IMM vs Kalman Filter: When We Should Use the Last

In this paper a comparison of accuracy between the Kalman Filter Estimator and the Interacting Multiple Model (IMM) Estimator was performed. The comparison was carried out following a study which compared the two estimators’ performance in tracking single targets, in order to quantify the condition under which an IMM estimator is more desirable than the Kalman Filter. The writers of said paper defined a “Maneuvering Index” to be used to quantify the maneuverability of the targets. Their results show that the IMM achieves significant improvement over the Kalman Filter. In this paper their experiments were reconstructed and repeated, and the results differ from theirs as they were found to achieve less improvement than their study had stated. This improvement was gained by the smaller state estimate errors provided from the IMM in non-maneuvering segments of the simulations. As such the proposed “0.5 rule”, which states that an IMM is preferable to the Kalman when a target’s index is 0.5 or above, should be revised and updated.