Multi-sensor target
tracking has traditionally been performed using a single processor to monitor
several sensors (centralized fusion), but this method is demanding of both
computational power and communication bandwidth. Distributed sensor fusion
is a method of addressing these limitations. However, the distributed sensor
fusion problem is more complex due to the correlation of separate track
estimates. A method known as measurement reconstruction has recently been
shown to address this problem in a specific architecture. This paper extends
the measurement reconstruction approach to a more generalized architecture
using two new algorithms. Computational and communication requirements
are compared with centralized sensor fusion, and Monte Carlo simulation
studies are used to compare the performance of these algorithms.