실제 도로상 자율주행자동차 평가 프로세스 도출 연구

Abstract

As the hardware and software that make up the automated driving system become more complex, consumers are increasingly concerned about the driving safety of automated vehicles in various fields such as failures of automated driving systems, risks to cybersecurity, and ethical issues of automated driving algorithms. Of course, efforts have been made to test the driving safety of automated vehicles in the simulation and proving ground. However, with current technology, there are clear limitations in implementing a perfect reality situation in simulation and proving ground. Therefore, it is time to focus on the evaluation process and evaluation indicators related to this safe evaluation of automated vehicles on the real roads. Against this background, this study attempted to define processes and evaluation indicators that can evaluate the driving safety of real automated vehicles on real roads. To this end, the evaluation process and evaluation indicators to be observed when evaluating automated vehicles on real roads are presented through related trends and research reviews. In addition, three prerequisites to be observed before performing the evaluation were presented. First, only vehicles that have been certified for temporary operation permission after sufficient testing on the proving ground are eligible. Second, a pre-safety inspection of the vehicle must be performed before driving on a real road. Third, a supervisor is required to ride and manage a real automated vehicle to test it, and a vehicle accompanying it is needed in case an accident occurs. Under the above prerequisites, this study proposes four steps for safe and effective evaluation of automated vehicles. The first step creates a matrix that can represent Korea's transportation environment. The second step is to select a test section suitable for the operational design domain of an automated vehicle. Third, specify the driving conditions only when possible or necessary. Finally, the evaluation is performed by driving a real automated vehicle along the set test section. In order to evaluate an automated vehicle on a real road, it is necessary to analyze the data collected while the automated vehicle is driving in real time and conduct several tests to be evaluated. At this time, the final evaluation report shall be prepared by comparing and analyzing the evaluation results of the simulation and proving ground. As a result of reviewing related studies to select indicators for evaluating real automated vehicles, a total of 66 evaluation indicators was selected. This was divided into 7 types according to the characteristics of each evaluation indicator and presented with a definition. It is believed that a basic design for the safe test of real automated vehicles can be derived through the real automated vehicle evaluation process and evaluation indicators derived from this study.