A Comparative Analysis of Machine Learning Methods for Lane Change Intention Recognition Using Vehicle Trajectory Data - Research and report

Aug 5, 2026

Upload by Patrick

Accurately detecting and predicting lane changes improves autonomous driving’s understanding of surrounding traffic, helps identify potential safety hazards, and supports safer, more efficient road operations. The paper compares multiple machine learning methods for recognizing lane change (LC) intention from high-dimensional vehicle trajectory time-series. Using 1,023 trajectories extracted from the CitySim dataset, results show ensemble methods achieve 98% classification accuracy and reduce Type II/III error impacts, while LightGBM improves training efficiency about sixfold over XGBoost without sacrificing recognition accuracy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Introduction

Vehicle Trajectory Data

Data sets used (NGSIM, CitySim, HighD)

Methodology Overview

Problem framing and model comparison dimensions

Experimental Setup and Results

Accuracy and training complexity comparison

Conclusions and Limitations

A Comparative Analysis of Machine Learning Methods for Lane Change Intention Recognition Using Vehicle Trajectory Data Renteng Yuan (First author) Jiangsu Key Laboratory of Urban ITS School of Transportation Southeast University, Nanjing, Jiangsu, P. R. China, and 210000 Email: [rtengyuan123@126.com](rtengyuan123@126.com) Abstract： Accurately detecting and predicting lane change (LC)processes can help autonomous vehicles better understand their surrounding environment, recognize potential safety hazards, and improve traffic safety. This paper focuses on LC processes and compares different machine learning methods' performance to recognize LC intention from high-dimensionality time series data. To validate the performance of the proposed models, a total number of 1023 vehicle trajectories is extracted from the CitySim dataset. For LC intention recognition issues, the results indicate that with ninety-eight percent of classification accuracy, ensemble methods reduce the impact of Type II and Type III classification errors. Without sacrificing recognition accuracy, the LightGBM demonstrates a sixfold improvement in model training efficiency than the XGBoost algorithm. Introduction Lane-changing behavior induces spatiotemporal interactions between vehicles, which have a significant impact on both traffic efficiency and safety. Statistical data shows that lanechanging behaviors are responsible for 18% of all roadway crashes and contribute to 10% of delays in China[1] . In 2015, the National Highway Traffic Safety Administration (NHTSA) reported approximately 451,000 traffic accidents in the US related to lane-changing behavior[2] . Timely identification and prediction of lane-changing behaviors are crucial for reducing accidents, enhancing traffic safety, and optimizing road operations. Lane-changing behavior is a complex process influenced by various factors, including human, vehicle, road, and environmental factors[3] . Lane change intention is defined as the planned or intended action of a driver to change lanes while driving. In previous studies, various indicators are used to characterize lane-changing intentions, such as vehicle dynamics parameters (e.g., steering wheel angle, rate of steering angle change, brake pedal position, and turn signal status)[4-6], driver's physiological indicators (e.g., eye movements and head rotation angle)[7-9], and vehicle operating state indicators (e.g., speed, acceleration, and headway distance)[10-12]. However, the practical application of vehicle dynamics parameters is limited due to variations in driving habits, affecting the recognition performance of related models. For example, turn signal utilization rates for lane-changing vehicles have been reported to be 44% and 40% in the United States and China, respectively[13, 14] . The acquisition of driver's physiological indicators faces challenges related to experimental conditions, such as small sample sizes and high data homogeneity, which can impact the transferability and reliability of trained models. Monitoring driver's physiological characteristics faces challenges related to data quality, cost, and potential discomfort to the driver. With the development of vehicle-tovehicle communication and vehicle-to-infrastructure technologies, access to personalized, high-precision vehicle trajectory data has increased. Vehicle operating statue indicators can be extracted directly from the vehicle trajectory and are increasingly used for lane change intention recognition due to their easy accessibility and large sample size[15-17] . Compared to conducting real-world or driving simulator experiments, vehicle trajectory data is more accessible and overcomes limitations related to small sample sizes and data homogeneity. This study focuses on using vehicle trajectory data to identify lane-changing behaviors by considering interactive influences among adjacent vehicles. From the methodology perspective, lane-changing intention r

FAQs

What is the main goal of the lane change intention recognition study?

The study aims to recognize lane change intention from high-dimensional vehicle trajectory time-series to improve traffic understanding and safety.

Which dataset and data size are used to validate the models?

Validation uses 1,023 vehicle trajectories extracted from the CitySim dataset.

How do ensemble methods and LightGBM perform compared with alternatives?

Ensemble methods reach 98% classification accuracy and reduce the impact of Type II and Type III classification errors; LightGBM offers about sixfold higher training efficiency than XGBoost without losing accuracy.

Aug 5, 2026

Upload by Patrick