Soft Robotic Gripper with Integrated Perception and Autonomous Classification

Chu, Ashley

Etd

Soft Robotic Gripper with Integrated Perception and Autonomous Classification

Public

Robotic grippers are arguably the most crucial components of robotic manipulation tasks. While rigid robots can easily hurt humans or delicate objects when interacting with them, soft robotic grippers are able to circumvent this issue with their inherent safety. However, most existing soft grippers suffer from grasping challenges that arise from limited mechanical capabilities, and lack of integrated perception and intelligence abilities. The goal of this work is toward intelligent soft robots. The main focus of this work is to introduce a novel soft robotic gripper. To address perception and classification problems, we presented a fully electric, sensorized soft adaptive gripper with three-dimensional self-conforming fingers. We implement a Support Vector Machine classifier to recognize various fruit items based on the embedded sensory measurements from a single grasp without relying on external vision systems. The 3-D compliance of a finger enables mechanical intelligence in terms of adaptability to varying object sizes and shapes. The proposed integrated system demonstrates the ability to classify different fruit items of similar color and similar size, which would be challenging to detect using vision systems. We also demonstrate the ability to determine the freshness of the same fruit items as they age. The integrated system aims to enable perception, action, and decision-making to fulfill autonomous grasping and classification on the fly to enable real-world applications in grasping, sorting, and packaging complex objects.

Creator