Multisensor Estimation Fusion for Wireless Networks with Mixed Data Delays

Abstract

Motivated by the extensive application of sensor networks in the multisensor target tracking systems, the problem of data fusion with mixed time delays which includes short timedelay and long time-delay is considered in this paper. In order to overcome several primary problems occurred in the existing fusion methods based on the "Out-Of-Sequence" Measurements (OOSMs), such as high investment for tracking systems, huge computational complexity and bad real-time performance and so forth, a universal linear predict estimate weighted fusion algorithm is proposed. Different from the OOSMs methods in which the transmitted data from local sensors to the fusion center are measurements, the proposed fusion algorithm sends the local estimates to the fusion center for fusion and is called "Out-Of-Sequence" Estimates (OOSEs). The time calibration for fusion is finished by predict operation in the fusion center. The proposed OOSEs method can avoid many problems existed in the OOSMs method, but its fusion accuracy is lower than that of the OOSMs method after a sort. This is because that the OOSMs is a smoothing estimate and the OOSEs is a predict estimate. The superiorities of the proposed method are illustrated by performance analysis.