Hou, Qingqing (Nanyang Technological University), Lu, Yun (Nanyang Technological University), Su, Rong (Nanyang Technological University), Jia, Chengfeng (Nanyang Technological University)

A Deep Auto-Encoder Approach for Clustering Cut-In Driving Styles from Naturalistic Data

Scheduled for presentation during the Video Session "On-Demand Video Presentations" (VP-VP), Saturday, November 22, 2025, 08:00−18:00, On-Demand Platform

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on April 2, 2026

Abstract

Vehicle cut-ins are a critical type of lane change that can introduce collision risks. The significant differences between the cut-ins and other lane changes suggest the need for dedicated research on cut-ins. Identifying and understanding the cut-in driving styles can help better manage such high-risk maneuvers. This study investigates the clustering of the cut-in driving styles by using naturalistic data. Particularly, cut-in events were extracted from the NGSIM dataset. We select driving features of the cut-in vehicle (CIV) and its interaction with surrounding traffic during the preparation and execution stages of the cut-in maneuver. A deep auto-encoder is built to reduce the dimensionality of the selected features. The K-means++ algorithm is applied to cluster the cut-in driving styles. The results show that the proposed deep auto-encoder outperforms traditional methods, including t-distributed stochastic neighbor embedding (t-SNE) and principal component analysis (PCA), achieving optimal clustering performance when the cluster number is three. Accordingly, the cut-in driving styles are divided into three classes, i.e., cautious, normal, and aggressive, and their driving features are further examined through statistical analysis.