Ruyi Wang
Ruyi Wang (王濡翼)
Embodied AI | Robot Learning | Reinforcement Learning | Tongji University
975967579 [at] qq.com  /  2333059 [at] tongji.edu.cn
CV Google Scholar ORCID

About

I am Ruyi Wang (王濡翼), a master's student at Tongji University, working at the intersection of embodied AI, robot learning, reinforcement learning, and physics-informed machine learning.

My recent work focuses on humanoid and legged robot locomotion, including PPO-based policy training, imitation learning, motion retargeting, and Sim2Real / Sim2Sim transfer on platforms such as Isaac Gym, MuJoCo, and ROS2. I have internship experience in robotics algorithm development at AgiBot, Xiaomi Robotics, and Anker Innovations.

I also conduct AI4Science research on physics-structured world models for PEM water electrolyzers, with interests in multimodal time-series modeling, electrochemical diagnosis, and scientific machine learning.

I am actively seeking opportunities in Embodied AI, robotics, and reinforcement learning. Feel free to reach out via email.

Research Interests

  • Embodied AI & Robotics
    Humanoid / legged robot locomotion, loco-manipulation, whole-body control, motion retargeting.
  • Robot Learning
    Reinforcement learning, imitation learning, policy optimization, reward design, contact modeling.
  • Sim2Real & Simulation
    Isaac Gym, Isaac Sim, MuJoCo, ROS2, sim-to-sim / sim-to-real transfer and deployment.
  • AI4Science
    Physics-informed machine learning, multimodal time-series modeling, electrochemical diagnosis for hydrogen systems.

Education

Tongji University, Shanghai, China
  • M.S. in Energy and Power Engineering (New Energy Power Systems)
    Sep 2023 – Jul 2026 (expected)
  • B.Eng. in Vehicle Engineering (Automotive Electronics and Intelligence)
    Sep 2018 – Jun 2023

Experience

2023.09 – 2026.07 (expected)
M.S. Student & Researcher @ Tongji University
Embodied AI and AI4Science research.
Working on humanoid / legged robot learning, reinforcement learning, motion retargeting, and physics-informed modeling for PEM electrochemical systems.
Supervisors: Xuezhe Wei, Jiangong Zhu, Xiaoping Wu
2025.09 – 2026.01
Robotics Locomotion & Manipulation Algorithm Intern @ AgiBot, Shenzhen, China
Project: Whole-body Loco-manipulation (Sim2Real)
• Debugged AIMRT–ROS SDK communication for full-stack HW/SW integration.
• Developed dynamics-constrained motions with Crocoddyl + Pinocchio for dataset expansion.
• Built VR teleoperation whole-body controller and validated Sim2Sim / Sim2Real transfer.
Instructor: Shaohe Du
2025.04 – 2025.09
Embodied Locomotion Control Algorithm Intern @ Xiaomi Robotics, Beijing, China
Project: Biped Locomotion with Toe Modeling
• Investigated the role of toe DoF in biped locomotion and gait stability using Cassie-style simulation.
• Built motion data processing and visualization pipeline for RL-ready trajectories from retargeted mocap data.
• Integrated AMP imitation learning in Legged Gym and trained PPO locomotion policies in Isaac Gym.
• Optimized rewards, contact forces, and PPO hyperparameters for locomotion stability.
Instructor: Jianhua Feng
2025.01 – 2025.03
Embodied Perception Algorithm Intern @ Anker Innovations, Shanghai, China
Project: 3D Simulation Asset Generation & Multi-engine Survey
• Surveyed Gazebo, MuJoCo, Isaac Sim, Unreal, Unity, ROS2/LCM, and 3D/4D Gaussian Splatting pipelines.
• Implemented ROS2 control interfaces in Isaac Sim for robot arm command and sensor streaming.
• Built an end-to-end pipeline from image capture to 3DGS reconstruction and simulation scene generation.
Instructor: Minyi Chen
2024.03 – 2024.10
ADAS Functional Test & Automation Intern @ NIO, Shanghai, China
Project: Automated Testing for Production ADAS
• Deployed WorldSim and communication modules on Ubuntu for automated ADAS testing.
• Authored regulatory scenario test cases covering AEB, BSD/LCA, ACC, ALC, and parking features.
• Developed automation scripts for scenario execution, result collection, and report generation.
Instructor: Xinghai Huang
2023.04 – 2023.08
Data Communication Intern @ Porsche Engineering, Shanghai, China
Project: ADAS Simulation Scenarios & Communication Pipeline
• Built a dual-system CARLA + ROS co-simulation stack for cross-platform communication and integration.
• Reconstructed real-world road segments and implemented FCW / SCW scenarios in CARLA.
• Designed a multi-node middleware pipeline from perception to decision to QT dashboard visualization.
Instructor: Mojun Qian
2018.09 – 2023.06
B.Eng. Student @ Tongji University
Major in Vehicle Engineering, with early research experience in hydrogen energy and fuel cell systems.
Supervisor: Haifeng Dai

Publications ( / )

A Physics-Structured Latent State World Model for Ion Contamination Diagnosis in Proton Exchange Membrane Water Electrolyzers
Ruyi Wang, Jiangong Zhu, Haoyu Zhang, Chao Wang, Xiaoping Wu, Ming Wu, Jianqiang Xu, Hao Yuan, Wei Liu, Haifeng Dai, Xuezhe Wei
Journal of Power Sources, revision / under review, 2026
A Review of the Portability of Fault Diagnosis Methods from PEM Fuel Cells to PEM Water Electrolyzers
Ruyi Wang, Jiangong Zhu, Haoyu Zhang, Chao Wang, Xiaoping Wu, Ming Wu, Jianqiang Xu, Hao Yuan, Wei Liu, Haifeng Dai & Xuezhe Wei
Discover Applied Sciences, 2025

Selected Projects

KudouZala/ion_detect Research Project
Physics-structured machine learning framework for ion contamination diagnosis in PEM water electrolyzers. The project integrates voltage data, impedance spectroscopy (EIS), and physics-informed modeling to build a multimodal Transformer-based diagnostic system.
Python · PyTorch
KudouZala/GitHub Profile Repository Collection
Personal GitHub repositories including robotics learning experiments, reinforcement learning training pipelines, and AI4Science research code.
Robotics · RL · AI4Science

Selected Honors & Awards

  • Second Prize, 2024 New Energy Vehicle Battery AI Fault Prediction Competition (University Group)
  • National First Prize, Formula Student Electric China (“NIO Cup”)
  • Third Prize, Tongji University Physics Competition
Top