When sensor data from multiple platforms is collected and fused in centralized manner, sensor measurements can arrive out-of-sequence at the central processor due to varying pre-processing times at the platforms and uncertain data transmission delays in communication networks. As a result, classical filters for orderly measurements, such as Kalman filter, cannot be used directly. There are three classes of OOSM (out-of-sequence measurements), respectively called single-lag OOSM, multiple-lag OOSM, and mixed-lag OOSM. For the first two classes of OOSM, C.A. Stelios et al (1988), R.D. Hilton et al., (1993), Y. Bar-Shalom (2002), and M. Mallick et al. (2001) have studied some optimal or suboptimal filters. This paper develops a recursive filter algorithm called UOOSMF (unified out-of-sequence measurements filter), which can sequentially process the OOSM in order of the time of arrival and is suboptimal in the linear MMSE sense under one acceptable approximation. The filter is accommodated with all the above three classes of OOSM. With the same measurements, the estimation accuracy of the UOOSMF is almost the same as standard Kalman filter for orderly measurements, and its computations and memory requirements are low. Numerical example proves the above conclusions.
[1]
J. A. Roecker,et al.
Comparison of two-sensor tracking methods based on state vector fusion and measurement fusion
,
1988
.
[2]
Yaakov Bar-Shalom.
Update with out-of-sequence measurements in tracking: exact solution
,
2002
.
[3]
Stelios C. A. Thomopoulos,et al.
Distributed filtering with random sampling and delay
,
1988,
Proceedings of the 27th IEEE Conference on Decision and Control.
[4]
Stefano Coraluppi,et al.
Comparison of Out-of-sequence Measurement Algorithms in Multi-platform Target Tracking
,
2001
.
[5]
S. Coraluppi,et al.
Advances in asynchronous and decentralized estimation
,
2001,
2001 IEEE Aerospace Conference Proceedings (Cat. No.01TH8542).
[6]
Richard D. Hilton,et al.
Tracking with Time-Delayed Data in Multisensor Systems
,
1993
.