Theoretical performance of a multisensor track-to-track correlation technique

A multisensor track-to-track correlation technique, originally proposed by Singer and Kanyuck (1971) is analysed. It is shown that the probability of correlating two tracks is a function of two dimensionless scalar parameters, T and W. T is a threshold which guarantees a selected successful correlation probability when no biases exist between the two tracks state estimates. W is a normalised parameter which indicates the magnitude of the biases in position and rates which may exist between tracks. Results are presented for the probability of track-to-track correlation for various T and W values. The analysis is extended to quantify the probability of false association when two tracks from sensor B correlate with one track from sensor A and the conflict is resolved incorrectly. The false association probability is shown to be a function of three dimensionless parameters T, W and Δ where T is the same threshold parameter as previously, W is a measure of the biases between the incorrectly paired tracks and Δ is a relative measure of uncertainty between the two track pairings. Numerical results for the false association probability are presented for various T, W and Δ values. The results will be of interest to scientists and engineers involved in multisensor system design