ReContact: Reconstruct and Retarget Human Contacts from Video for Contact-Rich Humanoid Manipulation
Overview
This paper introduces ReContact, a framework that reconstructs and retargets human-object interactions from video to enable contact-rich manipulation for humanoid robots. By defining an object-centric contact plan that specifies phases, pairs, anchors, directions, and modes, the method stabilizes pose reconstruction under occlusion and ensures forces align with planned contacts. In simulation, ReContact achieves 83.5% success in thigh-supported regrasp tasks compared to 0.16% for baselines, and improves tube-transfer success from 20.4% to 97.8% using a single policy trained on 247 references. On a real Unitree G1, the system executes seven contact-rich manipulation and two loco-manipulation tasks.
Generated from attributed reports · Updated 1 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Environment: simulation
“econstruction through occlusion. A residual policy over a frozen whole-body tracker is rewarded for making contact at planned anchors with forces aligned to the specified directions. In simulation, on a thigh-supported regrasp task, ReContact achieves the planned two-palm bottom hold in 83.5 of episodes over three training seeds, versus 0.16% for the best adapted baseline. One policy trained on 247 references re-solv”
Exact source · revision 1Source owner not reported
Reported quantity · number of tasks: 7 other · Basis not reported · Differing source assertions
“ed for new placements raises in-workspace tube-transfer success from 20.4% (source reference only) to 97.8%. On a real Unitree G1, policies execute seven contact-rich manipulation and two loco-manipulation tasks.”
Exact source · revision 1Source owner not reported
Reported quantity · number of tasks: 2 other · Basis not reported · Differing source assertions
“ed for new placements raises in-workspace tube-transfer success from 20.4% (source reference only) to 97.8%. On a real Unitree G1, policies execute seven contact-rich manipulation and two loco-manipulation tasks.”
Exact source · revision 1Source owner not reported
Reported quantity · number of reference videos: 247 other · Basis not reported
“econstruction through occlusion. A residual policy over a frozen whole-body tracker is rewarded for making contact at planned anchors with forces aligned to the specified directions. In simulation, on a thigh-supported regrasp task, ReContact achieves the planned two-palm bottom hold in 83.5 of episodes over three training seeds, versus 0.16% for the best adapted baseline. One policy trained on 247 references re-solv”
Exact source · revision 1Source owner not reported
Reported quantity · number of training seeds: 3 other · Basis not reported
“econstruction through occlusion. A residual policy over a frozen whole-body tracker is rewarded for making contact at planned anchors with forces aligned to the specified directions. In simulation, on a thigh-supported regrasp task, ReContact achieves the planned two-palm bottom hold in 83.5 of episodes over three training seeds, versus 0.16% for the best adapted baseline. One policy trained on 247 references re-solv”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalReContact: Reconstruct and Retarget Human Contacts from Video for Contact-Rich Humanoid Manipulation
This paper introduces ReContact, a framework that reconstructs and retargets human-object interactions from video to enable contact-rich manipulation for humanoid robots. By defining an object-centric contact plan that specifies phases, pairs, anchors, directions, and modes, the method stabilizes pose reconstruction under occlusion and ensures forces align with planned contacts. In simulation, ReContact achieves 83.5% success in thigh-supported regrasp tasks compared to 0.16% for baselines, and improves tube-transfer success from 20.4% to 97.8% using a single policy trained on 247 references. On a real Unitree G1, the system executes seven contact-rich manipulation and two loco-manipulation tasks.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).