Disengagement causes analysis of automated driving system

The failure and disengagement cause analysis of automated driving system is significant for the evaluation of automatic driving ability and system defect mining. California DMV has conducted data collection work on the disengagement of automated driving system in several companies. This paper comprises two parts: the disengagement causes analysis based on automatic driving system technology links and data cleaning, classification and reconstruction of DMV which involves scenarios, causes and enterprise factors. The analysis shows that planning and decision system is the most prone link and street is the most instability scenario. Furthermore, the automated driving capabilities of the enterprises is polarized based on DMV's data, and the results provide reference for the test and design of the automatic driving vehicle.