Research On Passive Localization Algorithms Based On Adaptive Networks

Recently, the research and application of adaptive networks have attracted extensive attention. To accurately estimate unknown interested parameters in adaptive networks, many effective and innovative algorithms have been proposed. As the most important and popular algorithm, diffusion LMS algorithm have been studied and analysed deeply. And many valuable variants are proposed to further improve estimation performance. On the other hand, localization as important and special parameter estimation is always a hot topic in signal processing. Among various localization methods, direct position determination methods become popular due to its high accuracy.Therefore, it is meaningful in both theory and application to study distributed direct localization of an emitter using a wireless sensor network. The topic of this research is passive localization in adaptive networks.In this thesis, localization of single emitter in adaptive network is considered firstly.It is assumed that each sensor can respectively receive the signal transmitted by the emitter, estimate the noise variance and share information with its neighbors. Herein a noise-constrained distributed adaptive direct position determination (NCD-ADPD)algorithm is proposed to localize the emitter by exploiting the a priori knowledge of the noise variance. The NCD-ADPD algorithm turns out to be a variable step-size extension of the recently proposed distributed adaptive direct position determination (D-ADPD)algorithm, where the variable step-size scheme arises naturally from the noise constraint.Compared with its predecessor, the NCD-ADPD algorithm is endowed with markedly improved localization performance, which is validated by simulation results, at the cost of a slight increase of computational complexity. To further enhance the tracking ability of the NCD-ADPD, a change-detection mechanism is given to cope with an abrupt change of the emitter position.Additionally, localization of multiple emitters simultaneously in wireless sensor network is discussed. In this scenario, there are multiple emitters in network. Each sensor can receive signal transmitted by only one emitter. Via MSD optimization, the combination coefficients denoting reliability between neighbors from original D-ADPD algorithm is updated. Eventually the proposed multiple emitters’ localization algorithm is obtained. Along with the localization algorithm iterates, the weights update scheme gradually enhance cooperation between sensors localizing identical emitter, while weaken the links between sensors localizing different emitters. Numerical experiments verify that the proposed algorithm can cluster correctly, further can determine the positions of multiple targets simultaneously in adaptive networks.