Barrier Functions (CBFs) with Model Predictive Control (MPC).
	Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots Z. Li, X. Cheng, X. Peng, P. Abbeel, S. Levine, G. Berseth and K. Sreenath International Conference on Robotics and Automation (ICRA), 2021 Paper / Video / Seminar / Media: MIT Technology Review • Tech Xplore • Inverse • MathWorks • heise (German) • DeepTech (Chinese) A bipedal robot learns to walk, turn, and squat, with significant robustness.
	Robotic Guide Dog: Leading a Human with Leash-Guided Hybrid Physical Interaction A. Xiao^, W. Tong^, L. Yang^, J. Zeng, Z. Li†, and K. Sreenath International Conference on Robotics and Automation (ICRA), 2021 Best Service Robot Paper Finalist, ICRA 2021* Paper / Video / Seminar / Media: New Scientist • Daily Mail • Tech Xplore • Daily Californian • Independent • Futurism • China Daily • DeepTech (Chinese) Quadrupedal robots can serve as guide dogs!
	Safety-Critical Control with Optimal-decay Control Barrier Functions with Guaranteed Point-wise Feasibility J. Zeng, B. Zhang, Z. Li, and K. Sreenath American Control Conference (ACC), 2021 Paper How can we ensure feasibility using Control Barrier Functions (CBFs) with bounded inputs?
	Animated Cassie: A Dynamic Relatable Robotic Character Z. Li, C. Cummings and K. Sreenath International Conference on Intelligent Robots and Systems (IROS), 2020 *Best Entertainment and Amusement Paper Finalist, IROS 2020* Paper / Video / Seminar / Media: Video Friday A bipedal robot can behave emotively to be a better human companion!
	Toward A Ballbot for Physically Leading People: A Human-Centered Approach Z. Li, and R. Hollis International Conference on Intelligent Robots and Systems (IROS), 2019 Paper / Video / Media: Video Friday A ballbot that can lead a blind-folded person to navigate environments. This is how my story begins.

Zhongyu Li | 李钟毓