Intelligent image analysis for error detection and correction in automated laboratory robot systems

Abstract

Error detection and correction are essential steps in developing robust automated laboratory systems involving robots. Single-point sensors can be used to detect errors when the anticipated number is small. But the wide range of errors that can occur in an automated laboratory system makes this an impossible or impractical approach. Computer vision and image analysis techniques can be used to significantly broaden the range of dynamically identifiable error conditions. Object recognition and neural networks can be used to provide further characteristics of an identified error. By combining these techniques, it is often possible to extract sufficient information to direct a laboratory robot to clear or avoid an identified error, allowing the automated laboratory task to continue without human intervention. This has the potential to dramatically improve the overall reliability of an automated laboratory robot system. In this article, we report on the application of computer vision and neural networks to the detection of errors that can occur in a robot system designed to automate the loading and unloading of a laboratory centrifuge. © 1998 John Wiley & Sons, Inc. Lab Robotics and Automation 10: 273–282, 1998