Distributed Estimation And Tracking Based On Wireless Sensor Network

In recent years,wireless sensor network(WSN)is widely used in the positioning,monitoring and tracking of the system.The application of WSN has been popularized in military,agricultural production,ancillary industrial production,infrastructure status monitoring,intelligent transportation system,medical system,ecological monitoring,smart home and other fields.Distributed estimation based on WSN is the basic problem of the location monitoring and tracking and the key point of current research.Consensus protocol is wildly used in design of the distributed estimation algorithm.This dissertation studies the weight design problem in the consensus based Kalman filtering algorithm and the distributed target tracking problem.Traditional consensus problem is mainly based on the adjacent matrix and Laplace matrix of the topology.However,it needs the information of the whole network and the weights to compute the eigenvalues of the Laplace matrix of the topology.It will be more difficult to design the proper weights to configure the Laplace matrix.The main contributions of this dissertation are the following:1?In distributed estimation problems,the target node is required to be globally reachable.This dissertation mainly proposes a static weight design method based on the nearest path lengths of the nodes,which can be obtained from the nearest path theorem(Dijkstra theorem,Floyd theorem,Bellman-Ford theorem,etc).This kind of weight design method is efficient and easy to calculate.2? The traditional distributed filtering algorithms are mainly leaderless.This dissertation uses the leader-follower filtering protocol.For the leaders(nodes that can direct measure the target),we directly use the Kanlman filter which is the optimal state estimation protocol.For the following nodes,we use weighted average consensus filtering algorithm based on the weight design.We further provide the stability condition of the adjusting parameter in the weights.We also design the grading method of the following nodes based on the leader-follower filtering method,and cut off some nodes of the network to obtain the optimal subgraph.This approach ensures that only the nodes that are nearer or equal to the target are used in the consensus algorithm.This design approach can reduce the estimation errors and accurate the convergent speed.3?Consider the problem of network delay,and package loss,we study the influence of the network delay on the estimation and tracking of the target.We use the delay compensation strategy to design the consensus Kalman filtering to reduce the negative effect of the time delay.And we give the conditions of the adjusting parameters in the weights to guarantee the stability of the algorithm.We also compare the filtering algorithms by simulations.4?We study the distributed estimation and tracking of the target with unknowing periodic signal input.We propose a target tracking algorithm based on the distributed estimator,and divide the design into the estimation and the tracking.Firstly,we use the Fourier harmonic to model the periodic signal input,and treat the Fourier coefficients as a new state vector to be estimated.And then we design the consensus estimation algorithm so that all nodes can get estimation of the target state and the unknown periodic signal input.Then we apply the model reference adaptive control(MRAC)to design the tracking algorithm of the target based on the estimation.