Exploring Driving Style Variations When Driving in Work Zone: A Driving Simulation Study

Abstract: Traffic safety hinges on individual driving styles, which can vary within a single trip through a work zone. A work zone, with its limited visibility, heavy machinery, and unexpected traffic flow, is a major contributor to a high number of traffic accidents. Driving style has a significant effect on driving behavior and directly impacts driving safety. However, studies on the variation in driving styles in the specific scenario of driving through a work zone are still missing. Also, most studies used either surveys or machine learning methods for classifying driving styles, while there is a lack of comparison between these two classification methods. To address the gaps, this study aims to detect and classify variations in driving styles when driving through a work zone and compare the results obtained from the self-evaluation method with those from the machine learning method. Firstly, a lane closure work zone was simulated by Webots and SUMO, based on a real-world case of an urban road section in Indiana state. Daytime and nighttime scenarios were included to analyze variations in driving styles. Secondly, sixteen participants were invited to drive through the road section with a lane closure work zone using a driving simulator. Their driving speed, as well as acceleration and deceleration, were collected by Webots. Then, the K-means algorithm was used to classify three types of driving styles (aggressive, normal, or calm) based on the total non-linear traits in the driving data (eg, speed, acceleration). Finally, a self-evaluation survey on driving styles was conducted after the driving simulation experiment, and a comparative analysis was performed between the self-evaluation survey data and the driving simulation data. The results show that 1) The percentage of aggressive driving style was 51.6%. Participants tended towards a calm driving style at nighttime compared to daytime, and the normal driving style remained consistent across both daytime and nighttime; 2) There were significant differences between self-evaluation and K-means method on driving styles. Compared to the results from the K-means method, drivers tended to overestimate their normal driving style and underestimate their aggressive driving style based on the results from the self-evaluation method.; and 3) There was an observed increase in calm driving style before and during the work zone, contrasting with a rise in aggressive driving tendencies after exiting. The results may help understand the variance of driving styles in a work zone and improve the classification accuracy of driving styles by comparing the differences between driving data evaluation and post-survey data evaluation.

Keywords: Driving style classification, Driving style variations, Work zone, ANN model

DOI: 10.54941/ahfe1005775

Cite this paper:

​