Get Started
Research questionHow can dual-arm vision-language-action policies avoid self-collisions with grasped objects?Coordinated arms and the objects they hold can enter one another’s workspace during continuous control. Instruction-following policies may not adequately model these interactions, creating safety failures during manipulation.
Alignment & Safety
Multimodal Models
Robotics
Latest papersRecent research connected to this question, newest first.CoFreeVLA: Short-Horizon Collision-Free Dual-Arm Manipulation via Vision-Language-Action Model and Risk EstimationThe source studies dual-arm VLA policies using proprioceptive states, visual embeddings, and candidate action sequences for short-horizon collision-risk estimation and closed-loop control. Evidence comes from five bimanual tasks, six VLA backbones, and 30 trials per variant; it reports changes in collision and task-success rates but does not establish performance beyond these settings.research paper · Sep 4, 2026
Related questions
How can vision-language-action policies follow execution details beyond a robot task’s goal?How can robotic manipulation policies combine vision, language, and touch for reliable contact-rich tasks under occlusion?How can vision-language-action policies act reliably when irrelevant sensors are corrupted or only one informative sensor remains?How can vision-language-action systems reliably execute long-horizon manipulation while tracking state and conditional dependencies?
Home
Topics
Search
Library