Dynamic driving environment complexity quantification method and its verification

To meet the requirements of scenario-based testing for Autonomous Vehicles (AVs), driving scenario characterization has become a critical issue. Existing studies have concluded that complexity is a necessary criticality measure for supporting critical AV testing scenario identification. However, the existing scenario complexity quantification studies mainly have two limitations, namely, subjective quantification methods highly rely on human participations and are difficult to apply to big data, and existing objective methods lack consideration of human driver characteristics and hence the performance cannot be guaranteed. To bridge these research gaps, a general objective quantification framework is proposed to quantify human drivers' judgement on driving environment complexity, by describing the vehicle-vehicle spatial-temporal interactions from the perspectives of quantity, variety, and relations. The model mainly contains three parts. First, to describe quantity information, a dynamic influencing area was set to identify the surrounding vehicles that contribute to driving environment complexity based on ResponsibilitySensitive Safety (RSS) theory. Second, considering the various surrounding vehicles' driving statuses, behaviors, and intentions, a basic vehicle-pair complexity quantification model was constructed based on encounter angles. Then, nonlinear relationships based upon information entropy theory were introduced to capture the heterogeneous longitudinal and lateral complexities. Third, a vehicle-pair complexity aggregation and smoothing step was conducted to reflect the characteristics of human driver's cognition. To demonstrate the abovementioned model, empirical Field Operational Test (FOT) data from Shanghai urban roadways were used to conduct case studies, and it can be concluded that this model can accurately describe the timing and extent of the complexity change, and reveal the complexity differences due to scenario type and spatial-temporal heterogeneity. Besides, Inter-Rater Reliability (IRR) index was calculated to validate the consistency of scenario complexity judgement between the proposed model and human drivers, and for performance comparison with the existing models. Finally, the applications of the proposed model and its further investigations have been discussed.