PROJECT TITLE :

A Comprehensive and Reproducible Comparison of Clustering and Optimization Rules in Wi-Fi Fingerprinting

ABSTRACT:

The use of Wi-Fi fingerprinting as a method for indoor positioning is a well-established practice. It makes use of a pattern recognition technique that makes use of a similarity function in order to compare the operational fingerprint that was captured with a set of reference samples (radio map) that had been collected in the past. The matching algorithms have a scalability problem in large deployments with a huge density of fingerprints because this results in an unmanageably high number of reference samples being stored in the radio map. This paper presents a comprehensive comparative analysis of existing methods, with the goal of reducing the complexity as well as the size of the radio map that is used during the operational stage. According to the findings of our empirical research, the majority of the methods reduce the amount of work that needs to be done on the computer, but they do so at the expense of accuracy. Only k-means, affinity propagation, and the rules based on the strongest access point strike the right balance between the positioning accuracy and the amount of computational time required among the methods that were investigated. In addition to the comparative results, this paper presents a new evaluation framework that makes use of multiple datasets. The purpose of this framework is to obtain results that are more general and to contribute, going forward, to the increased reproducibility of newly proposed solutions.