Detecting Cognitive Workload Using Driving Performance and Eye Movement in a Driving Simulator

The aim of this study was to develop cognitive workload estimation algorithm using driving performance and eye movement data. The algorithm adopts radial basis probabilistic neural networks (RBPNN) to construct cognitive load estimation models. In order to train and test the models, recordings of driver’s gaze and driving performance were captured in a driving simulator during three levels of cognitive demand. As a result, it was found that the proposed RBPNN models were able to differentiate driver’s high cognitive workload from the normal driving with high accuracy. The best performance was achieved with a combination of standard deviation of lane position (SDLP) and gaze dispersion of X and Y coordinates over 30 seconds time window. The highest cognitive workload detection accuracy rate in overall model performance was 85.0%.