Indoor localization based on RSS fingerprinting using Wi-Fi

Abstract

With the rapid advances in wireless technologies, indoor localization has been active in the field of research in the last 10 years, especially after the success of GPS in outdoor environments. Many indoor localization methods and algorithms have been proposed using different wireless technologies. However, there exists no wireless technology that can provide better infrastructure for localization as much as the popular and widespread Wi-Fi. And since signal triangulation methods do not perform well indoors, RSS fingerprinting seems to be the better choice for a localization method, but this method requires an accurate data matching and classification algorithm to perform location estimation. Support Vector Machine (SVM) is a strong machine learning algorithm for data classification and regression problems, and has shown to be the most accurate data classification algorithm in many applications especially in the case of large data sets. In this project, an indoor localization system based on Wi-Fi will be proposed. The localization method used will be a deterministic RSS fingerprinting method with a new MultiClass SVM algorithm (KM-SVM) for location estimation. The system will be tested using KM-SVM and the popular KNN as location estimation algorithms to compare the results.