Understanding lane-changing behaviour via time-to-frequency conversion: a dynamic time warping approach based on continuous wavelet transform

Lane-changing stage division plays a pivotal role in guiding the driving decisions of connected vehicles (CVs). To overcome the limitations of existing methods in capturing local details, a dynamic lane-changing time warping (DLTW) approach is proposed, leveraging the continuous wavelet transform to extract lane-changing duration and segment lane-changing stages. Interactive lane-changing field tests were conducted on urban roads in Beijing to simulate mixed traffic flow. By considering both driver and vehicle elements, the DLTW method translates the lateral coordinates and acceleration of lane-changing vehicles into the frequency domain, enabling the identification of lane-changing duration and segmenting the process into three stages: preparation, action, and adjustment. Results indicate that the DLTW approach accurately captures key lane change moments for CVs in mixed traffic, achieving an average error of only 0.32s, reducing errors by more than 31.91% compared to previous methods. Additionally, the practical significance of wavelet energy has been explored.