Research On Fingerprint Update Algorithm Based On Crowdsourcing Data

With the wide application of the internet of things(IoT)technology,the demand for indoor positioning systems has grown rapidly with regard to location-based services.In the outdoor environment,GPS as a mainstream positioning solution can accurately locate users.However,in the indoor environments,GPS can not accurately calculate the actual location of users due to the shelter of walls.Faced with the increasing demand for location-based services,academia and industry have made in-depth research on indoor positioning,and various positioning algorithms have been proposed,including indoor positioning based on WiFi,Bluetooth,ultra-wideband,geomagnetic and visible light.Indoor positioning has once become a hot research direction of top conferences and journals.Considering the cost and accuracy of indoor positioning,WiFi-based fingerprint positioning technology stands out.As a cost-effective choice,WiFi-based indoor positioning has attracted great increasing research attentions because it does not require external devices installed in target environment.In addition,the deployment of WiFi is everywhere,which helps WiFi-based indoor positioning accurately perceive users in various locations and return the users accurate position.Even though WiFi-based fingerprint positioning technology has attracted increasing research attentions because of its ubiquitous deployment,in the complex environment and long-term deployments,the automatic adaptation of radio map has not been fully studied and there still remain some problems.When the locations of some APs change,the initial radio map will no longer be applicable and the positioning accuracy will drop heavily.In addition,with the passage of time,the temperature change and the position movement of furniture will also lead to the mismatch between the initial radio map and the current signal space environment,and the positioning accuracy will gradually decline.The traditional solution is to conduct site surveying regularly to update the outdated radio map,however the whole process is time-consuming and laborious,and requires the participation of professionals.Although some automatic updating algorithms are proposed to update the initial radio map,they mostly rely on additional reference points and inertial sensors,which increases the deployment cost and consumes more power.Aiming at automatically updating the radio map,we provide two solutions for fingerprint updating and the main work is as follows:(1)We first proposed a crowdsourcing positioning system based on ensemble learning,namely AAIFU.AAIFU automatically updates the radio map regularly and accurately estimate the user's location.In indoor positioning,the altered APs are the most important factor that causes the positioning accuracy to drop heavily.When altered APs exist in the environment,updating the RSS values of the altered APs in the initial radio map is the key to improving the positioning accuracy.In the AAIFU positioning system,we first detect and identify the altered APs that have changed location.After obtaining the altered APs,we use the relationship between the received signal strength of the altered APs and the unaltered APs to train a prediction model for updating the radio map.Besides,we have also solved the issue of device diversity in positioning.Our solution is simple and efficient,and does not rely on additional infrastructures and high power inertial sensors.We carry out a series of experiments in a teaching building.The experimental results show that our solution can effectively adapt to the altered APs,eliminating the impact of the altered APs and improving the positioning accuracy.(2)Considering the problem of outdated radio map caused by the altered APs and the slow changes of indoor environment,we propose an indoor positioning system AAIMSS.AAIMSS automatically updates the radio map and locates users based on our enhanced transfer learning approach Enhanced-TCA.The traditional transfer learning approach TCA cannot update the radio map when it contains outlier features.The visual representation of the outlier features are the altered APs,which will seriously interfere the searching of the mapping space.In the AAIMSS system,Enhanced-TCA can effectively eliminate outlier features between the source and target domains,and make the best of a small amount of labeled crowdsourced data in the target domain to search the mapping space.Our lightweight solution does not rely on other high power devices and inertial sensors.Experimental results show that AAIMSS can accurately deal with the problem of outdated radio maps caused by the altered APs and the slow changes of the indoor environment.