Multiple sensor tracking architecture comparison

There are several multiple sensor tracking architectures referenced in the literature [1]-[11]. The basic architectures have individual sensors observing the same targets and sending their information about those targets to a fusion center, where a system track, which is a fusion of the individual sensor data, is formed. The fusion center could be a distributed architecture itself, but this article will only examine the roles of the individual sensors, the fusion center (whether at one central location or distributed), and the information passed between them. These architectures range from each sensor tracking the targets and sending a state vector and covariance matrix to the fusion center to each sensor sending all measurements (targets and clutter) to the fusion center for tracking and state estimation. Choosing which architecture to use involves decisions between optimal estimation, practical use of communication bandwidth between the sensors and the fusion center, and computer resources required at either location. A comparative study of architectures in the literature has been limited to varieties of one general architecture (track-to-track fusion) [4]-[11]. This article will examine advantages and disadvantages of each system and show that there is one architecture approach rarely mentioned in the literature that has more advantages than the others for a general multiple sensor tracking system. This article uses the term tracking architectures to describe the different approaches to sensor fusion, which is used elsewhere in the literature [2]. The term tracking configurations has also been used [1]. The use of the term architectures in this article is equivalent to the term configurations found in [1].