Manoeuvre detection and track correction by input estimation

The paper gives a simplified solution to the input estimation method of tracking a manoeuvring target. Manoeuvre detection is decoupled from input estimation and does not require estimates of the manoeuvre inputs. Instead, the detection variable is a normalised sum of the differences between the measurements and the Kalman filter projections of positions. For the same probability of false alarm, the resultant detector has a higher probability of detecting than the innovations-based detector. The input estimator uses the same position differences to give a generalised least squares estimation of the manoeuvre inputs. No matrix inversion is needed and many calculations can be done offline. Thus, this input estimation detection scheme is applicable to cases where several trackers are run simultaneously and where real-time constraint is a problem. The theoretical development is verified by simulation results, which also contain some tracking examples of typical target manoeuvres.