Autonomous Driving System Design for Formula Racing

Abstract

This paper mainly introduces automatic driving schemes and algorithms for formula racing car. Formula racing is a good platform for testing the extreme performance of vehicles, but there is a risk of harming the drivers. We propose an automatic driving scheme to avoid such problems. The solution we proposed integrates three modules: environment perception, path planning and vehicle control. Using lidar, camera and GPS/INS integrated navigation as sensors, an innovative data fusion method is proposed. This paper proposes a new path planning method, which uses Delaunay triangulation and topological map-based path search and evaluation to obtain the best path. Finally, a control model based on pure pursuit algorithm is used to control the vehicle precisely. The whole system is developed based on ROS which has a unique loose coupling. This scheme is proved to have high accuracy and good robustness by simulation analysis and multiple vehicle tests. The proposed solution has good portability, and it can also be applied to the field of unmanned passenger cars, robots and drones after appropriate modifications.