Understanding Human Driving Behavior through Computational Cognitive Modeling

Abstract

As per an article in The Economist, someone, somewhere, dies in a road crash every 30 seconds, and about 10 people are seriously injured. Currently, there are about 1.3 million global deaths per year due to road accidents. Most of these deaths and injuries are caused by either factors that are internal to the driver (e.g., driving experience), or due to factors that are external to the driver (e.g., track complexity). However, currently little is known on how these factors influence human driving behavior. In this research, we investigate the role of an external factor (track complexity) on human driving behavior through computational cognitive modeling. Eighteen human participants were asked to drive on two tracks of the same length: simple (4 curves; N=9) and complex (20 curves; N=9). Later, we used two computational models to fit the human steering control data: an existing near-far-point model and a new heuristic model involving tangent and car-axis angles and a position-correction term. Our modeling results show that the fit of the heuristic model to human data on the simple and complex tracks was superior compared to that by the near-far-point model. We highlight the implications of our model results on human driving behavior.