Enhancing mixed traffic safety assessment: A novel safety metric combined with a comprehensive behavioral modeling framework

Abstract

In the context of future traffic systems, where automated vehicles (AVs) coexist with human-driven vehicles (HVs), ensuring road safety is of utmost importance. Existing safety assessment methods, however, are inadequate for the complex scenarios presented by mixed traffic conditions. These methods often fail to distinguish sufficiently between AVs and HVs, leading to inaccuracies in safety evaluations. To address these issues, this paper highlights the shortcomings of current surrogate safety measures (SSMs) in mixed traffic contexts and introduces a novel SSM, the Weighted Combination of Spacing and Speed Difference Rates (WS²DR). We propose a comparative analysis method to validate the effectiveness of WS²DR and to establish its safety threshold. Experiment results reveal that WS²DR outperforms traditional metrics such as time-to-collision and deceleration rate to avoid crashes, in terms of adaptability to both homogeneous and heterogeneous traffic environments and the detection of risk levels across a wider range of traffic conditions. Additionally, the paper presents a sophisticated mixed traffic modeling approach that accounts for different characteristics of AVs and HVs, incorporating factors such as errors of estimating the motion of other vehicles and the extended reaction time of HVs, as well as the perceptual and cooperative-active control capabilities of AVs. The results of the comparison analysis underscore the critical importance of considering the differences between AVs and HVs in modeling for accurate safety evaluations of mixed traffic. Simulation experiments confirm the positive impact on safety with increased AV penetration rates, emphasizing the necessity of employing refined modeling and safety assessment metrics to capture the full benefits of AV integration.