Research On Sensor Network Decision Fusion

Sensor network decision fusion is an important issue of sensor network data fusion. It is a detection technology which fuses the information from all the sensors in the sensor network, and makes decision of some target or event. Recently, with the development of sensor networking technology and signal detection theory, the sensor network decision fusion issues are attracting more attention. This paper focuses on the sensor network decision fusion issues based on distributed fusion mode. The optimization method of sensor network decision fusion systems with correlated observation and decision fusion rules of resource-constrained wireless sensor networks (WSNs) under nonideal channels have been deeply studied. This paper could be divided into six chapters, which are organized as follows:In Chapter 1, the basic concepts and research status of sensor network decision fusion are briefly introduced, the existing problems of the current research on sensor network decision fusion are also pointed out. Finally, the content arrangement of the research on sensor network decision fusion in this paper is given.In Chapter 2, a study on the optimization design of the sensor network decision fusion systems with correlated observation and fixed decision fusion rule is made based on Neyman-Pearson criterion. Firstly, the traditional optimization methods based on Newton-Raphson and discrete iteration gradient algorithms are briefly introduced. Then, a new optimization method based on artificial immune strategy algorithm is proposed.The performance of this method is compared with that of the above two traditional methods by numerical experiments.In Chapter 3, the decision fusion rules for WSNs using censoring information transmission strategy are studied under fading channels with different distributions. Firstly, the likelihood ratio (LR) decision fusion rule based on instantaneous channel state information (CSI) is introduced which can provide optimal detection performance. Then, a sub-optimal LR decision fusion rule based on channel statistics (CS) is derived under Rice fading channels. Moreover, when the channel signal-to-noise ratio (SNR) is low, two reduced nonparametric decision fusion rules based on this sub-optimal decision fusion rule are also given. The detection performance comparison between these sub-optimal decision fusion rules and optimal decision fusion rule is made using simulation. Subsequently, existing sub-optimal LR decision fusion rule based on series expansion and nonparametric decision fusion rule of censoring based mixed fusion are briefly introduced under Nakagami fading channels. Then a sub-optimal LR decision fusion rule based on Laplace approximation is proposed. The detection performance of the proposed sub-optimal decision fusion rule is compared with that of existing sub-optimal decision fusion rules and optimal decision fusion rule by simulation.In Chapter 4, in order to solve the problem of long-distance transmission of sensors'local decision, WSNs using multihop information transmission mode are constructed. The optimal LR decision fusion rule based on CSI and sub-optimal LR decision fusion rule based on Rayleigh CS are successively derived for multihop WSNs using traditional information transmission strategy. Furthermore, two reduced nonparametric decision fusion rules based on the sub-optimal decision fusion rule are given at high and low channel SNR, respectively. The detection performance analysis of these decision fusion rules is made using simulation. Then, for the sake of reducing the power consumption, the censoring information transmission strategy is introduced into multihop WSNs. The optimal LR decision fusion rule based on CSI and sub-optimal LR decision fusion rule based on Rayleigh CS are derived under censoring multihop WSNs. Meanwhile, a reduced form of sub-optimal decision fusion rule is given at low channel SNR. The detection performance of sub-optimal decision fusion rules is compared with that of optimal decision fusion rule using simulation.In Chapter 5, the decision fusion rules for dynamic large-scale WSNs are studied under Rayleigh fading channels. Firstly, under parallel channels architecture, false alarm controlled and voting decision fusion rules are derived for dynamic large-scale WSNs using traditional and censoring information transmission strategies, respectively. The detection performance comparison between these decision fusion rules and existing nonparametric decision fusion rules is made using simulation. Then, under multiaccess channels architecture, quadrature square sum decision fusion rule is derived. The detection performance of this decision fusion rule is compared with that of LR decision fusion rule based on quadrature components'square sum using simulation.In Chapter 6, a summary of the work in the whole paper is given and a prospect of the issues of sensor network decision fusion is also presented.