Integrating Surrounding Vehicle Information for Vehicle Trajectory Representation and Abnormal Lane-Change Behavior...

Area of Science:

• Traffic Safety
• Artificial Intelligence in Transportation
• Vehicle Dynamics

Background:

• Abnormal lane-changing behavior poses significant risks to road safety, impacting traffic management and law enforcement.
• Existing detection methods often overlook the crucial influence of surrounding vehicles on lane-changing maneuvers.
• A comprehensive approach is needed to accurately identify and mitigate risks associated with abnormal lane changes.

Purpose of the Study:

• To propose a novel framework for detecting abnormal lane-changing behavior in road vehicles.
• To develop a trajectory representation that incorporates data from surrounding vehicles.
• To enhance the accuracy and reliability of abnormal lane-changing detection models.

Main Methods:

• A new method for representing vehicle trajectories, integrating surrounding vehicle information.
• Extraction of feature parameters that capture inter-vehicle interactions and lane-changing phases.
• Development of an abnormal lane-changing behavior detection model using the Light Gradient Boosting Machine (LGBM) algorithm.

Main Results:

• The proposed framework demonstrates high accuracy in detecting abnormal lane-changing behaviors.
• The integration of surrounding vehicle information was found to be a critical factor in improving detection outcomes.
• The LGBM model effectively distinguished between normal and abnormal lane-changing maneuvers.

Conclusions:

• The developed framework offers a more robust and accurate method for abnormal lane-changing detection.
• Considering the dynamics of surrounding vehicles is essential for a comprehensive understanding of lane-changing safety.
• This research contributes to safer traffic management and improved road safety through advanced AI techniques.