Bearing only tracking using square root sigma point Kalman filter

The robustness and the r.m.s error settling performance of scaled square root SPKF and the same for EKF in a bearings-only-tracking are compared. While the error settling performances are comparable, the SPKF is found to be more robust, as quantified by the frequency of track loss cases in Monte Carlo simulation. A new quantitative test for track loss is introduced. The feasibility tuning the SPKF filter by utilising a design freedom is demonstrated. It is shown that the use of square root SPKF does not introduce substantial computational burden, making the SPKF a strong candidate for on board implementation.