Autonomous agents in prediction of element concentration: Real-time fuzzy classification paradigms

We describe a novel approach for real-time fuzzy classification of spectral data using autonomous agents. The immediate goal of this approach is to provide an interactive (real-time) image classification middleware as a feedback backbone for a control or monitoring system. The domain of this experiment is a controlled combustion chamber, with a monitoring system that checks the concentration of specified elements in the effuse. The proposed middleware acts as an intelligent peer to provide 'timely' and accurate feedback to the control system on real-time classification of the spectral data images. We discuss the results of implementing three classification paradigms for this middleware, and their performances in terms of accuracy and interactivity.