These days, autonomous robotics getting increasingly famous and widely used research area. One of the major users is car industry where autonomous vehicles are the focus of most developments. Autonomous functions, developed for vehicles, can be divided into two basic classes: DA (Driver Assistance) and AD (Automated Driving) functions. For both function classes monitoring the environment of the vehicle and building up a precise environmental- and vehicle-model is an indispensable requirement. Perception of the environment and vehicle parameters are realized by sensors. Data provided by the sensors is processed by digital processing units and calculation results are sent to the central processing unit. Based on the gathered information the central processing unit builds the environmental- and vehicle-model, which are used by other units responsible for higher level functionalities.
The purpose of this thesis is the development of a method which enables global localization based on data acquired from long range sensors. In this context, the gathering and analysis of possible matching methods and selection, implementation and evaluation of the one that is the most suitable for the current application.
Range measurements, serving as the input of the algorithm, are provided by a simulation framework developed especially for this work and a UTM-30LX laser range finder. During the literature survey three significantly different approaches have been chosen for further analysis, which are the following: iterative, spectral and feature-based matching methods. For the implementation of the simulation framework and the algorithms, and for the evaluation of results MATLAB 2016b software was used.
In this thesis, the introduction part is followed by the detailed description of the aforementioned matching methods. Thereafter, test results and performance evaluation of the implemented algorithms are shown. Finally, a framework for the global localization of the vehicle based on the chosen method is presented.