Pengyuan Wang profile photo

Pengyuan Wang

Undergraduate Student
Zhejiang University
wpy[dot]wangpengyuan[at]gmail[dot]com


About Me

I am Pengyuan Wang, an undergraduate student in Robotics Engineering (Honors Program) at Zhejiang University, expected to graduate in 2027. Currently, I am a research intern at the THU Spatial Intelligence Lab, working with Prof. Yiming Li.

My long-term research vision is to develop machine learning-based control for dynamic systems, encompassing reinforcement learning, imitation learning, and the perception-action loop, to achieve high-level autonomy. I aim to fully leverage the hardware advantages of diverse robotic platforms to enable safe and autonomous deployment in both industrial and daily-life settings.

My recent work spans humanoid locomotion and soccer control, autonomous terrain exploration, agile UAV control, and embedded systems for RoboMaster robots.

RESEARCH PROJECTS

  1. End-to-End Humanoid Soccer Control via Depth-Based Active Perception
    End-to-End Humanoid Soccer Control via Depth-Based Active Perception
    Yixiao Huo*, Pengyuan Wang*, Jiakang Jin*, Yinan Han*, et al.
    Developed an end-to-end RL framework for a 25-DOF humanoid to perform complex soccer maneuvers. Integrated depth-based active perception and auxiliary prediction heads for robust sim-to-real hardware deployment.
    ongoing
  2. Autonomous Humanoid Exploration in Multi-Level Terrains
    Autonomous Humanoid Exploration in Multi-Level Terrains
    Yinan Han*, Yixiao Huo*, Pengyuan Wang*, Jiakang Jin*, et al.
    Designed a goal-free heuristic exploration policy for a 25-DOF humanoid using active depth sensing. Synthesized competitive reward dynamics to induce emergent, terrain-adaptive locomotion across complex environments.
    ongoing

COMPETITIONS

  1. RoboMaster University Series: Embedded System Architecture
    RoboMaster University Series: Embedded System Architecture
    Zhejiang University RoboMaster Team "Hello World"
    Architected the embedded software and multi-layer state machines for heterogeneous competitive robots. Developed high-bandwidth bit-field protocols and cascaded PID controllers to ensure deterministic, high-speed motion control.
    Technical Report

ENGINEERING PROJECTS

  1. General Motion Tracking on Noetix E1 Humanoid Platform
    General Motion Tracking on Noetix E1 Humanoid Platform
    Internship at Noetix Robotics
    Reproduced and optimized General Motion Tracking frameworks (BeyondMimic/HoloMotion) via Isaac Lab. Leveraged adaptive sampling and robust sim-to-real pipelines to achieve stable whole-body imitation of high-dynamic dance sequences.
  2. Agile UAV Flight Control via Deep Reinforcement Learning
    Agile UAV Flight Control via Deep Reinforcement Learning
    Research project at FAST Lab, Zhejiang University
    Built a high-mobility UAV control stack bridging PPO-based RL and cascaded PID. Achieved autonomous agile maneuvers (e.g., figure-8) using extensive domain randomization for robust sim-to-real hardware transfer.