Abstract: The steering wheel motion and lateral position were used to evaluate the driver’s fatigue based on the driving experiments on driving simulator. The driver’s fatigue detection system was developed by neural network. Data of ten driver’s driving experiments are collected and divided into two levels of fatigue based on the PVT (psychomotor vigilance task) testing and record video. And then, the data are processed offline to find the difference between alert and fatigue. As the input data of neural network, the steering wheel angle and lateral position should be discretized and normalized. The neural network is trained by BP algorithm until the expected error is reached. The experiment results show that this method is effective and practical to identify the driver’s fatigue level.