Event
FAWAM: Force-Aware World Action Model Improves Success Rate in Contact Tasks
1 reports · 1 sources
Haotian He and other authors propose the force-aware world action model FAWAM, which introduces force information at three levels: perception, prediction, and closed-loop execution: first, historical six-axis force/torque signals are encoded to modulate action generation; then, future actions and end-effector force wrenches are jointly predicted to explicitly model contact evolution; and the predicted force wrench trajectory is used as a reference during execution, with real-time force feedback used to correct actions online. The authors claim that in real-world experiments on multiple contact-rich tasks, FAWAM's average success rate is 36.25% higher than pure vision baselines and 21.25% higher than existing force-aware baselines.
Report timeline
- FAWAM: Force-Aware World Action Model for Closed-Loop Contact-Rich Manipulation
The authors propose FAWAM, a force-aware world action model that incorporates force information at three levels: perception, prediction, and closed-loop execution. It first encodes historical 6-axis force/torque signals to modulate action generation, then jointly predicts future actions and end-effector wrenches to explicitly model contact evolution, and uses the predicted wrench trajectory as an execution-time reference while correcting actions online with real-time force feedback. In real-world experiments across multiple contact-rich tasks, FAWAM achieved an average success rate 36.25% higher than vision-only baselines and 21.25% higher than existing force-aware baselines.
arXiv cs.RO ·