Abstract: Safety of the intended functionality(SOTIF) is one of the biggest challenges in the commercialization of intelligent connected vehicles. In the process of conducting SOTIF analysis on intelligent connected vehicles, establishing appropriate risk acceptance criteria can provide more accurate evaluation criteria for hazard identification and risk assessment, and help reduce development costs, improve the SOTIF confidence and development efficiency of the entire vehicle. In order to establish appropriate risk acceptance criteria, some key parameters of driver traffic characteristics in natural driving data are extracted to validate the “extreme value characteristics” of driver behavior and the normal distribution of parameters. Three basic principles that need to be followed in establishing safety acceptance criteria are analyzed. Based on the practical application of the six standard deviation (6sigma, 6σ) theory in the field of engineering, it is proposed to use the 6σ theory to establish the 3σ acceptance criteria for SOTIF risk assessment. Based on this acceptance criteria, the calculation method of the standard value of the driving safety index(DSI) in the driving safety field(DSF) is redefined. Finally, the TrafficNet database is used to calculate the standard value of DSI(DSI^*) under different basic scenarios, quantifying the SOTIF risk acceptance criteria under the basic scenarios. The research results have improved the SOTIF evaluation system for intelligent connected vehicles. They are of significance for improving the safety level of intelligent connected vehicles fundamentally and avoiding endless accumulation of testing mileage.

Key words: intelligent connected vehicles, safety of the intended functionality, risk assessment, driving safety index, acceptance criteria