Humanoid Robots Reinforcement Learning Training

Training & Education
1

Hi ROS community,

We’re excited to announce a new hands-on, on-site training focused on reinforcement learning for humanoid robots.

You will work through a full humanoid robotics pipeline, starting from low-level robot access and ending with learned control policies running on physical humanoid robots.

During the training, you will:

  • Use a Unitree G1 EDU development PC and Unitree’s SDK to command, monitor, and control the robot at a low level

  • Build and train reinforcement learning policies for humanoid locomotion and motion control in simulation

  • Deploy learned policies on real humanoid robots

The training combines high-fidelity simulation, real robot execution, and hands-on debugging of simulation-to-real gaps that commonly appear in humanoid systems.

Training Schedule:

  • Day 1. RL for Humanoid Locomotion and Teleoperation
    You will work on:

    • Reinforcement learning based walking control for the Unitree G1 humanoid robot

    • Teleoperation and imitation learning using Isaac Lab Mimic

  • Day 2. RL for Humanoid Whole-Body Control
    You will work on:

    • Full-body control from motion tracking using reinforcement learning

    • Deployment of learned policies using ros2_control

  • Day 3. Vision-Language-Action Models and MuJoCo Lab
    You will work on:

    • Vision-Language-Action models for generalized humanoid skills using Gr00t

    • Training and validation of humanoid behaviors in MuJoCo Lab

Training details

Humanoid Platforms Used in the Training:

  • Unitree G1 Edu

  • PAL Robotics Kangaroo

Questions?
Contact info@theconstruct.ai

humanoid-robot-reinforcement-learning-training
humanoid-robot-reinforcement-learning-training1200×630 124 KB

2 Likes