Zecheng Zhu

Hii! I’m an undergraduate student in Electronic and Information Engineering at Imperial College London, supervised by Prfs. Stephen James at Safe Whole-body Intelligent Robotics Lab (SWIRL).

Prior to this, I am fortunate to work with Prfs. Zhongyu Li and Prfs. Bo Tao on Legged Robot and Reinforcement Learning(RL), and intern at Lightwheel AI on Simulation Framework and RL.

Find more in my Curriculum Vitae.

Selected Projects

View Project →

GenlocoV2: Transformer-based Generalized Locomotion Controller

Transformer-based generalized locomotion controller for morphology & topology-agnostic policy learning across diverse robotic embodiments including quadrupeds, bipeds, and humanoids. Implemented multi-robot training pipeline with procedurally generated morphological variations and novel graph-based robot representation architecture.

TD3 Adaptive Navigation Controller

Used the Twin Delayed Deep Deterministic (TD3) algorithm to train an adaptive controller for Jackal robot, achieving major improvements in navigation. Specifically, with the DWA algorithm, the controller increased the success rate from 72% to 89.47% and cut average navigation time from 76.64 seconds to 50.92 seconds across 300 simulation worlds in Gazebo.

Autonomous Campus Tour Guide Robot

Award-winning autonomous robot with vision, voice, and cloud integration for campus tours. Features self-balancing mobility using dual-loop PID/LQR control achieving <15cm positional accuracy, YOLOv8 Nano vision system, cross-platform Flutter app, and RAG campus AI chat assistant.

Languages: Python, C++, C, SQL, SystemVerilog
Skills: Isaac Sim/Gym/Lab, PyTorch, ROS, Linux, Git, Embedded Platforms (ESP/STM)