Humanoid Robotics · RoboCup
2022714goo@gmail.com
Hello! I am Ziqian Guo, an undergraduate student in Mechanical Design, Manufacturing and Automation at China Agricultural University.
I work on humanoid robot soccer, with a focus on reinforcement-learning-based fall recovery, motion control, ROS2 strategy integration, and simulation-to-real deployment. My current goals are robust get-up recovery, match strategy, and safe policy deployment on humanoid platforms.
China Agricultural University
B.Eng. in Mechanical Design, Manufacturing and Automation
September 2024 - Present · Beijing, China
CAU RoboCup Humanoid Robot Team
Core Member · Strategy and Locomotion Deployment
October 2025 - Present
Humanoid Fall Recovery RL (loco_recovery)
MJLab-based fall-recovery training for flat and rough terrain. It bundles robot models, loco_recovery task registration, motion data, training/evaluation scripts, and log inspection for humanoid get-up behavior.
Humanoid Motion Retargeting and Data Conversion Toolkit
A humanoid motion pipeline for clipping, previewing, retargeting, and converting CSV, PKL, and BVH motion data. It connects external human motion datasets to robot joint mappings and prepares reference data for recovery and imitation training.
Robot Policy Deployment and Safety FSM Framework
A lightweight finite-state-machine framework for standing, recovery policy execution, and safety protection. It validates joint order, observations, action scale, and control frequency, while adding command filtering and staged checks from passive mode to low-speed hardware tests.