Using a brain-like cognitive computational model to analyze the difference between desired speed and actual speed on rural highways for young drivers

Journal article, 2025

Due to the lower technical standards and complex driving environments on rural highways, drivers, especially young drivers, often encounter significant differences between their desired speed and actual speed (hereafter referred to as speed difference), resulting in frequent traffic accidents. Thus, this study proposes a brain-like cognitive computational model containing three subnetworks (i.e., perceptual, cognitive, and motor subnetworks) to analyze and predict young drivers’ speed difference through a “perception-cognition-action” loop. Simulated driving experiments were conducted on a 13.5-kilometer rural highway stretch with 50 young participants. Brainwave information (32-channel EEG) and actual speed were collected while driving, and the desired speed was obtained from participants’ self-reported data while they were watching their driving recording videos. A visual road environment model (VREM) was developed using deep neural networks to extract quantifiable parameters of the road environment perceived by drivers, which were then used as inputs for the perceptual subnetwork. In this study, the perceptual, cognitive, and motor subnetworks were composed of 4, 3, and 3 servers, respectively. Desired speed was the output of the cognitive subnetwork, while actual speed was obtained from the motor subnetwork. The brain-like cognitive computational model was calculated using the linear mixed-effects model that considers the driver heterogeneity. The results showed that using the brain-like cognitive computational model could predict the speed difference more accurately than using VREM alone. The findings could help to analyze speed difference causations and prevent risky driving behavior from an innovative brain-like perspective, thereby promoting the development of advanced driver assistance systems and human-like autonomous vehicles.