A Bayesian Reference Model for Visual Time-Sharing Behaviour in Manual and Automated Naturalistic Driving

Abstract

Visual time-sharing (VTS) behavior characterizes an inattentive driver. Because inattention has been identified as the major contributing factor in traffic crashes, understanding the relationship between VTS and crash risk could help reduce the crash risk through the development of inattention countermeasures. The aims of this paper are: 1) to develop a reference model of the VTS behavior and 2) reveal if vehicle automation influences the VTS behavior. The reference model was based on naturalistic eye-tracking data. The VTS sequences were extracted from routine driving data (including manual and automated driving). We used Bayesian generalized linear mixed models for a range of on- and off-path glance-based metrics. Each parameter was estimated with a probability distribution and summarized with credible intervals containing the model parameters with 95% probability. The reference model not only corroborates previous findings from the driving simulator experiments and on-road studies, but also captures the characteristics of on-path and off-path glance behavior in greater detail. The model demonstrated that: 1) there was minimal change in the VTS behavior due to automation and 2) the percentage of time that glances fell on-path (PRC) was greater for all routine driving (~80%) than for the VTS sequences (~50%). The PRC was the only metric that was sensitive to the VTS, but it did not differentiate between manual and automated driving. Our model, by describing a measure of inattention (VTS behavior), can be used in future driver models to improve the computer simulations used to design ADASs and evaluate their safety benefits. In addition, the model could serve as a detailed reference for inattention guidelines.