Track-to-track fusion is an important part in multisensor fusion.In practice, the process noise is not negligible due to targetmanoeuvring and sensors typically communicate infrequently to savecommunication bandwidth. In such situations, the measurements from twosensors are not conditionally independent due to the common processnoise from the underlying system. This dependence phenomena was alsoobserved by Bar-Shalom (1981) where a formula was derived to compute thecross-covariance of two track estimates obtained by different sensors.Based on the results, a fusion formula was subsequently derived(Bar-Shalom et al. (1986)) to combine the local estimates which tookinto account the dependency between the two estimates. This paper pointsout the implicit approximation made by Bar-Shalom et al. and shows thatthe result turns out to be optimal only in the maximum likelihood sense.A performance evaluation technique is then proposed to study theperformance of various track-to-track fusion techniques. The resultsprovide performance bounds of different techniques under variousoperating conditions which can be used in designing a fusionsystem
展开▼