A New Lane-Changing Model with Consideration of Driving Style

摘要

The lane-changing model is a hot spot in the field of traffic research, and there are already a lot of free lane-changing modelestablished mathematical statistical methods or machine learning algorithm. However, these models don’t consider the driver’sdriving style to the free lane-changing, and the accuracy of these models is low. This paper considers the driver’s driving style andproposes a new free lane-changing model based on machine learning. The new model splits the sample data into three drivingstyles: cautious, stable and radical. This paper selects the most effective multilayer perceptron model by comparing differentmachine learning methods based on the NGSIM trajectory data. In the analysis of the final accuracy of this paper, it can be seenthat the new model has a great improvement in accuracy.