Stable Grasping Strategy for Robots on Flexible Mesh via Integration of Haptic Sensing and VLM
The maintenance of deep-sea net cages, specifically net cleaning and chemical coating application, is currently hindered by intensive labor requirements and the inherent risks of high-altitude maritime operations. Consequently, developing automated climbing robots offers significant advantages in both safety and operational efficiency. However, unlike rigid surfaces, the flexible netting of sea cages is characterized by easy deformation, discontinuity, and an unstructured nature. During climbing, robots frequently encounter unstable grasping or unexpected interference between the gripper and the mesh, which serves as a primary bottleneck for engineering applications. To address these challenges, this paper develops a novel robotic gripper structure specifically designed for climbing flexible netting and proposes a gripper state monitoring system based on multi-modal information fusion. Experimental results conducted in both PyBullet physics simulations and on a physical flexible netting testbed demonstrate that this multi-modal fusion system effectively compensates for the inability of single-force sensors to perceive spatial geometric relationships. This integration significantly enhances grasping stability and operational safety on unstructured, flexible net surfaces.