TABLE OF CONTENTS The term Kalman filter stirs excitement in the heart of every student of engineering and science. Kalman's epoch-breaking solution [1] to the problems of discrete state estimation has been applied successfully to a wide and diverse area including aerospace and marine navigation, power grid control, plant operation control, biomedical sample cultivation, prediction of demographic distribution, and even crop harvest forecast.
Scores of books and hundreds of research reports have been published, some theoretical, some applied, and others in between. This chapter is written for those students and engineers who plan to broaden their knowledge of Kalman filters in a quick order.
Estimation is a process of extracting information from available measurements, though the measurements are invariably contaminated with errors. The most important information we wish to extract from the measurements may be a prediction, such as the future position of an aircraft or ocean liner or the future power demand and the future plant control required to produce goods with specified tolerance.
The object of this chapter is to provide practical working familiarization in the subject of estimation, prediction, and computation algorithms that extract the best information from contaminated measurements.
An implementation of the Kalman filtrering process on a computer illuminates the importance of the finite word length of computing machines. The finite word length causes a loss of precision through round-off error, destruction of symmetry in a matrix, accumulation of error build-up, and so on. We sometimes encounter a numerically unstable algorithm that leads to an utterly...
