Seite - 51 - in Proceedings of the OAGM&ARW Joint Workshop - Vision, Automation and Robotics

Design of an Autonomous Race Car for the Formula Student Driverless (FSD) Marcel Zeilinger1, Raphael Hauk1, Markus Bader2 and Alexander Hofmann3 Abstract—Formula Student Germany is a race car com- petition for student teams competing with self-designed, self- developed and self-built vehicles. These cars have been com- peting to win in various disciplines every year since 2006 at the Hockenheimring in Germany. In 2016, a new discipline, Formula Student Driverless, was announced for the following year, targeting autonomous race cars that complete tracks of 5kmover 10 laps as fast as possiblewithout thehelp of human racer pilots or remote control systems. This paper will cover the framework, the sensor setup and the approaches used by the Viennese racing teamTUWRacing. I. INTRODUCTION Formula Student Germany (FSG) is an international con- struction competition for students with the aim of promoting research in multiple disciplines on a concrete application. The competition is divided into three classes: electric, combustion and driverless vehicle. The last classification, which has existed since 2017, is called Formula Student Driverless (FSD), where either combustion or electric cars arepermitted.Contrary to theother twoclasses, thedriverless class rules permit the use of a team’s existing race car which has already attended previous FSG race(s). Therefore, formerly manually-driven race cars equipped with appropri- ate sensors, actors and computing hardware can be adapted to driverless cars. Nevertheless, vehicles have to comply with the restrictions of their appropriate class; in particular, manual operation must still be possible. Regardless of the class in which a team participates, the event itself is divided into static and dynamic disciplines, and points can be scored in each discipline. In the static disci- pline, a cost report and a business plan must be presented by each team.Beforeacar is allowed toparticipate inadynamic event, a technical inspection must be passed to ensure the car is mechanically and electrically safe, in accordance with the rules [3]. Since a car’s total score is comprised of both disciplines, the fastest car need not necessarily win. In the following sections TUW Racing will describe its concept for participation in the FSD 2017. First, an overview is provided of the solutions and academic approaches avail- able in the area of autonomous driving, with regard to similar 1Marcel Zeilinger and Raphael Hauk are with TUW- Racing, Adolf-Blamauergasse 1-3, 1030 Wien, Austria firstname.lastname@racing.tuwien.ac.at 2Markus Bader is with the Automation Systems Group at Institute of ComputerAidedAutomation,TUWien,Karlsplatz13,1040Vienna,Austria firstname.lastname@tuwien.ac.at 3Alexander Hofmann is with the Institute of Com- puter Science at the University of Applied Sciences FH Technikum Wien, Ho¨chsta¨dtplatz 6, 1200 Vienna, Austria firstname.lastname@technikum-wien.at Fig. 1. Racecar Edge8 from the previous season (2016) which has been redesigned to drive autonomously solutions in general mobile robotics. Then an overview of the vehicle and the sensors, actuators, as well as the software components used is provided. Furthermore, we will describe our approach to the problems posed by the competition and reflect upon its effectiveness up to this point. II. RELATED WORK In order to describe related work, time was invested in studying competitions with a similar focus. In addition, stud- ies were conducted on commercially available solutions to autonomous driving such as the NVIDIA DRIVE1 products. The Carolo Cup2 is a competition where student teams build autonomous 1:10 sized cars, which have to cope with reality-inspired driving situations faced by passenger cars [12]. While there is a static competition in the form of a concept presentation as well, the dynamic events consist of parking and driving a free drive with and without obstacles. During the free drive, the model car needs to master chal- lenges like intersections, speed limits or dynamic obstacles suchasothervehiclesorpassengerscrossing the track,which is marked with white lines on a dark ground similar to real road surface markings. In the paper [12], Technical Evalua- tion of the Carolo-Cup 2014, the authors state under Lessons Learned the importance of high quality hardware products and the need for a robust computer vision algorithm in detecting road markings. Similar to this competition are the NXP Cup student competition, also known as the Freescale 1NVIDIA DRIVE http://www.nvidia.com/object/drive-px.html 2Carolo Cup: https://wiki.ifr.ing.tu-bs.de/carolocup/ 51