Voronoi Diagram and Crowdsourcing-Based Radio Map Interpolation for GRNN Fingerprinting Localization Using WLAN

Abstract

In the last decade, fingerprinting localization using wireless local area network (WLAN) has been paid lots of attention. However, this method needs to establish a database called radio map in the off-line stage, which is a labor-intensive and time-consuming process. To save the radio map establishment cost and improve localization performance, in this paper, we first propose a Voronoi diagram and crowdsourcing-based radio map interpolation method. The interpolation method optimizes propagation model parameters for each Voronoi cell using the received signal strength (RSS) and location coordinates of crowdsourcing points and estimates the RSS samples of interpolation points with the optimized propagation model parameters to establish a new radio map. Then a general regression neural network (GRNN) is employed to fuse the new and original radio maps established through interpolation and manual operation, respectively, and also used as a fingerprinting localization algorithm to compute localization coordinates. The experimental results demonstrate that our proposed GRNN fingerprinting localization system with the fused radio map is able to considerably improve the localization performance.

Developing a Solution for Mobility and Distribution Analysis Based on Bluetooth and Artificial Intelligence

“…Whatever application is intended to be used for, if RSSI is the parameter that is to be monitored to estimate distance, and to further calculate position of a mobile station with regard to a specific monitoring (fixed) station, a proper model for signal propagation would be always useful. Some authors [19] recommend certain models, based on experiments performed to determine the best suitable behavior for signal attenuation. One of these is denoted by:…”

“…Other research involves different procedures for improving the localization results [ 14 , 15 , 16 , 17 , 18 ]. The authors of [ 19 ] propose an optimized propagation model parameter for each Voronoi cell using the RSS and location coordinates of crowdsourcing points and estimate the RSS samples of interpolation points with the optimized propagation model parameters to establish a radio map—and reducing processing time. In [ 20 ] the authors present an algorithm named Dynamic Topology Detector (DTD) for extracting a GVD with topological information from a grid map.…”

“…Whatever application is intended to be used for, if RSSI is the parameter that is to be monitored to estimate distance, and to further calculate position of a mobile station with regard to a specific monitoring (fixed) station, a proper model for signal propagation would be always useful. Some authors [ 19 ] recommend certain models, based on experiments performed to determine the best suitable behavior for signal attenuation. One of these is denoted by: where d is the distance between the communicating devices, n represents the signal transmission constant, AE is defined as the absolute energy—A comparison factor in terms of energy, taken at 1 m distance from the measuring point.…”

Indoor Localization within Multi-Story Buildings Using MAC and RSSI Fingerprint Vectors

“…In contrast to the above, [15,16] used differential barometric altimetry to assist with floor identification and achieved a judgement rate of 98%. In addition, in order to improve the matching efficiency of online positioning, different clustering algorithms were introduced [17,18,19,20,21,22,23], such as interior geometric clustering algorithm. Some researchers considered APs (access points) as Voronoi diagram’s generators to create Voronoi cells and used these cells to cluster fingerprints of database [22,23].…”

“…In addition, in order to improve the matching efficiency of online positioning, different clustering algorithms were introduced [17,18,19,20,21,22,23], such as interior geometric clustering algorithm. Some researchers considered APs (access points) as Voronoi diagram’s generators to create Voronoi cells and used these cells to cluster fingerprints of database [22,23]. After that, the location area of mobile nodes was determined by using the strongest access point, and finally used the dynamic KNN (K-nearest neighbor) algorithm to implement the fingerprint clustering algorithm.…”