A Naturalistic Driving Study for Lane Change Detection and Personalization

Lakhkar, Radhika Anandrao; Talty, Tim · 2024 · Crossref

DOI: 10.4271/2024-01-2568

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This thesis addresses the challenge of personalizing autonomous driving features, specifically automated lane changes, to align with individual driver behaviors. As Advanced Driver Assistance Systems (ADAS) and autonomous technologies become ubiquitous, ensuring user acceptance requires that vehicle maneuvers match the driver’s preferred style, whether conservative or aggressive. The research aims to characterize naturalistic lane-changing behaviors using real-world data to develop methods for detecting these events and classifying driving styles, thereby facilitating the personalization of autonomous control systems. The study utilizes the Second Strategic Highway Research Program (SHRP2) Naturalistic Driving Study (NDS) dataset, which contains over 5.4 million trip files from more than 3,400 participants. To extract reliable lane change instances from this massive dataset, the author developed a detection algorithm using machine vision variables, specifically left and right lane marker probabilities, to filter out noisy or erroneous data. Lane changes were identified by analyzing discontinuities in the lane offset variable, which represents the vehicle’s distance from the lane center. Once extracted, kinematic parameters—including vehicle speed, lateral displacement, lateral acceleration, steering wheel angle, and lane change duration—were analyzed to characterize the maneuvers. The findings demonstrate that analyzing kinematic parameters individually is insufficient for capturing the complexity of driving styles. Consequently, the thesis introduces a novel metric called the Lane Change Dynamic Score (LCDS), which aggregates the collective effect of these kinematic parameters. The LCDS effectively classifies lane change events and distinguishes between different driving styles, such as aggressive and conservative. The study validates that detected lane changes can be confirmed using vehicle kinematics data alone, provided the initial detection leverages machine vision probabilities to ensure data reliability. Histograms and statistical analyses of the extracted data reveal distinct patterns in lateral acceleration, steering angle, and duration associated with specific driver profiles. The significance of this work lies in its human-centric approach to autonomous vehicle development. By providing a robust method for detecting and characterizing naturalistic lane changes, the research lays the groundwork for personalized autonomous systems. The proposed LCDS metric offers a quantifiable way to classify driving styles, which can be used to condition future models, such as Conditional Generative Adversarial Networks (CGANs), to generate trajectories that mimic individual driver preferences. This personalization is critical for enhancing driver trust and comfort, ultimately improving the safety and acceptance of higher-level automation features.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• lane changing
• lane positioning
• steering pattern
• naturalistic crash near crash
• work zones

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data
• Methodological Resource: dataset resource
• Theoretical Contribution: computational model