Thesis Details
An Autonomous Driver of a TORCS Racing Car
This work describes the TORCS simulator and optimization algorithms used in the field of autonomous driving competitions. The main purpose of this work is to design a new controller solution based on genetic algorithms. The controller's behavior can be divided into two main parts which are exploited during the distinct stages of the competition. The warm-up stage serves for the track model sampling and the race line optimization. The race stage logic then benefits from the data obtained in the warm-up stage. The track optimization is done by a Genetic algorithm while the track is divided into several segments optimized separately. A genetic algorithm is applied once again to the track trajectory to smooth out gaps caused by the segment composition. In this work was shown that the track sampling and race line optimization by a genetic algorithm can be done during the warm-up stage. This makes the controller suitable for an autonomous driver competitions.
TORCS, car racing, autonomous driver, car controller, algorithms inspired by biology, genetic algorithm, racing line, the simulated car racing championship
Hrubý Martin, Ing., Ph.D. (DITS FIT BUT), člen
Janoušek Jan, doc. Ing., Ph.D. (FIT CTU), člen
Rychlý Marek, RNDr., Ph.D. (DIFS FIT BUT), člen
Sekanina Lukáš, prof. Ing., Ph.D. (DCSY FIT BUT), člen
Zbořil František, doc. Ing., Ph.D. (DITS FIT BUT), člen
@mastersthesis{FITMT11855, author = "Luk\'{a}\v{s} B\v{e}hal", type = "Master's thesis", title = "An Autonomous Driver of a TORCS Racing Car", school = "Brno University of Technology, Faculty of Information Technology", year = 2012, location = "Brno, CZ", language = "english", url = "https://www.fit.vut.cz/study/thesis/11855/" }