CoFreeVLA: Collision-Free Dual-Arm Manipulation via Vision-Language-Action Model and Risk Estimation

TL;DR

CoFreeVLA enhances dual-arm robotic manipulation by integrating a self-collision risk estimator with vision-language models, significantly reducing collisions and increasing task success rates in complex bimanual tasks.

Contribution

It introduces a novel risk estimation module that predicts self-collision likelihood and guides safe manipulation, improving safety and performance over existing methods.

Findings

Reduces self-collisions in dual-arm tasks

Improves success rates on bimanual manipulation

Effective risk-guided adjustments enhance safety

Abstract

Vision Language Action (VLA) models enable instruction following manipulation, yet dualarm deployment remains unsafe due to under modeled selfcollisions between arms and grasped objects. We introduce CoFreeVLA, which augments an endtoend VLA with a short horizon selfcollision risk estimator that predicts collision likelihood from proprioception, visual embeddings, and planned actions. The estimator gates risky commands, recovers to safe states via risk-guided adjustments, and shapes policy refinement for safer rollouts. It is pre-trained with model-based collision labels and posttrained on real robot rollouts for calibration. On five bimanual tasks with the PiPER robot arm, CoFreeVLA reduces selfcollisions and improves success rates versus RDT and APEX.