According to the personification development ideas of automated driving system , the structure of automated driving system can be divided into five parts: perception, identification, decision, control and execution. The system first develop a perception of surroundings with the sensors, then process and integrate the perception data, get the whole picture of the environment, and then generate a decision based on its algorithm and finally pass the decision to the control system and execute corresponding orders.In this process, vehicles can exchange information with surroundings such as road facilities and other vehicles through vehicle to everything system (V2X). This helps vehicles to acquire a wider range of real-time information and solve questions like “Where I am”, “What are the surroundings”, “What changes may happen” and “What should I do”.
According to the statistics,there is on average 1-3 fatal accidents for every 100 million km human driving, self-driving have to significantly improve the safety level of driving to get massive applied. Meanwhile, wide range of road tests also helps to collect data of different scenarios and is a necessary method to improve artificial intelligence technology such as perception and decision making technologies. Visual road test under simulation environment and shadow mode, which are not involved in actual controls, can serve as the supplement for normal tests to reduce cost of road rests.