COMMUNITY · /community/development
Software for humanoids: SDKs, ROS 2, simulation, control, robot learning, sim-to-real and deployment.
Which simulator do you use for humanoids, for what (RL training, controller testing, sim-to-sim validation, synthetic data), and what made you choose it? Share the friction too: installation, URDF/MJCF/USD import, GPU…
#gazebo#isaac-sim#mujocoWBH Editorial ·
A policy that walks in simulation and stumbles on hardware is a rite of passage. What was the cause in your case (actuator model, latency, observation noise, domain randomisation, control frequency, sim-to-sim checks…
Unitree G1#locomotion#reinforcement-learning#sim-to-realWBH Editorial ·
If you train or deploy on the G1, which variant do you use, and what had to change between the 23-DoF and 29-DoF configurations (action space, observations, URDF/MJCF, deployment)? Unitree's unitree_ros repository now…
Unitree G1#reinforcement-learning#whole-body-controlWBH Editorial ·
Once walking works, what next: getting up, carrying, whole-body manipulation, motion imitation? Share the approach (reward design, motion retargeting, imitation, teacher-student) and the open-source code you built on.…
#imitation-learning#reinforcement-learning#whole-body-controlWBH Editorial ·
A classic question with a moving answer. How would you explain the role of ZMP-based control today, where model-based balance (capture point, MPC, whole-body control) is still preferred, and where learned policies have…
#balance#locomotion#whole-body-controlWBH Editorial ·
How do you collect demonstrations for your robot: VR headset, exoskeleton, leader-follower arms, motion capture? What did setup and calibration involve, and how many episodes did a usable policy take? On…
Vega#imitation-learning#teleoperationWBH Editorial ·