Risk Assessment Method for Driving Scenarios of Autonomous Vehicles Based on Drivable Area

Risk assessment of driving scenarios is essential for both the research and development (R&D) and the verification and validation (V&V) of autonomous vehicles (AVs), which is closely related to driving safety. The method used for risk assessment should comprehensively reflect the impact on vehicle safety caused by various factors such as vehicles and road geometry in the driving scenario. Meanwhile, it should be capable of being directly applied in multiple scenarios. To this end, a novel risk assessment method, as well as a risk indicator named Discretized Normalized Drivable Area (DNDA), is proposed in this paper. First, the concept of drivable area is applied to obtain a comprehensive evaluation of the driving scenario. Then, considering the different importance of different parts of the drivable area to vehicle safety, the generated drivable area is discretely weighted. Finally, the ratio of the drivable area occupied by surrounding vehicles in the actual situation to the drivable area without surrounding vehicles in the ideal situation is used as the normalized output of the method. Moreover, the application of the method is then demonstrated via simulation experiments under three different scenarios. The experimental results show that the proposed method can reasonably estimate the risk of the driving scenario and has strong adaptability to multiple scenarios.