Abstract

Driver's mental workload (MWL) influences the driver's performance. A mental workload that is either too high or too low may endanger driving safety. Few previous studies have discussed how to quantify the mental workload in an objective way and whether or not it is possible to discover the driver's mental workload over-reduction. Furthermore, the influence of driver mental workload has not yet been considered in driver models. In this study, an ECG features-based driver mental workload estimation method was proposed. This measure is based on clustering analysis and Learning Vector Quantization neural networks. Furthermore, a fuzzy space model of the dangerous zone for a moving vehicle and the estimation method of driving risk level were proposed. Finally, two neural network based driver models with the input of driver mental workload were built up. The experiments and simulations show that the MWL estimation results are consistent with the evaluation of the Rating Scale of Mental Effort (RSME). The Driving risk level has potential to indicate the driver MWL over-reduction. In the driver steering performance simulation, Elman recurrent network based driver models show superiority than those based on multilayer perceptrons (MLP), and the MWL variance does have positive impacts on the performance of neural network based driver models. This study is limited to the available data source. Future work could be done on following aspects: more psychological signals indicating MWL variance should be included to make the quantification results more reliable. Besides, the influence of instant driving risk level to the driver's performance in the scenario of MWL over-reduction could also be studied.