Tracking and data fusion

This paper is concerned with the principles of data fusion for two or more Kalman filters tracking the same target. Each filter receives a sequence of measurements from its own sensor and the measurement errors are independent for different filters. However, the estimators of the target position which they produce are not independent of one-another because they involve the same process noise in the mathematical model assumed for the motion of the target. Estimators can be combined to improve precision by using conditional expectations of the target state. given the appropriate information. This data fusion requires a substantial extension of the standard theory for a single filter.