Identification of a Driver Steering Model, and Model Uncertainty, From Driving Simulator Data

Driver steering models have been extensively studied. However, driver model uncertainty has received relatively little attention. For active safety systems that function while the driver is still in the control loop, such uncertainty can affect overall system performance significantly. In this paper, an approach to obtain both the driver model and its uncertainty from driving simulator data is presented. The structured uncertainty is used to represent the driver’s time-varying behavior, and the unstructured uncertainty is used to account for unmodeled dynamics. The uncertainty models can be used to represent both the uncertainty within one driver and the uncertainty across multiple drivers. The results show that the unstructured uncertainty is significant, probably due to randomness in driver behavior. The structured uncertainty suggests that an estimation and adaptation scheme might be applicable for the design of controllers for active safety systems.