자율주행 시스템을 위한 우회전 구조화에 관한 연구

Abstract

Drivers have a lot of questions and confusion about how to turn right at the intersection. The recent revision and enforcement of the Road Traffic Act about pedestrian protection obligations have increased citizens' interest, but drivers are still having difficulty taking a legitimate right turn. The causes of this confusion are the lack of facilities operating RTOR and the unclear meaning of traffic signals, but efforts are still needed to clearly and consistently organize the legitimate right turn method. In particular, in the future, autonomous driving systems should learn the legitimate right-turn methods and verify their capabilities with computer systems such as simulations, so they should not only accurately interpret and organize Road Traffic Act but also convert them into algorithm-friendly states. Since the legitimate right-turn method depends on the presence or absence of signals for vehicles and pedestrians, signal lights, and pedestrian conditions, various provisions are involved, and drivers' perceptions and driving behaviors are different We need to convert the right-turn method into algorithm-friendly states. Therefore, this study attempted to organize the legitimate right-turn method based on precedents and design it in the form of a flowchart, which is one of the algorithmic expressions. First, the obligation to comply with the signal and the pedestrian protection in the first stage, the obligation to comply with the signal in the second stage, and the the obligation to comply with the pedestrian protection in the third stage were problematic, and the specific duty of care was confirmed and the legitimate right turn method was examined. Next, the intersection situation was classified into 27 types, and according to the above criteria, whether there was a signal violation, whether there was a temporary suspension obligation, and whether a right turn was possible were analyzed by situation. Finally, based on the results of the case review and contextual analysis, the legitimate right-turn method was converted to the form of a flowchart, and the accuracy of the flowchart was confirmed by showing that the results were the same as those of the contextual analysis discussed above. I hope that this study can be used in the development of a system for learning and verifying the legal compliance ability of an autonomous driving system as one of algorithm-friendly Road Traffic Act conversions.