Developing Driving Behaviour Models Incorporating the Effects of Stress

Driving is a complex task and several factors influence drivers’ decisions and performance including traffic conditions, attributes of vehicles, network and environmental characteristics, and last but not least characteristics of the drivers themselves. in an effort to better explain and represent driving behaviour, several driving behaviour models have been suggested over the years. In the existing literature, there are two main streams of driving behaviour models that can be found. The first is approaching driving behaviour from a human factors and cognitive perspective while the second is engineering-based. Driving behaviour models of the latter category are mathematical representations of drivers’ behaviour at the individual level, mostly focussing on acceleration/deceleration, lane-change and gap-acceptance decisions. Many of these factors are captured by existing driving behaviour models used in microscopic simulation tools. However, while the vast majority of existing models is approximating driving behaviour, primarily focusing on the effects of traffic conditions, little attention has been given to the impact of drivers’ characteristics.
The aim of the current thesis is to investigate the effects of stress on driving behaviour and quantify its impact using an econometric modelling framework. This main research question emerged as a result of a widely acknowledged research gap in existing engineering-based driving behaviour models related to the incorporation of human factors and drivers’ characteristics within the model specification. The research was based on data collected using the University of Leeds Driving Simulator. Two main scenarios were presented to participants, while they were also deliberately subjected to stress induced by time pressure and various scenarios. At the same time, stress levels were measured via physiological indicators. Sociodemographic and trait data was also collected in the form of surveys.
The data has been initially analysed for each main scenario and several statistics are extracted. The results show a clear effect of time pressure in favour of speeding, however relations related to physiological responses are not always clear. Moreover, two driving behaviour models are developed, a gap-acceptance and a car-following model. In the former model, increase in physiological responses is related to higher probability of accepting a gap and time pressure has a positive effect of gap-acceptance probability as well. In the car-following model, stress is associated with increased acceleration and potentially a more aggressive driving style.
The aforementioned analysis is based on data collected in a driving simulator. Given the potential differences in driving behaviour between real and simulated driving, the transferability of a model based on the latter data to field traffic setting is also investigated. Results indicate significant differences in parameters estimated from a video and the simulator dataset, however these differences can be significantly reduced after applying parameter updating techniques.
The findings in this thesis show that stress and drivers’ characteristics can influence driving behaviour and thus should be considered in the driving behaviour models for microscopic simulation applications. However, for real life applications, it is suggested that the extent of these effects should be treated with caution and ideally rescaled based on real traffic observations.