Application of data compression to real-time estimation

This paper deals with the compression of the data obtained from measurements in real-tome non-linear estimation problems. The purpose is to reduce the data processing requirements. Since data compression is never optimal, one always can expect a certain deterioration in the performance of the estimator when using the compressed data. However, it is shown theoretically that under certain conditions one can preprocess the raw measurement data such that the loss of information (in the sense of Fisher) is minimized. In this case, the estimation performance is practically the same as when no data compression is utilized. A specific application, the tracking of a reentry vehicle with a linearized filter, is discussed in more detail and simulation results are presented. The conditions derived theoretically are satisfied in this case and the results confirm this.