Trade Offs between Measurement and Track Fusion at the System Level

A hotly debated topic that has been around for decades-probably as long as the Kalman Filter-is what advantage is there to fusing measurements as opposed to fusing tracks at the system level. Many systems use the measurements and have been designed to do so. On the other hand, some systems only utilize tracks from sensors and have had excellent success in this manner. One line of thinking is that it depends on the target that is being tracked. If the target is expected to maneuver in a significant manner then it may be better to use measurements in order to reduce overall lag in the system. Conversely if the target is not expected to maneuver significantly then tracks from each sensor may be a better alternative to allow for bias estimation and removal. The motivation for this paper is not to provide the definitive answer to this topic, but rather the purpose of this paper is to explore and document many of benefits of each approach. It will begin with a basic three-sensor scenario where both measurement and track fusion are performed with the same update rates.