Integration of Multi-Camera Vision System for Automatic Robotic Assembly

Abstract

An automated assembly cell does not come with the luxury of intelligence. A non-adaptive robot manipulator must have more accurate parts and assembly fixtures to a greater accuracy than their assembly tolerances. Accurate parts presentation and the fixturing of assemblies can be a complex and costly portion of an automated cell. Machine vision can be integrated with the assembly process to effectively and efficiently guide the assembly process. This paper presents a framework for an automatic assembly system that consists of an industrial robot and a three-camera vision system, where the tasks involve object detection, pose estimation during pick and place operations, and assembly verification. Given a pocket calculator as an application, the proposed framework can successfully perform autonomous assembly of a pocket calculator. The experiment results verified the efficacy of the proposed method.