Real-Time Localization In Large-Scale Underground Environments Using RFID-Based Node Maps

This thesis presents a system for localizing a sensor equipped mining vehicle in a large-scale underground environment, where GPS is not available. Such a system would increase vehicle drivers' situational awareness and enable underground mining companies to monitor their vehicles and manage operations remotely, all of which would increase efficiency and safety. Previous work in this area has been successful in mapping large-scale environments using RFID tags as unique landmarks.;The localization system was first tested offline using simulated and previously collected real underground mine data, and online in the 4-kilometre long Carleton University underground tunnels. Experimental results from various localization tests as well as qualitative and quantitative analyses are presented. A GPS-like GUI developed for the mining vehicle operator as well as a website interface for monitoring mining vehicles locations remotely are also shown.;The localization system presented in this work incorporates the use of RFID tags, a particle filter and a set of 2D local maps built a priori , referred to as node maps, that represent the environment. The overlapping structure of node maps allows for efficient localization from the point of view of the processing power and the memory required. The use of sporadically-placed passive RFID tags as unique landmarks allows for the creation of the locally consistent node maps and for efficiently solving the global localization problem in very large, challenging, unstructured, and uniform appearance environments.