A method for characterizing track goodness when design mismatch occurs

While the Kalman filter has been theoretically shown to be optimal as a tracking filter in sensor applications, the performance of the filter can be sub-optimal when the filter parameters are poorly selected or when the filter is used to track a class of target other than that for which the filter is optimized. Constant-gain, steady-state versions of the Kalman filter have properties that can be easily characterized analytically and then optimized for a selected design criteria. In this paper, we use the mean squared error in velocity as the design criterion to motivate the concept of track goodness and develop two indicators of track quality. We then use one of these indicators to predict the performance of a Kalman filter and an alpha-beta filter on an accelerating target for a range of designed acceleration parameters. Finally, we compare the predicted performance of both filters against the actual performance for design parameters at above and below the actual target acceleration.