Harvard Unveils GRASP for Contact-Rich Motion Retargeting

Harvard's computer robotics group released GRASP, which refines human demonstrations in physics simulation to help RL master contact-rich tasks like pen spinning. It combines sampling-based exploration with gradient-guided refinement, recovering gradients by fitting local cost models from past rollouts without requiring differentiable simulators.

2026-11-01 ~ 2026-11-01 · 2 related posts

• Harvard's GRASP refines human demos in physics sim for dexterous hands and Unitree G1 — hankyang94 · 2026-11-01
• Harvard's GRASP recovers gradients from archived rollouts for dexterous retargeting — hankyang94 · 2026-11-01