As a promising strategic industry group that is rapidly evolving around the world, autonomous vehicle is entering a critical phase of commercialization from demonstration to end markets. The global automotive industry and governments are facing new common topics and challenges brought by autonomous vehicle, such as how to test, assess, and administrate the autonomous vehicle to ensure their safe running in real traffic situations and proper interactions with other road users. Starting from the facts that the way to autonomous driving is the process of a robot or a machine taking over driving tasks from a human. This paper summarizes the main characteristics of autonomous vehicle which are different from traditional one, then demonstrates the limitations of the existing certification mechanism and related testing methods when applied to autonomous vehicle. Based on the above analysis, a novel assessment mechanism focusing on complete vehicle behaviors is proposed, which takes safety, timeliness, accuracy, and smoothness as the four rules of autonomous vehicles. Then, a novel comprehensive evaluation scheme based on scenario is proposed, which combines simulation test, real vehicle test and verification test under actual driving situation. In addition, a novel administration mechanism covering vehicle development, type approval and in-use management is proposed to address the fact that the performance of autonomous vehicles is dominated by software rather than hardware and is changed by software updates throughout the life of the vehicle.
