Research On Distributed Cooperative Localization And Tracking Method In Wireless Sensor Networks

Wireless sensor network has become the most important research area in modern technologies,and location awareness capability is one of most important technique in wireless sensor network.This technique has been widely used in many frelds including military defense,unmamed navigation and mobile robot localization.In cooperative localization process,conventional centralized localization technique has large communication overhead and computation complexity and is not robust to the agent node failures.Novel cooperative localization techniques has better scalability and accuracy by using the potential of agent nodes,which has better applacation prospects in wireless localization.In this paper,we focus on the research of distributed cooperative localization and tracking method in wireless sensor network.Firstly,it research the distributed mobile agent cooperative localization techniques under nonlinear Gaussian system.Then,it discusses distributed agent cooperative localization and object localization techniques in nonlinear non-Gaussian system.Finally,the research be extended to distributed agent cooperative localization and object tracking.The main contributions are presented below:Aiming at the problem that conventional cooperative localization algorithm has poor accuracy and large computation complexity in nonlinear Gaussian system.A distributed cubature kalman filters cooperative localization algorithm based on belief propagation is proposed.Factor graph be created according to joint posterior probability density function(PDF)of all agent nodes position variables,and belief propagation be used to compute the position posteriors distribution of each node on factor graph.Node's position satisfies Gaussian distribution and measurement mode nonlinearity,as aresult,the closed-form analysis of the marginal posterior distribution cannt be obtained.We are utilized Gaussian parameter belief propagation with reconstruction method,and cubature kalman filters algorithm,and derived multivariate Gaussian distribution model for each belief and posterior distribution of nodes on the factor graph.It is proposed algorithm only need to transmission mean vector and covariance matrix among the adjacent agent nodes.Therefore,lower communication overhead and computational complexity are more feasible in practical applications.Aiming at the problem that conventional cooperative localization algorithm has large communication overhead and computation complexity by use nonparametric belief propagation in nonlinear non-Gaussian system.A distributed cooperative localization algorithm based on parametric belief propagation is proposed.The hyperbola asymptote model be introduced for estimated position distributed of unknown position nodes by utilization measurement capability of anchors.Double-layer factor graph belief propagation model be constructed according to joint posterior PDF of all nodes position variables.The nodes position distributed estimation is achieved through belief position propagation in the first-layer factor graph.We are utilized the result of position distributed as priori information,and parametric belief propagation be used to compute the position posteriors distribution of each node position variables,and derived the parameter compute expression for the variable node and factor node on the second-layer factor graph.Proposed algorithm only need transfer parametric belief among adjacent nodes for reduce communication overhead and computation complexity,and improve the feasibility of cooperative localization techniques in practical applications.Aiming at the problem that agent localization and object tracking in mobile networks,a novel cooperative localization model be extended to cooperative localization and object tracking.A distributed cooperative localization and object tracking algorithm base on nonparametric belief propagation is proposed.Factor graph be created according to joint posterior PDF of all nodes,and belief propagation algorithm be used to compute the posterior distribution of each node.Because,node's position satisfies non-Gaussian distribution and measurement mode nonlinearity,the closed-form calculation of the posterior distribution cannt be obtained.It is proposed a low computation complexity nonparmametic belief propagation with reconstruction method,and computed the local joint probability density distribution by combining high-dimensional nonparametric message.The marginalization method be preformed to compute position marginal posteriors distribution of each node.Since the objects characteristic are non-cooperative and passive tracking,the belief propagation and computing on the factor graph related to the objects are required to complete by all adjacent agents and anchors.Thus,it is achieved agent cooperative localization and object tracking.