Adaptive Cruise Control (ACC) system has attracted great attention in recent years, since it is recognized to have the potential to reduce traffic crashes. However, due to the limitation of sensing and automation technologies, the current ACC system is demonstrated by some studies to be inferior to human drivers in longitudinal safety performance. This study revisits this issue from a different perspective, the heterogeneity of human drivers versus the standardization of ACC system. Although the safety performance of one single ACC vehicle may not be superior to one human driver, the vehicle string equipped with a standardized ACC system may outperform the string consisting of heterogeneous human drivers. Specifically, a micro-level simulation platform is first developed in MATLAB software, and models for ACC system and human drivers are coded and incorporated into the platform. Surrogate safety measures, derived from time to collision (TTC), are utilized to evaluate longitudinal safety in simulations. The empirical trajectory data is employed to calibrate model parameters of human drivers to delineate heterogeneity. Extensive simulations are conducted to test parameter settings for ACC system and for sensitivity analyses. Results indicate that: (1) the heterogeneity of human drivers in practice negatively affects longitudinal safety; (2) the standardized design of ACC system could outperform human drivers in some conditions; and (3) the ACC model is greatly affected by the different parameters setting. Results of the study could provide valuable information for safety improvement of mixed traffic flow in the near future. Key words: adaptive cruise control, longitudinal safety, intelligent driver model, simulation, time to collision

CICTP