Fast retargeting method speeds humanoid robot motion learning
Fast Motion Retargeting Pipeline for Scalable Humanoid Robot Learning
Robotics
Summary
Capturing human movements is helpful for teaching robots how to move, but copying those motions directly onto different robots is tricky and usually slow. The authors offer a faster method called FRDV that quickly adapts human motions to various humanoid robots while checking the moves against the robots' physical limits to avoid impossible actions like sinking into the ground or floating feet. They show this method works better and faster than previous approaches and even tested it on a real small humanoid robot.
What this means in practice
- •For humanoid robot developers: Generate physically feasible humanoid robot motions from human demonstrations rapidly and without manual tuning.
- •For robotics reinforcement learning teams: Train motion tracking policies efficiently using scalable retargeted human motion data for improved robot movement fidelity.
Authors
Daniel Marew, Sangjun Park, Nisal Perera, Donghyun Kim
Abstract
While large-scale human motion capture datasets provide a rich source of demonstrations for humanoid robot learning, retargeting the diverse motions to different humanoid robots remains a nontrivial challenge. Existing kinodynamic trajectory optimization pipelines are computationally expensive and often require manual adjustments per motion. We present FRDV (Fast Retargeting with Dynamics Validation), a scalable framework consisting of rapid constrained kinematic retargeting and an additional dynamics validation pipeline. Its kinematic stage integrates a custom Sequential Quadratic Programming solver, Real-Time Iteration, and parallel chunk-based processing, achieving up to 9,156 frames per second, approximately up to 136 times faster than GMR and ProtoMotions, while robot-specific constraints are incorporated into the optimization to reduce common retargeting artifacts, such as ground penetration and self-collision. For motions requiring additional verification, a contact implicit iLQR controller is utilized to test the retargeted reference under full rigid body dynamics. This dynamics-based validation identifies dynamically infeasible motions and refines contact-related artifacts such as foot floating and foot skating. We evaluate retargeting quality by training motion tracking reinforcement learning policies using several state-of-the-art pipelines and comparing their tracking fidelity and success rates. Furthermore, sim-to-real transfer is demonstrated on a Unitree G1 humanoid. Videos and code are available at: https://daroslab.github.io/frdv/.