Robot Learning

01

A research-oriented project connecting simulation, learning-based control, and real robot deployment thinking.

Project Detail

Reinforcement Learning Based Locomotion Control for an Omnidirectional Wheeled-Bipedal Robot

2026.01 - PresentProf. Hua Chen, ZJUI

Keywords

Reinforcement LearningIsaacLabMuJoCoPPOSim-to-Real

Overview

This project focuses on learning-based locomotion control for an omnidirectional wheeled-bipedal robot operating on complex terrain. The work connects simulator setup, policy training, controller understanding, and deployment validation.

My Contributions

  • Studied the code structure, training workflow, and control framework of reinforcement learning locomotion projects.
  • Configured simulation environments with IsaacLab and MuJoCo Playground.
  • Reproduced PPO-based locomotion policies and organized the training pipeline.
  • Completed a workflow covering simulation training, sim-to-sim validation, and sim-to-real deployment preparation.
  • Analyzed robot models, observation spaces, reward design, and deployment results to understand terrain adaptation mechanisms.

Outcomes

  • Built a clearer understanding of learning-based robot motion control pipelines.
  • Connected reinforcement learning policy training with practical deployment considerations.
  • Prepared this project as the main research highlight for future graduate research applications.

Project Signal

A research-oriented project connecting simulation, learning-based control, and real robot deployment thinking.