Flexible cables, ubiquitous in both domestic and industrial settings, exemplify semi-deformable linear objects (SDLOs) characterized by a rigid connector section and an elongated, flexible cord segment. This study aims to introduce a vision-based method for grasping SDLOs in cluttered environments.
The proposed method for grasping SDLOs in cluttered environments focuses on the rigid connector component for stability. It begins by assessing SDLO occlusion to determine grasping order based on spatial relationships among cables, minimizing entanglement risks. The key innovation is a one-stage grasp detector that treats grasp detection as keypoints detection in image space. Each grasp candidate is identified as a pair of keypoints, allowing effective detection of grasping poses for thin, elongated objects with minimal texture. This approach enhances the ability to handle SDLOs in complex environments.
The efficacy of the method was evaluated through experiments conducted on a real robotic experimental platform. Results indicate that the proposed approach is viable for sequentially extracting cables from cluttered scenes.
The proposed method incorporates several key innovations to address the challenges associated with grasping SDLOs in cluttered environments. Enhanced stability is achieved by targeting the rigid component of the SDLO, typically the connector, while an anti-entanglement strategy is implemented by assessing the occlusion of SDLOs and determining the grasping order based on spatial relationships among cables. The approach introduces an innovative one-stage grasp detector that conceptualizes grasp detection as keypoints detection in image space. This efficient identification method represents each grasp candidate as a pair of keypoints, enabling effective detection of grasping poses for thin, elongated objects with minimal texture.
