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

2026-01-29

AI Summary

This work addresses the safety risks in dual-arm manipulation under vision-language instructions, where unmodeled self-collisions can lead to collisions between robot arms or with objects. The authors propose a risk-aware, end-to-end Vision-Language-Action (VLA) framework that integrates, for the first time, a short-term self-collision risk estimator with a VLA model. By fusing proprioceptive and visual embeddings, the system dynamically predicts high-risk commands and halts their execution in real time. Additionally, it incorporates a risk-guided state recovery and policy optimization mechanism. The method is pretrained in simulation using collision labels and fine-tuned on a real PiPER dual-arm robot. Experiments across five dual-arm tasks demonstrate a significant reduction in self-collision rates and higher task success compared to existing approaches such as RDT and APEX.

Technology Category

• Computer Vision: Language and Vision
• Natural Language Processing: Safety and Robustness
• Intelligent Robots: Manipulation

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

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.

Problem

• dual-arm manipulation
• self-collision
• Vision-Language-Action model
• collision avoidance
• bimanual tasks

Innovation

• Vision-Language-Action Model
• Self-collision Avoidance
• Risk Estimation
• Dual-arm Manipulation
• Policy Refinement