Robustness analysis of car-following models for full speed range ACC systems

TL;DR

This paper evaluates the robustness of various car-following models used in full speed range adaptive cruise control systems, analyzing their stability under real-world disturbances like noise, latency, and heterogeneity.

Contribution

It provides a systematic comparison of classical car-following models, highlighting their stability robustness and the factors that can induce stop-and-go waves.

Findings

Many models are sensitive to latency, noise, and measurement errors.

Even stable models can be destabilized by external perturbations.

Heterogeneity and kinetic constraints significantly affect system stability.

Abstract

Adaptive cruise control systems are fundamental components of the automation of the driving. At upper control level, ACC systems are based on car-following models determining the acceleration rate of a vehicle according to the distance gap to the predecessor and the speed difference. The pursuit strategy consists in keeping a constant time gap with the predecessor, as recommended by industrial norms for ACC systems. The systematic active safety of the systems is tackled thanks to local and string stability analysis. Several classical constant time gap linear and non-linear car-following models are compared. We critically evaluate the stability robustness of the models against latency, noise and measurement error, heterogeneity, or kinetic constraints operating in the dynamics at lower control level. The results highlight that many factors can perturb the stability and induce the formation of stop-and-go waves, even for intrinsically stable car-following models.