Exploitation of String Stability to Predict Disturbance-triggered Platoon Collisions in Mixed Traffic Comprising Automated and Conventional Vehicles

Autonomous driving has drawn much attention and been widely studied in recent years, given its remarkable superiority in perception and control compared with human driving. However, it is foreseeable that the full deployment of automated vehicles (AVs) will take a long time, due to challenges associated with technology, ethics, and policy. Consequently, mixed traffic streams comprising both AVs and human-driven vehicles (HDVs) will exist in transportation systems. This paper devises a novel approach to estimating the platoon collision probability after a certain disturbance in such a mixed traffic environment. First, the disturbance transfer models for AVs and HDVs are derived in the frequency domain. Next, the platoon expectation stability (PES) and a simple PES safety (PESS) model are developed. Finally, based on complex vehicle combinations and random AV appearance, an improved PESS model is constructed. Comprehensive simulation experiments reveal that the simple PESS model achieves a reasonable performance, with an overall absolute percentage error of 20.41 % across various AV deployment levels, disturbance amplitudes, and platoon lengths. The improved PESS model further reduces the error to 3.95%, thereby accurately estimating the platoon collision probability and demonstrating its great potential for use in collision warning and avoidance.