The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

Accurately estimating driving style is crucial for designing personalized autonomous driving to enhance market acceptance. This paper focuses on the task of driving style estimation while driving. Unsupervised learning Deep Neural Networks (DNNs) is employed to model the driving style estimation process. A novel model defined as deep clustering is proposed, in which the data space is reconsidered and a parameterized non-linear embedding from the original data space to a low-dimensional feature space by using DNNs is employed. The naturalistic driving data collected from the Next Generation Simulation (NGSIM) dataset is used for framework development and verification, and four-dimensional driving style is obtained with reasonable performance. Compared with k-means, Fuzzy c-means (FCM) and Gaussian Mixture Model (GMM), results showed that the deep clustering model can be applied to estimate driving style reliably and superior to traditional methods in behavior analysis. Moreover, deep clustering has a stable effect on driving style estimation for different vehicle classes, showing the universality and effectiveness.

Deep Clustering;Driving Style Estimation;NGSIM