Formula Student Driverless is a student competition where teams have to design and build an autonomous race car, that can navigate on an unknown track without a driver. This thesis presents the idea of using model-based navigation in autonomous racing for better performance. By creating a map of the environment, the vehicle is able to plan further ahead on the track and move much faster.
This thesis covers the main principles for solving Simultaneous Localization and Mapping (SLAM), such as the Extended Kalman Filter (EKF), the Particle Filter (PF), and the graph-based approach. Also presented is the fastSLAM algorithm, which was chosen to be implemented for the race car.
This thesis has resulted in the modeling of the vehicle and the sensors used for measurement. The SLAM algorithm was implemented in CPP language, in the Robot Operating System environment, which was then later tested in simulation and on real-world measurements. The results show that the fastSLAM algorithm is a suitable solution to the problem.