Event

FORCA: Force-Aware Coding Agent for Contact-Rich Manipulation

1 reports · 1 sources

Xiao Tan, Yixuan Wang, Zhexi Wang, and Wenzhen Yuan propose FORCA, a force-aware coding agent that takes visual observations and wrist force/torque measurements as input, writes Python programs for contact detection and force-responsive motion, and, between execution intervals, combines force records with visual results to diagnose failures and revise motion and force-feedback control logic. The authors state that FORCA requires no demonstrations, task-specific training, or simulation. In five real-robot tasks, they demonstrated its effectiveness on three types of contact-rich manipulation: improving manipulation precision, adapting to changes in object or surface properties, and regulating contact under surface changes and object motion. Compared with a vision-only version of the same agent, the authors report that FORCA improves grasp precision and task success rates and reduces contact loss during wiping and peeling.