Discrepancies between Simulated and Real-World Car-Following Behavior of Adaptive Cruise Control Vehicles in Mixed Traffic

Abstract

In recent years, most newly manufactured vehicles in the United States have been equipped with adaptive cruise control (ACC). However, publicly accessible datasets that (i) confirm ACC engagement and (ii) provide detailed car-following trajectories in mixed traffic with human drivers remain limited. As a result, many traffic studies rely on microscopic simulation to examine interactions between ACC-equipped and human-driven vehicles. The reliability of these simulations depends on the predictive accuracy of the underlying car-following models. Therefore, this study assesses the discrepancy between real-world ACC car-following behavior and simulated ACC car-following behavior in relation to speed, space-headway, response-time, macroscopic traffic-flow parameters, and flow-stability. The study uses the OpenACC dataset for real-world ACC car-following, while relying on two prevalent car-following models: the PATH-ACC model developed by PATH Lab at UC-Berkeley and the intelligent driver model. Although we expected the calibrated car-following models to mirror observed ACC behavior, the findings reveal substantial differences. Both calibrated models failed to predict ACC car-following without a significant difference. Among the two, the PATH-ACC model demonstrates stronger predictive capability, particularly for speed and space headway. Simulated ACCs exhibit shorter response times (0.4–0.6 s) than observed ACC vehicles (0.8–1.1 s), therefore, reflecting idealized reaction assumptions and ignoring real-world heterogeneity. Flow stability analysis revealed more stable simulated ACC trajectories (relative mean absolute deviation [RMAD] ∼2.5–2.9) than those observed (RMAD ∼2.8–3.6). Both models show higher critical density and flow values than real-world ACC vehicles. These results underscore the need for advanced, data-driven modeling approaches to improve the behavioral realism of ACC simulations in mixed traffic.