Gulliver - Vidareutveckling av ett system för testning av autonoma bilar

Abstract

With an increasing population and new transportation habits, many problems regarding the traffic system will become more prominent. These include road congestion, traffic accidents and emission of carbon dioxide. Many of these problems could be reduced by letting computers drive for us. Developing driverless cars is however a very expensive process and there are many difficulties. At Chalmers there is a project named Gulliver, where a simulated computer model is combined with small physical cars to achieve a good combination of realistic and cost effective tests. This bachelor thesis further develops the Gulliver project, and the goal is to improve and develop the test-bed as well as to evaluate new technologies that can be relevant in the future of the project. The thesis has focused on three subprojects: • Improving the localization system of the cars by integrating the software GulliView into the Gulliver graphical tool. GulliView enables the use of an independent source in the form of a web camera that uses image recognition to localize the cars. • Improving the graphical tool that is used when testing in order to speed up and simplify the testing procedure. • Evaluating the algorithms UDP Broadcast and Paxos in terms of performance, especially the time it takes to reach consensus, for direct communication between the cars.