Research On Fusion Algorithm For Wireless Sensor Networks Information Based On Estimation

With the development of communication technology and sensor technology,wireless sensor networks are increasingly used in both military and civilian fields,such as,target tracking,environmental monitoring,auxiliary navigation,disaster forecast,resource exploration,and accident investigation.Limited by cost,working environment and communication capability,it is particularly difficult to fuse wireless sensor network information by using conventional state estimation methods.It is necessary to consider the problems of energy limitation,bandwidth limitation,non-Gaussian noise,and unknown noise variance.Commonly used multi-sensor information fusion strategies include the centralized strategy with a unique fusion center and the decentralized distributed strategy.In the centralized strategy,the fusion center collects and centralizes the information of all sensors.Its communication and computing capabilities limit the number of sensors available,and are suitable for scenarios with small sensor scales.On the other hand,in distributed strategies,each sensor only interacts with its neighboring sensors,which can extend the sensor's sensing range,improve system reliability,and reduce the cost of a single sensor.It is suitable for applications with limited energy and bandwidth,large sensor volume,and wide spatial distribution.This paper studies the fusion estimation of wireless sensor network information under non-ideal conditions from the perspective of centralized and distributed,respectively,and verifies the effectiveness of the proposed algorithms through typical target tracking applications and other typical simulation scenarios.The main research contents are as follows:A multi-sensor fusion estimation algorithm based on event-driven strategy is proposed for multi-sensor information fusion under energy-limited conditions.Firstly,an event-driven strategy of dead-zone type event is defined.And then the basic filtering algorithm is derived by using some conclusions of Tobit model.With the help of its information filter algorithm,an event-driven multi-sensor fusion estimation method is presented.In this algorithm,the sensor transmits data to the fusion center only when the defined event is triggered,and the energy efficiency is improved by reducing the communication frequency.The simulation results show that the proposed filtering algorithm is equivalent to its information form,and its estimation accuracy is better than several existing algorithms.The proposed event-driven multi-sensor fusion estimation algorithm can significantly reduce the communication frequency between sensors and the fusion center.The state estimation algorithm based on maximum correntropy is proposed for systems with non-Gaussian noises.Considering that the existing maximum correntropy filtering algorithm is easy to diverge,a numerically stable maximum correntropy unscented Kalman filter algorithm is obtained by defining a new cost function,and its information filtering algorithm is also derived accordingly.In order to further improve the estimation accuracy of the proposed algorithm,based on the more general cost function of filtering update,the iterative maximum correntropy filtering algorithm is obtained through Gauss-Newton and LevenbergMarguardt methods,respectively.The simulation results show that the numerical stability of the proposed algorithm is better than the existing ones,and the proposed iterative algorithm can further improve the estimation accuracy.An adaptive interactive multi-model state estimation algorithm based on variational Bayes is proposed for systems with unknown noise variances and inaccurate models.The process and measurement noise variances are modeled as inverse Wishart distribution.Under the framework of the interactive multi-model algorithm,the variational Bayesian method is used to simultaneously estimate the system state,process noise variance and measurement noise variance,and then use the weighted Kullback-Leibler method to fuse the estimation results of each model.Finally,we can obtain the adaptive interactive multi-model algorithm based on variational Bayes.The simulation results show that the proposed adaptive interactive multimodel algorithm can approximate the classic ones with real noise variances,and the amount of computational burden of the proposed algorithm is moderate.A fusion estimation algorithm based on diffusion strategy is proposed for the distributed fusion estimation problem of linear systems.Based on the equivalent deformation of the existing diffusion Kalman filter algorithm,the track fusion type diffusion Kalman filter algorithm considering error covariance propagation is derived and its simplified algorithm is also given.The unbiasedness and convergence of the proposed algorithms are analyzed by means of mathematical induction.The simulation results show that the accuracy of the proposed track fusion algorithm is better than the existing diffusion algorithms,and its simplified algorithm can realize periodic information fusion and further reduce the communication frequency of the algorithm.A sigma point approximation distributed algorithm based on hybrid consensus strategy is proposed for the distributed fusion estimation problem of nonlinear systems.For the nonlinear sensor network with intermittent measurements,the diffusion cubature Kalman filter algorithm with intermittent measurement is derived firstly.Based on that,a statistically linearized method is used to obtain a hybrid consensus distributed filtering algorithm for nonlinear systems using sigma point approximation,and its stability is analyzed by means of statistical linearization using recursive method.The simulation results show that the proposed algorithm has better estimation accuracy than the existing related methods,and the estimation accuracy of the proposed algorithm can approximate that of the centralized algorithm as the number of consensus iterations increases.A distributed maximum correntropy fusion estimation algorithm for systems with nonGaussian noise is proposed.On the basis of the previous maximum correntropy algorithm,by defining the maximum correntropy cost function considering multiple sensor measurement equations,the maximum correntropy distributed fusion estimation algorithm for linear systems is obtained with the help of the information filter.And then the distributed linear algorithm is extended to nonlinear systems.The simulation results show that the proposed distributed maximum correntropy algorithms can significantly suppress the pollution of non-Gaussian noise,and the estimation accuracy is better than the existing conventional distributed algorithms.Besides that,they can approximate the results of the centralized maximum correntropy filtering algorithms.