Research On Sensitivity-based Algorithm For Indoor Geomagnetic Localization

Now most of human activities have been happened indoors.And the demand for location services and location-based applications is increasing.Indoor positioning technology is gradually becoming the current research focus.Indoor position will be confronted with three aspects.Firstly,multiple obstacles and interference sources,variability of indoor circumstances;secondly,measurement error,indoor environment is composed of many available sources superposition;lastly,dependency on hardware devices and limitations of the devices.Based on the difficulties above and in view of distribution characteristics of the geomagnetic field.In indoor area,without of any interference sources,geomagnetic field is stable or changes weakly.But many sources of interference are random distributed in the real indoor space which forms a unique distribution of indoor magnetic field.Even in the partial corners,the magnetic field strength can still reach the average.In this paper,we use the interference source to complete the localization process.In order to eliminate the interference of auxiliary equipment,we use the eigenvalue gradient to establish the feature fingerprint database.However,geomagnetic field which used as a positioning carrier still faces two major problems.First,for a certain geomagnetic field eigenvalue measurement,the magnetic field characteristics of the distribution is obvious but the measurement error is greater.Second,smooth areas will also exist in complex indoor distribution,which are known as the "blind zone",for example,corridors,aisle.In order to solve the above two problems,in the indoor nonlinear environment,the particle filter which is based on mathematical statistics method uses overall sample distribution to estimate the true position distribution.Using particle filter to match the more obvious feature distribution.In addition,this paper establishes the pre-treatment process to carry on area division.After the recognition is completed,we establishes the adaptive particle filter to solve the "blind zone" localization problem.The paper chooses a indoor space which covers 400 square meters.We construct four kinds of databases of fingerprint characteristics by four interpolation methods which are called nearest,cubic,spline,linear,after gathering intensity and X,Y,Z three-component values of geomagnetic field by smart-phone.Then we divide the space into some small area and utilize the measured values to find the best match in the feature database.Finally,integrating perceptual information into particle filter to build the adaptive processing.The processing makes filter adjust filter's parameters to match different feature points automatically on the basis of original.This pre-treatment adds the relation between pointsand feature databases to particle filter and makes the recognition process fast and accurate.This paper has finished three aspects of the locating problem.Constructing indoor maps by a finite number of sample points.Then dividing indoor space into several subspace by sensitivity based on the relation between measurement and feature distributions.Finally,integratin information of relation between points and maps into particle filter,optimization of particle filter parameters automatically.After a series of tests,the results indicate that positioning errors of cubic,spline,linear are better than nearest's.Further,comparing to the traditional particle filter,the estimation accuracy is more stable.At different resolutions,positioning error have an average loss of0.51 m,1.09 m,1.54 m,2.02 m respectively,after 100 random tests.