Abstract

Multidimensional sensors such as spectral sensors have been widely used in monitoring tasks. They provide simultaneously many measurements distributed over the spectra. These measurements are normally highly correlated and the important information lies in a lower dimensional space. Their direct use in control schemes has been limited by the multidimensional nature of the sensor output and the fact that spectral measurements can be affected by many factors. Thus some processing is required to extract the important information out of them. In this work, the use of a spectral sensor to control a mixing process is addressed. In this application, the measurements are described as a linear combination of some spectral basis functions and the dynamics by linear time invariant systems. Two control strategies based on the information provided by a spectral sensor are proposed and analyzed. A simple example illustrates the main characteristics and differences of the proposed approaches.