A Computational Driver Model to Predict Driver Control at Unsignalised Intersections

Abstract

The interaction between a cyclist and a driver at unsignalized intersection remains a risky situation which may result in a collision with severe consequences, especially for the cyclist. Crash data show that the number of cyclist fatalities at unsignalized intersections has been stable the last years, indicating that more efforts should be given to improve safety in this specific scenario. Safety systems can help drivers avoid collisions with cyclists. However, systems addressing this conflict scenario are difficult to design, not only because of the technical aspects (e.g., sensor, or control limitations) but because those systems need to predict how drivers will or would control their car to be effective. A handful of studies focused on describing driver behaviour in this traffic scenario, but no computational model that can predict driver control can be found in the literature. The present study presents a driver model based on a biofidelic human sensorimotor control modelling framework predicting driver control in this traffic scenario. Two visual cues were implemented: 1) optical longitudinal looming, and 2) projected post-encroachment time between the bike and the car. The model was optimized using test-track data in which participants were asked to drive through an intersection where a cyclist would cross their travel path. The performances of the model were evaluated by comparing the simulated driver control process with the observed controls for each trial using a leave-one-out crossvalidation process. The results showed that the model performed rather well by reproducing similar braking controls, and kinematics, compared to the observations. The extent to which the model could be used by safety systems’ threat-assessment algorithms was discussed. Future research to improve the model performances was suggested.