Comparative analysis of indoor localization using machine learning models

Indoor localization has become one of the most talked about services in today’s technology. We have observed that there have been huge demand of Indoor Location services due to increase in smartphone market in last few years. GPS is widely used to find the real time location information of different mobile users mainly in outdoors. This is the main reason there is a large demand for real time location prediction of various mobile users. However, GPS is not effective in indoor buildings due to its weak signal strength and has been found in the research that GPS does not work properly in indoor environment. Wi-Fi access points are widely used in indoor localization techniques which are based on Wi-Fi fingerprint data. Many research papers have been proposed on Wi-Fi fingerprinting based methods using its signal strength generated from Wi-Fi access points. They have proposed that Wi-Fi fingerprinting can increase indoor localization accuracy using different methods like collected Wi-Fi signal strength in various locations. Indoor localization using machine learning techniques is still an open area in which so much research is still going on to find out the best indoor localization. Our goal in this project is to do the comparative analysis of indoor localization using the different machine learning models. We will collect Wi-Fi fingerprinting data which has different Wi-Fi signal strength, and access points on different locations using android application in an android smartphone Motorola G4. After then, we will implement and evaluate the indoor positioning accuracy, time complexity and test error on different machine learning/ neural network models. We will present the comparative analysis of best model out of them to use for indoor localization technique in various closed environment. This project will be implemented on the third floor of California state university- Sacramento, Department of computer science after collecting the Wi-Fi fingerprinting data and will perform different test results. We have also found there has not been much research and comparative analysis happened in indoor localization field using the different machine learning algorithms in different scenarios like time duration, accuracy, error and overall performance. Implementation will be done in both the environments CPU and GPU using python programming language and different related libraries.