Self-Optimizing Robot Vision for Online Quality Control

Abstract

An online quality control system based on robot vision designed as a quality control agent is presented. It is part of a multi-agent system conceived for control of a production line at factory level, integrating quality control with process control. The implementation of self-optimizing behaviors aimed to improve reliability and confidence level of the diagnosis in a quality control system is also explained. Focus is on integration of advanced state-of-the-art experimental techniques to a real-world industrial problem. These concepts are developed for a robot vision system in eye-in-hand configuration for in-line inspections, which allows the flexibility necessary to implement agent behaviors. The design of the quality control station is presented. The set of self-optimization strategies, which have been implemented, is outlined. They consist in (a) the spatial repositioning of the vision system so as to optimize the line of sight and allow imaging of parts, whichmay randomly be hidden by other components and (b) self-optimization in image acquisition and processing. This increases the level of confidence in the diagnostic output. Tests on the laboratory prototype show the performance improvement of the inspection and allow to discuss the potential of agent technologies for in-line quality control.