Research On Distributed Cooperative Location Method For Sparse Dynamic Communication Network

Reasonable and effective fusion of multi-source information from different sensor nodes is a crucial part of the collaborative positioning method.Multi-source information is integrated through a communication network composed of multi-sensor nodes.At present,sparse and random dynamics are typical instability factors in communication topologies.This article focuses on these two unstable factors:1.For sparse Wireless Sensor Networks(WSN),that is,sparse communication networks,according to the effectiveness of DKF for estimating the state of targets in sparse networks,we introduced the distributed information filter based on unscented transformation.And,this paper deduces the form of Unscented Kalman Filter under multi-sensors in detail,finally is combines it with consensus filter Distributed Unscented Information Filter(DUIF)to gather up it to nonlinear systems.The simulation proves that the state estimation under the distributed structure can realize the estimation result of the centralized structure at each node.Compared with the real track under the simulation,the distributed information filter based on the scaled unscented transformation is verified to be suitable for sparse WSN in the non-linear system.2.There are sometimes useless sensor nodes in sparse WSN,especially when applied in a large-scale and complicated actual environment.At this time,DUIF will tend to diverge from the average consistency error due to invalid multi-source measurement information.To solve this problem,we propose a DUIF algorithm based on weighted average consistency filtering.Under the cascade structure of the original DUIF algorithm,the upper and lower filters are changed to Local Unscented Information Filter(LUIF)and Weighted Average Consensus Filter(WACF).LUIF only obtains the multi-source information sent by neighboring nodes of the sensor node to obtain the local information matrix and information vector,and then uses them as the input of the weighted average consistency filter,thereby obtaining the one that does not contain average consistency error posterior estimation of target state in distributed structure.Among them,the weighted average consistency filter uses a quasi-Laplace matrix as a weight,and an information matrix and an information vector as input filters.It also proves that the input of the average consensus filter is changed,and the weighted average consensus value that does not affect its output is the weighted value of the posterior estimate of the target state.Finally,the effectiveness of the improved algorithm is verified by simulations comparing with the original DUIF algorithm.3.In order to solve the sparse and random dynamic problems in WSN,we propose a parallel fusion DUIF algorithm.On the basis of the improved DUIF algorithm proposed before,we changed the cascade structure of its two filters to a parallel structure.At the same time,an instant filter update mechanism is established to prevent possible asynchronous problems of the filter.LUIF iteratively filters through the communication topology itself,and no longer takes the posterior state estimation output by WACF as input.WACF performs a weighted average consistency process on the real-time local prediction information matrix and information vector output by LUIF,and finally obtains a distributed target posterior state estimation that weakens the adverse effects of sparse WSN.In addition,we added rate convergence scheme while WACF was running.It can increase the mean square convergence rate of WACF,so as to improve the global efficiency of algorithm.Finally,The simulation results show that the algorithm can efficiently track the target in sparse dynamic WSN.