Utilizing Hand-Eye Active Visual Servoing for Automated Bolt Removal

Abstract

Due to the decreasing lifespan of products, the accumulation of electronic waste (e-waste) continues to increase in each year. Disassembling products without disposal can significantly contribute to resource utilization and energy conservation. Since most of the disassembling is currently performed manually, recycling rates remain below 50% [2]. Compared to the research and development efforts invested in production automation, those dedicated to disassembly have been far less. The accumulated e-waste consists of various types in small quantities. Therefore, promoting the automation of the disassembling process using positioning fixtures is difficult. This situation emphasizes the need for a universal robotic system capable of disassembling products of diverse sizes and shapes without positioning fixtures. In the disassembling process, our primary focus is on the removal of bolts, which are commonly found in many types of products. For the automatic removal of the bolts, we suggest a robotic system capable of precise positioning of an electric screwdriver using visual servoing. Visual servoing is a well-known technique for positioning with robotic vision. However, the conventional visual servoing method has certain limitations particularly when dealing with objects lacking distinct visual features such as e-waste. To solve this issue, we propose an automated bolt removal system using an active visual servoing(AVS) method. AVS involves projecting optimal patterns onto target objects to enhance image features. The optimal projection pattern has been derived theoretically on the assumption that the intensity of light leaving the projector and light incident on the camera is equal. In this paper, we relax this assumption to handle various types of e-waste. AVS is implemented on hand-eye system. The proposed system robustly achieves the task of bolt removal without positioning fixtures. It effectively handles misalignments of the target object and maintains a positioning accuracy of 0.5mm.