Online Adaptation of Parameters using GRU-based Neural Network with BO for Accurate Driving Model

Testing self-driving cars in different areas requires surrounding cars with accordingly different driving styles such as aggressive or conservative styles. A method of numerically measuring and differentiating human driving styles to create a virtual driver with a certain driving style is in demand. However, most methods for measuring human driving styles require thresholds or labels to classify the driving styles, and some require additional questionnaires for drivers about their driving attitude. These limitations are not suitable for creating a large virtual testing environment. Driving models (DMs) simulate human driving styles. Calibrating a DM makes the simulated driving behavior closer to human-driving behavior, and enable the simulation of human-driving cars. Conventional DM-calibrating methods do not take into account that the parameters in a DM vary while driving. These "fixed" calibrating methods cannot reflect an actual interactive driving scenario. In this paper, we propose a DM-calibration method for measuring human driving styles to reproduce real car-following behavior more accurately. The method includes 1) an objective entropy weight method for measuring and clustering human driving styles, and 2) online adaption of DM parameters based on deep learning by combining Bayesian optimization (BO) and a gated recurrent unit neural network. We conducted experiments to evaluate the proposed method, and the results indicate that it can be easily used to measure human driver styles. The experiments also showed that we can calibrate a corresponding DM in a virtual testing environment with up to 26% more accuracy than with fixed calibration methods.