Research On Indoor Wi-Fi Positioning Algorithm Based On Crowdsourcing Fingerprint Approach

With the advent of the IoT(Internet of Things)era,more and more academic and commercial organizations have hegun to pay attention to LBS(Location-Based Services).Therefore,the research and development of IPS(Indoor Positioning System)have made great progress.Most IPS require site survey and sampling to meet the requirements of indoor LBS for positioning accuracy.However,site survey and sampling are very time-consuming and laborious,which brings limitations and challenges to the realization of scalable IPS.At present,academia has proposed methods based on semi-supervised and unsupervised crowdsourcing to reduce the cost of site surveys.The semi-supervised method still requires a certain amount of site survey and sampling,while most of the existing crowdsourcingbased unsupervised algorithms need to introduce additional sensors,such as inertial sensors built into mobile phones.Different from the traditional WiFi fingerprint positioning method,we proposes an indoor positioning framework based on crowdsourcing to collect WiFi data.In view of the manifold hypothesis for the true RSS distribution of WiFi,it can be reasonably considered that the point cloud of RSS observation data collected by crowdsourcing is embedded near the two-dimensional manifold.Based on the assumption that the RSS observation data point cloud is embedded near the two-dimensional manifold,we proposes an iterative method to reconstruct the two-dimensional manifold from the RSS observation data point cloud which based on the idea of elastic map.Different from the previous method of assigning location labels to crowdsourced data,after obtaining the manifold,we regard the nodes of the twodimensional manifold and the internal topological coordinates of the manifold as the proxy of the standard radio map which sampled in site survey.In the location stage,we propose a weighted positioning algorithm with the manifold distortion effort as the weight.The manifold distortion effort integrates both measurement error and manifold shape preservation.,and it can reflect the degree of matching between measurement and nodes in manifold.We verify the proposed positioning framework in simulation and real indoor environments.Under different noise levels of the simulation,the robustness of the manifold reconstruction method is verified by visual analysis.In addition,on the basis of crowdsourcing fingerprints,the positioning algorithm proposed in this paper is compared with other positioning methods.The results show that the positioning algorithm in this paper is better than other schemes in the comparison experiment in terms of evaluation indicators,no matter in simulation or in real experimental environment,which verifies the feasibility and stability of the algorithm in this paper.