An integrated method based on wavelet modulus maxima and local Holder exponents for automatic phase detection and labelling of lane-changing execution

Abstract

This paper presents a novel automated method for detecting and labelling lane change (LC) execution phases using trajectory data. By integrating Wavelet modulus maxima lines with the local Holder exponents (WTMM-LHE), WTMM-LHE accurately identifies the commencement of LC execution. This methodology addresses the generalization challenges faced by traditional fixed-interval and rule-based approaches across different datasets. Furthermore, it improves upon the previous Wavelet transform modulus-based methods by effectively eliminating confounding results, thereby enhancing its robustness even with challenging trajectory profiles. Experiments on both synthetic and naturalistic trajectories were conducted to test this method’s performance. Results show that the proposed approach significantly enhances the reliability and accuracy of LC phase identification, improving data availability for calibrating, training, and modeling LC behaviors. Additionally, this study demonstrates the application of the proposed automatic labelling methods on machine learning-based LC prediction models, highlighting its ability to improve the accuracy of training data labelling, with potential implications for advanced driver assistance systems (ADAS) and connected and autonomous vehicles (CAVs).