Abstract
Localization is a set of techniques that help in identifying the location of an object or a person, by taking into account different input parameters. This set of techniques is governed by algorithms ranging from traditional to advanced machine learning. These set of techniques are then incorporated into a full-fledged system in order to map out the location of the object, be it a person or even a car for example.
The concept of localization is generally divided into two parts – Outdoor localization and Indoor localization. The problem of outdoor localization has been long solved using the concept of the Global Positioning System (GPS). GPS helps in finding real-time location information of different mobile users in the outdoor environment. However, GPS is ineffective in indoor environments due to a couple of reasons, primarily reduced signal strength. The signal loss occurs due to the presence of obstacles. This limitation adversely affects implementing GPS in indoor environments.
In this project, the main goal is to solve the problem of indoor localization by using an algorithmic approach to map the location of the users. The general idea is to develop radio maps which also help in mapping unknown environments. The algorithms have been designed using the Kriging approach. A novel approach has also been followed where a neural network based on functional data analysis has been used.
Wi-Fi fingerprint data has been used for this project. The approach leverages the information provided by different access points (AP). For this, the dataset called UJIIndoorLoc has been used, which is publicly available on the UCI Machine Learning Repository. The information pertaining to Received Signal Strength Indication (RSSI) has been used to develop an approach using appropriate machine learning techniques, where location can be predicted and mapped out using radio maps. The entire code has been written in Python with the use of necessary libraries and algorithms.