XiyuanWang

Introduction

I am a postgraduate student at The Chinese University of Hong Kong, where my work focuses on embodied intelligence and humanoid robotics.

I completed my undergraduate studies in Robotics Engineering at Lanzhou University of Technology. Before joining CUHK, I worked as a research assistant under Professor Donglin Wang at Westlake University.

Research Interests

My research interests center on whole-body motion control for humanoid robots, as well as motion planning, simultaneous localization and mapping (SLAM), and perception for indoor and outdoor environments. More recently, I have been exploring world-model-based robotic brain perception and control (WAM), with the goal of improving a robot’s environmental understanding, decision-making, and autonomous execution of complex tasks.

I have developed and deployed algorithms for humanoid robots with different body scales and mechanical architectures, including the Unitree G1, Booster Robotics PM01, Westlake Robot O1, and Kepler K2, as well as a confidential Russian humanoid platform. In addition, I have implemented motion-planning and perception strategies for indoor and outdoor locomotion over uneven terrain on the Unitree G1 and quadruped robot platforms.

My previous work focused primarily on humanoid motion-control algorithms and full-body motion tracking. Using a motion-capture suit or a VR headset, I developed teleoperation systems that enable humanoid robots to track human motion in real time with low latency and high fidelity, forming an embodied-avatar system (project description). Currently, I am particularly interested in integrating vision-language models (VLMs) with low-level control, enabling embodied-intelligence systems to perform complex tasks with greater autonomy and efficiency.

Whole-body motion control for humanoid robots
Motion planning and SLAM for indoor and outdoor environments
World-model-based robotic perception and control (WAM)
Integration of vision-language models with low-level control

Publications

A Review of Point Cloud 3D Object Detection Methods based on Deep Learning

Wang, X., Lin, J., Yang, L., Wang, S. (2024). A Review of Point Cloud 3D Object Detection Methods Based on Deep Learning. In: Zhang, M., Xu, B., Hu, F., Lin, J., Song, X., Lu, Z. (eds) Computer Applications. CCF NCCA 2023. Communications in Computer and Information Science, vol 1959. Springer, Singapore. https://doi.org/10.1007/978-981-99-8764-1_3 (2023)

[Paper] [Slides]

A Study on Semantic Segmentation for Small Objects in High-resolution Aerial Images based on Mask R-CNN and HRNet

Sicong Wang, Xijun Zhang, Yujie Yin, Xiyuan Wang, and Longrui Yang. "A Study on Semantic Segmentation for Small Objects in High-resolution Aerial Images based on Mask R-CNN and HRNet." 2024 16th International Conference on Machine Learning and Computing, October 2023. Accepted. (2023)

Review of TSP Solution Methods

Longrui Yang¹, Xiyuan Wang¹, Jie Lin (Corresponding Author), Zhaoqi He, and Sicong Wang. "Review of TSP Solution Methods." 18th International Conference on Bio-inspired Computing: Theories and Applications (BIC-TA 2023), October 2023. Accepted. (2023)

Awards

Rising Star Award, RoboMaster 2023: Infantry Confrontation
2nd Prize, The 6th China University Intelligent Robot Creative Competition
Presided over the 2023 National College Student Innovation Plan and Scientific Research Fund
1st Prize, The 19th National Smart Car Competition for College Students: Outdoor Car Challenge Group (ROS)
1st Prize, The 18th National Smart Car Competition for College Students: Unicycle Group (ROS)
2021 School First Class Scholarship and Outstanding Students