Research On Key Technologies Of WLAN-Based High Precision Indoor Location System

With the rapid development of economic and increase of large scale single-family construction and large venues,a variety of indoor location based service(LBS)has grown.Satellite positioning system which is widely used outdoor cannot meet the need of indoor positioning accuracy because the satellite signal is blocked by obstacles.It restrict the development of application technology of LBS.Requirement of indoor LBS and wide coverage of wireless local area network(WLAN)make that indoor positioning has important practical implications and wide application prospects.There are two main classes of WLAN positioning techniques:ranging technique and fingerprinting technique.Fingerprint technique is recently adopted extensively in the indoor positioning system because of its high degree of accuracy and its anti-interference capability features.There are still a lot problems in rapid positioning,fingerprint location database construction,positional accuracy.We studied to these problems,including:Research on clustering and modeling method of dimensions reducing and refactoring for fingerprint location database and research on positioning algorithms of mesh matching in time-space domain.The main work and innovative results of the paper are as follows:1.Signal clustering method based on minimum round coverIn fingerprinting positioning technique,we need cluster the fingerprint location database to constrain the positioning area to local as much as possible.To accelerate the positioning,traditional K-means method cluster the fingerprint database according to the signal correlation feature,but fingerprint points with similar signal characteristics may be far apart in physical space.It results in fingerprint matching offset and influence the positioning accuracy.We propose an off-line location fingerprint clustering method based on minimum round cover to overcome this problem.We cluster neighboring spatial fingerprint points to one or several cluster on the neighborhood relation in spatial.Experimental results show that the clustering method based on minimum round cover has good continuity in position fingerprint.It reduces the positioning error while the positioning speed is improved.2.Modeling method of redundancy positioning database based on GrDOP dimensionality reductionThere is much redundant information in the process of fingerprint database.It is a problem to optimize the redundant information modeling on the premise of not reducing accuracy.We propose a redundancy positioning database dimensionality reduction modeling method on the base of GrDOP.After analyzing the relationship of the signal strength changing slop and its contribution to the positioning results.The redundant information with high level contribution is conserved and the low is dropped.We reduce the mesh matching operations without loss of positioning accuracy.With the proposed method the fingerprint database is reduced and mesh matching time is shorter and the system performance is improved.3.Mesh characteristic matching positioning algorithm in signal domain-space domainThe traditional position fingerprinting techniques,such as KNN algorithm,calculate the position by measuring the similarity between the positioning point and the reference point through calculating signal Euclidean distance.Signal space is nonlinear and calculating similarity with this method will lead to positioning error.We propose a mesh matching positioning algorithm based on signal domain-space domain.We change the signal space to distance space to calculate the similarity.The nearest grid is selected and the precise position is got by weighted calculating.We verify the feasibility and effectiveness of the proposed algorithm by lots of experiments,simulations and comparison with other algorithms.In addition,WLAN signal filtering and acquisition method are researched.An indoor positioning platform applied to large gymnasium is developed,which consists of a server,several mobile hand-held terminals and location fingerprint database.The architecture,function,software and implementation of all modules are described in detail.The system experiment scene and performance of the system are also given.Finally,we conclude the research work and give the next research plan in the future.