Robust Lane Detection with Binary Integer Optimization

Abstract

Formula Student Driverless (FSD) is a competition where student teams compete to build an autonomous racecar. The main dynamic event in FSD is trackdrive, where the racecar traverses an unknown track whose boundaries are demarcated by cones. One challenge of the event is to determine the track boundaries based on cone locations in the presence of false positive cone detections, sharp turns and uncertain cone color information while traversing the track. In this work, we present a binary integer optimization that encapsulates this problem, along with taking into account competition rule specifications, such as cone spacing and track width. This optimization routine is implemented in simulation, and on an autonomous electric racecar. We present our approach, and analyze its latency, accuracy, and robustness to uncertain cone detections. This approach is used on-vehicle to solve the real-time boundary generation problem during the competition.