Track-before-fuse error bounds for tracking passive targets

In tracking applications, data fusion involves combining data from multiple sensors to produce lower error target state estimates than were possible from using data from individual sensors alone. The main focus of this paper is performance modelling and evaluation for two different fusion paradigms: the track-before-fuse (or track fusion) approach and the fuse-before-track approach. Track-before-fuse is often required in distributed, low-bandwidth environments, while fuse-before-track is an option when centralization is possible. It is anticipated that fuse-before-track will provide superior performance over track-before-fuse because it postpones hard decision making until the most data is received, analogous to track-before-detect. This paper computes approximate error bounds for a track-before-fuse approach and verifies them via simulation. It also shows that a fuse-before-track approach out-performs a standard track-before-fuse method.