Research On Improvement Of Distributed Particle Filter Tracking Algorithm Based On Wireless Sensor Network

With the development of sensor technology and information technology,wireless sensor networks are widely used in areas such as smart cities and smart transportation.Target tracking is a key task of wireless sensor network,and its important research content is filtering tracking algorithm.Among them,distributed particle filter tracking algorithm can adapt to complex tracking environment due to its nonlinear non-Gaussian point,and it has received extensive attention in recent years..However,the distributed particle filter tracking algorithm will cause excessive node communication traffic,increase wireless communication energy consumption,and affect the working life of the wireless sensor network.Therefore,how to reduce the communication traffic has become an urgent problem to be solved by the distributed particle filter tracking algorithm in the wireless sensor network.In this paper,combining the actual needs of wireless sensor network nodes with limited communication flow and energy consumption,the distributed particle filter tracking algorithm is the core of the subject.Aiming at the problem of excessive communication flow of sensor network nodes caused by the algorithm,an improved distributed particle filter tracking algorithm is proposed.At the same time,in order to verify the actual performance of the improved algorithm deployed in the wireless sensor network,a dedicated simulation platform was designed.Specifically,in terms of theoretical innovation,based on the in-depth analysis of the principle of distributed particle filter tracking algorithm,this paper proposes a method based on BP to solve the problem of synchronous sampling of node particles and excessive communication traffic caused by the local weight of interactive particle samples.Improved distributed particle filter tracking algorithm based on neural network approximate likelihood.The algorithm fixes the random seed used in the particle importance sampling and resampling process,and chooses to use the global weight to filter the particle samples in the particle resampling stage,so as to realize the particle synchronous sampling mechanism and ensure that all nodes can obtain the same batch of particle samples.On the basis of particle synchronous sampling,a BP neural network model is designed and trained to map a large number of local weights into a small number of model parameters,and instead of local weights to participate in the node interaction process,the purpose of reducing the communication traffic between sensor nodes is achieved.The simulation results show that although the tracking accuracy of the improved algorithm proposed in this paper decreases slightly,it significantly reduces the communication traffic between nodes.In terms of engineering innovation,research and design a wireless sensor network simulation platform dedicated to target tracking.Compared with the existing simulation platform,the node energy model associated with the internal components of the sensor node is introduced,and the sensor network structure and sensor node are supported.The flexible configuration of internal components is more in line with the target tracking scenario requirements of this article.The design of the platform follows the principle of modularity,and each module is subdivided according to its function,with a friendly user interface and strong scalability.