Locating and identifying components in a robot's workspace using a hybrid computer architecture

Abstract

This paper describes a system that locates and identifies components in an automated manufacturing process. The system uses a network of processors (an array of transputers) to construct and hold the workspace model, and to extract the feature measurements used to facilitate component identification. A MLP artificial neural network is then used to identify the components using the feature measurements obtained from the model. In an earlier version of this system goodness-of-fit was used to classify components, however, that method has drawbacks that neural networks overcome. The original design of the system was modular enabling a straightforward substitution of the component classification methods.