Key points are not available for this paper at this time.
Multimodal soft grippers can adapt grasping strategies to diverse environments, yet integrating sensors under large, coupled deformations remains challenging. Inspired by the mechanosensation of sea anemones and their ability to swallow, this article presents a soft gripper that integrates eight toroidal optical waveguides to realize three modes—contacting, expansion, and swallowing—with continuous proprioceptive feedback. Deformation-induced optical attenuation, processed by machine-learning pipelines, enables perception of object shape, hardness, and surface texture. Experiments show a 0.04 N detection limit, a 0.006 N resolution, a 55 ms response time, and sensitivity >1.4 dB/N, with machine-learning classification achieving >89% accuracy. Mode-specific experiments demonstrate sensing across the entire soft gripper with integrated optical waveguides. The outer surface localizes contact after inflation, the inner surface provides circumferential contact sensing of irregular objects during swallowing; and the pedal interface at the base distinguishes surface hardness and texture, achieving perception on the outer, inner, and bottom interfaces. We also demonstrate multi-object swallowing that grasps and counts 1–4 transparent bottles in real-time and a breakfast task that switches grasping modes to grasp a bowl and cup, swallowing fragile items without damage. These results show that our design enables mode-switching interactive perception and expands opportunities for soft robotics in fragile product handling and laboratory automation.
Xiang et al. (Fri,) studied this question.