A Practical Multi-sensor Localization and Mapping Approach for the Formula Student Driverless Competition

Formula Student is an international engineering competition where student teams design, build, and race small formula-style race cars. The competition provides a platform for students to apply their engineering knowledge in a practical, hands-on project and involves various disciplines such as mechanical, electrical, and computer engineering. One category of this competition is Driverless. This competition challenges student teams to design, build, and program autonomous vehicles that can navigate and compete in various dynamic and static events without a human driver. This thesis focuses on building a localization and mapping system for the Formula Student team from the University of Padova, RaceUp. The main goal is to transition the algorithms from a simulator environment to a real-world setup. The focus of this work is on vehicle sensorization, visual perception algorithms, and simultaneous localization and mapping (SLAM). The SLAM process is typically divided into two main parts: the front-end, which involves sensor data collection and feature extraction, and the back-end, which focuses on optimizing the vehicle's estimated trajectory and the map of its surroundings. The evaluation of the pipeline was conducted using real-world data obtained by equipping sensors onto the vehicle and traversing the track. By employing a blend of contemporary and traditional methodologies, we analyze data generated by the stereo camera to localize the colored cones outlining the track. The derived positions of the cones subsequently feed into critical modules of the vehicle, including the control system and the SLAM pipeline.