Modelling decisions of control transitions and target speed regulations in full-range Adaptive Cruise Control based on Risk Allostasis Theory

Adaptive Cruise Control (ACC) and automated vehicles can contribute to reduce traffic congestion and accidents. Recently, an on-road study has shown that drivers may prefer to deactivate full-range ACC when closing in on a slower leader and to overrule it by pressing the gas pedal a few seconds after the activation of the system. Notwithstanding the influence of these control transitions on driver behaviour, a theoretical framework explaining driver decisions to transfer control and to regulate the target speed in full-range ACC is currently missing.This research develops a modelling framework describing the underlying decision-making process of drivers with full-range ACC at an operational level, grounded on Risk Allostasis Theory (RAT). Based on this theory, a driver will choose to resume manual control or to regulate the ACC target speed if its perceived level of risk feeling and task difficulty falls outside the range considered acceptable to maintain the system active. The feeling of risk and task difficulty evaluation is formulated as a generalized ordered probit model with random thresholds, which vary between drivers and within drivers over time. The ACC system state choices are formulated as logit models and the ACC target speed regulations as regression models, in which correlations between system state choices and target speed regulations are captured explicitly. This continuous-discrete choice model framework is able to address interdependencies across drivers’ decisions in terms of causality, unobserved driver characteristics, and state dependency, and to capture inconsistencies in drivers’ decision making that might be caused by human factors.The model was estimated using a dataset collected in an on-road experiment with full-range ACC. The results reveal that driver decisions to resume manual control and to regulate the target speed in full-range ACC can be interpreted based on the RAT. The model can be used to forecast driver response to a driving assistance system that adapts its settings to prevent control transitions while guaranteeing safety and comfort. The model can also be implemented into a microscopic traffic flow simulation to evaluate the impact of ACC on traffic flow efficiency and safety accounting for control transitions and target speed regulations