urichZurich, SwitzerlandAbstract Autonomous racing in robotics combines high-speed dynamics with the necessity for re-liability and real-time decision-making. While such racing pushes software and hardwareto their limits, many existing full-system solutions necessita show annotation
due to financial and safety con straints. This limits their reproducibility, makingadvancements and replication feasible mostly for well-resourced laboratories with compre-hensive expertise in mechanical, electrical, and robot ics fields. Researchers interest show annotation
nt time with familiarization and integration. The ForzaETH Race Stack addressesthis gap by providing an autonomous racing software platform designed for F1TENTH, a1:10 scaled Head-to-Head autonomous racing competition, which simplifies replication byusing commercial off-the-shelf hardware. This approach enhances the competitive aspect show annotation
thm as described in Section 8.1. The two trajectories different from the nominal racing line can be seen in Fig. 32. The cars were set up inthe same racing scenario as the official F1TENTH Head-to-Head format , where the vehicles are positio show annotation
trajectories.6.2 Global Planner The global planner is based on the work presented in [Heilmeier et al., 2020]. Their work describes theplanning of a minimum curvature trajectory using a quadratic optimization problem formulation. To optimizea global path around show annotation
ng trackboundaries from the map. The occupancy grid, generated by SLAM, can be interpreted as an image by transforming each cell intoa pixel as illustrated in Fig. 15a. In the first step, the occupancy grid is binarized. Subsequently, thebinarized image is smoothed with a morphological open filter [Bovik, 2009] to mitigate most of the LiDARscans located outside of the track. Following this, the centerline is extracted from the filtered image usingthe morphological skeleton method [Kong and Rosenfeld, 1996]. The centerline, characterized by its angularshape, can pose challenges for path optimization due to abrupt directional changes. Therefore, the centerlineis smoothed using a Savitzky-Golay filter [Orfanidis, 1995], and the resulting centerline is depicted in Fig. 15b. Finally, the centerline, in comb
ting real-time visual feedback. This state machine isresponsible for orchestrating different high-level behaviors through the information obtained from the differ-ent autonomy modules and supplying the Control module with the correct waypoints obtained either fromthe static global planner or the dynamically updated local planner. To balance and prioritize the a show annotation