Chapter 14

Performance Measures for the Family of Kalman Filters

In this chapter, we will address several general methods to measure the performance of any of the Kalman filters developed in Chapters 6–12. The performance measures fall into two classes. The first class is applicable when ground truth tracks are not available or where the number of ground truth tracks available for a given scenario are not statistically significant, such as in the analysis of a small number of field tests. For this case, only the filters estimated track covariance matrix can used as an input to generate successive track error ellipses as a performance measure. On the other hand, a larger class of performance measures are available when a simulation is used to generate a completely known target track. This is true because a large number of Monte Carlo observation data sets can be generated that allow for the use of a variety of specific tracker performance measures, such as the root mean squared (RMS) error, the Cramer–Rao lower bound (CRLB), and the percentage of divergent tracks. In this chapter, all of these performance measures will be discussed and specific implementation issues associated with each will be addressed.

Get Bayesian Estimation and Tracking: A Practical Guide now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.