Autonomous vehicle’s structure consists of four parts: The Perception, localization, motion planning and motion control. For implementing navigation, data obtained from various sensors are collected and then required information is extracted from them. Based on this information, vehicle location, vehicle status to its surrounding environment and position of all identified static and dynamic obstacles around vehicle can be achieved. Then, set of actions or paths that would guide vehicle to accomplish its mission are designed in motion planning segment and given to motion control segment for executing. It can be said that the most important factor in navigation is localization. In an urban environment, a map that contains all roads and intersections information would be a great help for localization process. This information may be provided in the form of a graph. In this graph, links and nodes are respectively represented as roads and intersections of the real road network map. With entering road network information into localization domain, a new concept called Map Matching is introduced. The main purpose of map matching is finding the link of the graph that vehicle is located on. This link is called chosen link. A map matching algorithm usually behave as follow, at the first step, all links which has the possibility of being the chosen link are selected. These links are called candidate links. Then the chosen link is selected among these candidate links. This means vehicle is on this link. And at the end, vehicle location is estimated on the chosen link. GPS data has an important role in map matching algorithms. In this thesis, some of the most commonly used map matching algorithms that follow the mentioned behavior are presented and examined. As an example of more advanced map matching algorithms, particle filter based map matching algorithms are presented. Disadvantages of these algorithms and proposals to their improvements are expressed. All algorithms are evaluated on real dataset and are compared based on two parameters: correct estimation on map matching and execution time. The mentioned autonomous structure is modeled with this difference that it has been assumed that there is no dynamic obstacle in the environment; therefore, the motion planner is only executed at the first step. Dijkstra algorithm is used for motion planner. The proposed map matching method that gives the best performance is selected and used in localization segment of autonomous navigation architecture. Simulated vehicle has similar dynamic constrains of a real car and it can be guided just by controlling speed and steering angle in motion control segment. All map matching evaluation and autonomous experiments are executed and simulated in MATLAB software environment. Autonomous experiments are performed for different GPS errors and in all of them, the vehicle is accomplished its mission and reached the destination point successfully. Presented architecture has the ability to be executed on every possible map. Keywords: autonomous navigation, localization, urban environment, digital map, map matching.