Event
CLARO: VLM agent achieves open-vocabulary grasping in cluttered scenes
1 reports · 1 sources
Sungsu Kim and other authors submitted a preprint proposing CLARO (CLutter-Aware Robot Orchestrator), a VLM agent that orchestrates skills such as open-vocabulary object detection, segmentation, 6-DoF grasping, and non-prehensile pushing to locate and extract a language-specified target in cluttered environments. It records action history to support long-horizon execution and outputs grasp poses or 3D waypoints, and can transfer to different robotic arms without task-specific training. The authors report that CLARO achieved an average success rate of 89.1% in dynamic simulated environments, and in highly cluttered scenes improved by more than 16% relative to the strongest comparison method; it achieved a success rate of 88.8% across various cluttered scenes with a real robotic arm. The paper has been accepted to WACV 2027.
Report timeline
- CLARO: Open-vocabulary grasping in cluttered scenes with a VLM agent
CLARO is a VLM agent that orchestrates skills such as open-vocabulary object detection, segmentation, 6-DoF grasping, and non-prehensile pushing to locate and retrieve a language-specified target in cluttered environments. It records past actions to support long-horizon execution and outputs grasp poses or 3D waypoints, enabling transfer to different robot arms without task-specific training. In a dynamic simulation environment, CLARO achieved an average success rate of 89.1%, a relative improvement of over 16% on the strongest competing method in highly cluttered scenes; on a real robot arm, its success rate was 88.8%.
arXiv cs.RO ·