An optimization model for target tracking of mobile sensor network based on motion state prediction in emerging sensor networks

Abstract

According to the defects of the standard particle filter algorithm in target tracking of mobile sensor networks, such as low accuracy, large network energy consumption and poor anti-noise ability, an optimization model is proposed for target tracking of mobile sensor network based on motion state prediction. First, centroid algorithm was adopted to construct the node localization model, and then the features of the position and the direction of the moving target in the mobile sensor network were as the measurements. The method of integral point assignment was adopted to self-adaptionly optimize the weights of the standard particle filter algorithm, and the introduced modifying factor, the value assignment of integral point was for self-adaption correction, then the difference between the observed and predicted values of the system was provided news residual interest knowledge in the re-sampling phase, to self-adaption modify the sampling particles by measuring the news. And then improve the operation efficiency of the particle filter algorithm with asymmetric kernel function, and provide new residual interest knowledge with the difference between the system current time and forecast values in the re-sampling phase, self-adaptive adjusting of sampling population through measuring the new rates. The simulation experiments show that the proposed improved particle filter algorithm has the higher accuracy and better stability for target tracking, and has lower energy consumption of the network.