Research On Key Technologies Of Indoor Location And Trajectory Tracking Based On Statistical Inference

The location and trajectory of target are the most basic information of WSN,and are the foundation of implementing location-based service.A satellite positioning system provides a solution for the problem of target localization and trajectory tracking in outdoor environment.According to statistics,most people spend more than 80%of their time indoors.Providing indoor location or trajectory with high accuracy is the foundation of many applications.Satellite wireless signals will be severely affected by buildings,walls and other obstacles,and cannot effectively work indoors.The problem of wireless network localization and trajectory tracking involves the related research fields of computer,communication,automation and physics.In order to provide indoor positioning service,researchers at home and abroad have proposed a variety of indoor positioning systems based on different technical principles in the past twenty years,but so far there is no leading role.The problem of wireless network indoor localization and trajectory tracking is very valuable in indoor application,and the problem has attracted many researchers' attention.The location information of the indoor moving object includes two aspects:the current position and the moving trajectory.At present,the proposed trajectory tracking model defines the trajectory of the moving object as the position sequence perceived through the localization devices in the field,that is,to perform continuous localization and then form the trajectory.The positioning error is unavoidable,and the original location information contained in the trajectory is not necessarily continuous in space.The reason for this phenomenon is that location information is produced in the adjacent locations.In this thesis,the problem of wireless network localization and trajectory tracking are studied based on position fingerprint.A trajectory tracking model consists of two major processing phases is proposed based on the prior knowledge that the corresponding physical position of successive fingerprints in time is necessarily adjacent in space.The localization and trajectory tracking are considered together in this thesis.The statistical characteristics of RSSI fingerprint vectors are fully utilized to mine the constraint conditions.Specifically,the main research contents and achievements in this thesis include:(1)A track tracking algorithm based on the temporal and spatial correlation of RSSI fingerprint vector is proposed for grid area.The original RSSI fingerprint vectors are treated as a number of time sequences classified according to the beacon nodes.In order to reduce the influence of noise and reflect the change trend of RSSI clearly,boundary time series are processed by moving least square to detect the spanning boundary events.RSSI time window statistics related to region are calculated according to the detected boundaries.Using the classifier based on the minimum error rate Bayesian decision rule,the grid area of an unknown node corresponding to the time window is determined.Then,the dynamic time warping RSSI fingerprint matching algorithm is used to obtain trajectory within the grid area with the heuristic information constraint.The principles of constraint conditions detection in spatio-temporal association trajectory tracking model are strictly proved in mathematics.The field experiments and the simulation experiments show that the algorithm has good environment adaptability,smooth trajectory,no outlier and the error does not accumulate among the regions.Compared with the existing methods,the accuracy of trajectory tracked is improved.(2)A trajectory tracking model based on Delaunay triangulation is proposed for arbitrary shape regions.A site is divided into a series of triangular areas,and the common edges of the adjacent triangular areas are treated as boundaries.Each Delaunay triangle is the optimal triangle made up of three nearest beacon nodes.Using the method of accumulating statistics based on successive APIT tests,the triangle region is determined by the method of moving least square fitting to the RSSI time series information of the nodes.The accuracy rate of the regional decision is raised from 86%to 95.4%.The location of switching Delaunay triangle of the unknown nodes is the boundary point.Based on the dynamic time warping location fingerprint matching constrained by heuristic information,the trajectories within the Delaunay triangle region are obtained.The field experiments show that the region can be accurately determined when the density of the nodes is>6 and the APIT test number is not less than 5 times within a single triangle region,thus forming a high precision trajectory with continuous smooth and no cumulative error.(3)A multilateration algorithm based on the optimal reference equation in grid area is proposed.In the case of the anchor node grid deployment,the region where the unknown nodes are located in a period of time is determined using the statistical characteristics of RSSI fingerprint vector.The corresponding beacon nodes attached to the grid region are used to construct the multilateral positioning equations.The least square solution is used as the positioning result.First,for the grid area,the selected regional vertices conform to the MaxMean criterion.Secondly,the optimal reference equation is chosen according to the wireless signal transmission model as small as possible.Finally,the equations set up in the grid region are linearized and have excellent characteristics.The field experiments show that the optimal reference equation multi location algorithm has higher positioning accuracy than the simple consideration of beacon nodes with MaxMean criterion.(4)An adaptive constrained convex programming algorithm based on RSSI ranging is proposed.In the case of the anchor node arbitrary deployment,the Delaunay triangle region where the unknown nodes are located in a period of time is determined by the statistical characteristics of RSSI fingerprints vector.According to triangulation theory,the Delaunay triangle is the most close to the equilateral triangle in all triangulation schemes,so these anchor nodes attached to the Delaunay triangle are selected to localize.On the one hand,considering the selection of anchor nodes in the triangle area,the effect of the triangles of different shapes in the location calculation is different.On the other hand,the Delaunay triangle selected by the algorithm is the closest to the unknown node in the area of the unit.Then,the adaptive constrained convex programming localization algorithm based on RSSI distance ranging is designed.The algorithm introduces adaptive mechanism to ensure that the overlap area can always be formed and the area of the overlapping area can be reduced effectively by regional restriction.The field experiment shows that the location accuracy of this algorithm is greatly improved compared with the typical convex programming localization algorithm.In summary,this thesis studies the regional information constructed by the spatial deployment of different anchor nodes,and uses statistical characteristics to fully mine the constraint conditions related to the region.The trajectory tracking models are constructed to gain the trajectory constrained by the constraint conditions.On the basis of heuristic information,the selection of anchor node is considered in the positioning procedure and the corresponding positioning algorithm is designed.These algorithms greatly reduce the influence of RSSI noise,eliminate the outlier out of the region,and obtain high precision position accuracy and track tracking accuracy.