A Novel Intelligent Approach to Lane-Change Behavior Prediction for Intelligent and Connected Vehicles

Area of Science:

• Intelligent and Connected Vehicles (ICVs)
• Autonomous Driving Systems
• Traffic Safety

Background:

• Predicting lane-change behavior is crucial for safe navigation in intelligent and connected vehicles (ICVs).
• Existing methods often lack comprehensive analysis of driving environments and vehicle trajectories.
• Accurate prediction enables proactive decision-making for safer lane changes.

Purpose of the Study:

• To propose a novel intelligent approach for predicting lane-change behaviors in ICVs.
• To develop a model that considers both driving style-based environments and trajectory parameters.
• To enhance the safety and efficiency of lane changes in connected vehicles.

Main Methods:

• A modified dataset was created using the Next-Generation Simulation (NGSIM) dataset of real vehicle trajectories.
• A hidden Markov model (HMM)-based model was employed to assess lane-change suitability based on environmental parameters.
• A learning-based prediction-then-judgment model was designed to predict ICV lane-change behavior.

Main Results:

• The HMM-based model accurately judged the surrounding lane-change environment, with lane-change probability values generally above 0.5.
• The learning-based prediction-then-judgment model achieved 99.32% accuracy in predicting lane-change behavior.
• The lane-change detection algorithm within the model demonstrated 99.56% accuracy.

Conclusions:

• The proposed approach effectively predicts lane-change behavior in ICVs.
• The system accurately assesses lane-change suitability and predicts future actions.
• This research contributes to safer and more reliable autonomous driving systems.