Research on Risk Assessment for the Operational Design Domain of Autonomous Vehicles Based on the Driving Risk Field

Background: Real-time driving risk assessment is an important basis to ensure the safe operation of automated driving systems. There are many patents and articles related to risk assessment methods, however, they are not applicable to the situation when the automated driving system approaches the boundaries of its Operational Design Domain (ODD), because of the abrupt increase in driving risk. Objective: This paper proposes a risk assessment method for autonomous vehicle ODD based on driving risk field Methods: First, the comprehensive driving risk is calculated by quantifying the risks of different elements within ODD based on a driving risk field model, where Chinese traffic accident data calibrate the environmental risk parameters. AEB scenario testing is conducted to compare the driving risk index with the reciprocal of the time to collision (TTC). Furthermore, to address the abrupt change of driving risks near the ODD boundary, the Critical Operating Design Domain (CODD) is proposed in this paper, which quantitatively analyzes the risks in different driving scenarios and the critical risk thresholds are determined. Finally, the driver takeover experiment is conducted through the joint simulation of PreScan and Simulink. Results: Tests conducted in the AEB scenario show that the driving risk index can accurately assess driving risks when considering environmental factors. The driver takeover experiment results show that the risk classification based on the risk assessment coefficient can accurately identify dangerous scenarios, and the CODD risk threshold can effectively assist the driver in taking over the vehicle in time. Conclusion: The proposed driving risk assessment approach based driving risk field addresses the issues of existing risk assessment approaches that have not adequately considered environmental factors and the abrupt changes in risks near ODD boundaries in autonomous driving technology and could identify the risks in ODD of autonomous vehicles under different hazardous scenarios, the simulation results verify its effectiveness. The proposed risk assessment method could effectively assist the driver to takeover the vehicle in time, thereby improving the safety of autonomous vehicles.