Vision-Based Self-Adaptive Gripping in a Trimodal Robotic Sorting End-Effector

Abstract

Recyclable waste management, which includes sorting as a key process, is a crucial component of maintaining a sustainable ecosystem. The use of robots in sorting could significantly facilitate the production of secondary raw materials from waste in the sense of a recycling economy. However, due to the complex and heterogeneous types of the recyclable items, the conventional robotic gripping end-effectors, which typically come with a fixed structure, are unlikely to hold onto the full range of items to enable separation and recycling. To this end, a trimodal adaptive end-effector is proposed that can be integrated with robotic manipulators to improve their gripping versatility. The end-effector can deploy effective modes of gripping to different objects in response to their size and porosity via gripping mechanisms based on Nano Polyurethane (PU) adhesive gels, pumpless vacuum suction, and radially deployable claws. While the end-effector's mechanical design allows the three gripping modes to be deployed independently or in conjunction with one another, this work aims at deploying modes that are effective for gripping onto the recyclable item. In order to decide on the suitable modes of gripping a real-time vision system is designed to measure the size and porosity of the recyclable items and advise on a suitable combination of gripping modes to be deployed. Integrated current sensors provide an indication of successful gripping and releasing of the recyclable items. The results of the experiments confirmed the ability of our vision-based approach in identifying suitable gripping modes in real-time, the deployment of the relevant mechanisms and successful gripping onto a maximum of 84.8% (single-mode), 90.9% (dual-mode) and 96.9% (triple-mode) of a specified set of recyclable items.