Research On Distributed Adaptive Kalman Filter In Wireless Sensor Networks

With the rapid development of sensor technology,wireless sensor networks(WSNs)in recent years has gradually attracted attention and research,has been more and more put into social lives,with great convenience.Distributed signal processing is the most popular signal processing and filtering method in wireless sensor networks due to its low energy consumption,flexible network structure and good robustness.WSNs work in a distributed way,which requires sensor nodes in the network to communicate and share information with a certain cooperation strategy.Among them,the diffusion strategy requires each sensor node to communicate only with other nodes in its one hop communication withf low communication burden,so it has higher research value in WSNs.Although WSNs have a wide range of applications,due to the limitations of the energy of the sensor itself,the rapid growth of the amount of data and the interference of the wireless environment,the sensor generally has random uncertainties,among which the energy constraint and data loss are the most significant.Therefore,based on this fact,this work will study the state estimation problem under energy bandwidth constraints and measurement data loss.Firstly,this work will use the corresponding mathematical model and parameters to elaborate on the WSNs in detail.In addition,the state-space model of WSNs is studied by introducing Kalman filter and distributed Kalman filter.Secondly,the bandwidth of communication between sensors is usually limited due to the power limitation of the sensor in WSNs.Considering quantization is an effective method to save communication bandwidth,we adopt a dither quantization to reduce the communication cost.Besides,based on the minimum mean square error criterion and diffusion strategy,the optimal local gain and neighboring gain are designed to fusion the raw and quantization information,and a distributed diffusion quantization Kalman filter algorithm is proposed.In addition,we also analyze the mean performance and mean square performance of the proposed algorithm,and find that covariance is bounded under variable bandwidth.The effectiveness of the proposed algorithm is verified by numerical simulation.Finally,a distributed diffusion Kalman filter under unknown missing measurements is proposed in this work.The missing measurements are established as a stochastic process that subjects Bernoulli distribution.According to the probability model,a Bayesian hypothesis test is proposed,which can detect whether the measurements are lost or not.Then,based on the detection results,the prior estimation of nodes is used to replace the part of missing measurements,and the diffusion kalman filter under missing measurements is proposed.Then,the performance of Bayesian hypothesis test,the mean performance and mean square performance analysis of the distributed diffusion Kalman filter are analysed.The experimental results show that the proposed algorithm is robust under unknown missing measurements.