Performance of state space multi-sensor track fusion with model mismatch
M.L. Krieg, Douglas Andrew Gray · 2002
Successful tracking requires assumptions to model the target's behaviour and sensor characteristics. The sensitivity of multisensor tracking filters to errors in these models is of particular interest to the designer. We consider the specific case of a dual sensor Kalman tracker and compare its theoretical sensitivity to process and measurement model mismatch with that of a single sensor Kalman tracker, confirming our results with simulations. We introduce the notion of a fusion gain as a comparative measure of tracking performance, and the concept of a mismatch factor as a comparative measure of sensitivity, between the dual and single sensor trackers.