Get Started
Home
Topics
Search
Library
Research questionHow can robotic manipulation policies combine vision, language, and touch for reliable contact-rich tasks under occlusion?Vision-language-action policies can lose task-relevant information when objects are occluded or interactions depend on precise physical contact, making fine manipulation difficult.
AI
Evaluation & Benchmarks
Machine Learning
Multimodal Models
Research Paper
Robotics
Technology
Latest papersRecent research connected to this question, newest first.TacVLA: Contact-Aware Tactile Fusion for Robust Vision-Language-Action ManipulationThe source studies a transformer-based vision-language-action policy augmented with tactile sensing for robotic manipulation. It reports constraint-locked disassembly, in-box picking, visual-occlusion robustness, and recovery under human disturbance; the abstract does not specify the tactile hardware or other implementation details.research paper · Sep 6, 2026
Related questions
How can vision-language-action policies act reliably when irrelevant sensors are corrupted or only one informative sensor remains?How can dual-arm vision-language-action policies avoid self-collisions with grasped objects?How can pretrained vision-language-action models reliably perform contact-rich manipulation when goals, scenes, and contacts change?How can vision-language-action policies follow execution details beyond a robot task’s goal?