Lu Shi
Postdoctoral Researcher,
Tsinghua University
email: shilu@air.tsinghua.edu.cn
Postdoctoral Researcher,
Tsinghua University
email: shilu@air.tsinghua.edu.cn
About me
I am a robotics researcher with nearly a decade of experience in robotics control systems, spanning from classical methods such as PID and model-based control to advanced intelligent control frameworks, including reinforcement learning (RL), model predictive control (MPC), and vision-language-action (VLA) models. My work bridges theory and practice across diverse robotic platforms, spans locomotion (quadrupeds, wheeled-leg robots, mobile platforms), manipulation (robotic arms, dexterous hands), aerial robots, and soft robots. During my Ph.D., I focused on combining model-based and data-driven approaches through operator-based control (Koopman Operator Theory). As a postdoctoral researcher, my primary work centers on building generalized Real-to-Sim-to-Real (RSR) pipelines, integrating world models, and advancing embodied AI systems capable of robust adaptation across tasks and environments.
I am currently a postdoctoral researcher at the Institute of AI Industry Research, Tsinghua University. Before joining Tsinghua, I obtained my degree of Ph.D. in the Department of Electrical and Computer Engineering at the University of California, Riverside (UCR). I completed my bachelor’s degree in Automation and Control at Xi’an Jiaotong University, Xi'an.
A quick shout-out to our new survey paper on an exciting (and still relatively new) topic — Koopman Operators in Robot Learning 👉 https://arxiv.org/pdf/2408.04200
You might want to check it out if:
1️⃣ You’re dealing with a highly nonlinear system or equations, and want a powerful tool for global linearization;
2️⃣ You’re working with a system that’s hard to model, but you have data, and you want to design a controller that’s interpretable and comes with safety guarantees.
We invited some of the most representative groups working on Koopman in robotics, covering both theoretical foundations and practical implementations.
If we’re working in the same field — you know what to do 😉 Cite it!
In the Classroom: Embodied Intelligence & Robot Learning
Recently, I had the pleasure of serving as an instructor for the Embodied Intelligence Training Camp hosted by Tsinghua AIR and Digua Robotics. In this role, I taught students about robot learning, Sim2Real techniques, and the principles and algorithms behind embodied intelligence, helping them gain hands-on experience with cutting-edge robotics research. The sessions combined theory with practical exercises, aiming to inspire curiosity and creativity in tackling real-world robotic challenges. You can check out the video of the course at the video link.
Spotlight:
Oct. 2024: Our paper "Bridging the Resource Gap: Deploying Advanced Imitation Learning Models onto Affordable Embedded Platforms" was accepted by IEEE ROBIO 2024.
Aug. 2024: Our survey paper "Koopman Operators in Robot Learning," in collaboration with Dr. Konstantinos Karydis, Dr. Jorge Cortes, Dr. Todd Murphey, Dr. Ian Abraham, Dr. Daniel Bruder, Dr. Masih Haseli, and Dr. Giorgos Mamakoukas, has been submitted to IEEE Transactions on Robotics (TRO).
July 2024: Our paper "Arm-Constrained Curriculum Learning for Loco-Manipulation of the Wheel-Legged Robot" was accepted for oral presentation at IEEE IROS 2024, placing in the top 10%.
Sept. 2023: I was honored to be accepted into the prestigious Shuimu Tsinghua Scholarship program!