Force delta vla improves robot manipulation by separating actions and corrections
ForceDelta-VLA: Distilling Force-Conditioned ActionCorrections for Contact-Rich Manipulation
Robotics
Summary
Robot arms can be tricky when they have to touch or move objects without causing problems. The authors found a way to teach robots to first plan a main action and then make quick adjustments based on force feedback. This method helps robots respond better to changes when handling things, making them more successful and gentler in contact tasks. Their approach works across different types of tasks and robots, improving success rates and reducing the force used.
What this means in practice
- •For robotics engineers: Improve industrial robots’ performance in tasks requiring delicate contact by enabling real-time, force-based action corrections.
- •For automation system designers: Design control systems that separate planned motions from force-sensitive adjustments for safer and more reliable robot operation.
Authors
Ju Dong, Yu Fu, Jian Chen, Yimeng Liu, Haocheng Zhao, Lei Zhang, Kaixin Bai, Liding Zhang, Diwen Zheng, Alois Christian Knoll, Angela P. Schoellig, Jianwei Zhang
Abstract
Force-aware Vision-Language-Action (VLA) policies improve contact-rich manipulation, but typically combine task-level motion and contact-dependent adjustment in a single action prediction. Demonstrations provide no explicit labels for decomposing that prediction into a reusable reference action and a correction. We present ForceDelta-VLA, a correction-distillation framework that constructs an explicit force-correction target using paired predictions from a frozen teacher's force-conditioned and learned force-agnostic modes. A separate delay-correction target accounts for reference-action mismatch and the change in reference state. Training uses asynchronous schedule replay with the cached task context available during execution. The resulting lightweight policy adjusts the reference actions using recent force history and robot state, responding to contact changes between reference-action updates without regenerating complete action chunks. Across nine single-arm and bimanual contact-rich tasks, ForceDelta-VLA achieves an 82.2% mean success rate, compared with 54.4% for the original ForceVLA baseline. Direct execution of our Stage-1 Temporal Teacher achieves 70.6%. Relative to ForceVLA, the complete system reduces mean peak contact force over successful trials by approximately 26% on both platforms.