A Lane-Changing Decision-Making Model of Bus Entering considering Bus Priority Based on GRU Neural Network.

Area of Science:

• Intelligent Transportation Systems
• Deep Learning in Public Transportation

Background:

• Mandatory bus lane changes for station entry reduce urban road capacity and cause congestion.
• Deep learning for bus lane-changing decisions is a growing research area, especially with Cooperative Vehicle-Infrastructure Systems.

Purpose of the Study:

• To explore bus lane-changing rules and decisions during station entry.
• To develop a deep learning model for predicting bus lane-changing maneuvers.

Main Methods:

• Utilized a real-world Vehicle-to-Everything (V2X) dataset.
• Processed image and point cloud data via coordinate transformation.
• Applied Kalman filtering for vehicle state evaluation.
• Developed a Gated Recurrent Unit (GRU) neural network trained with XGBoost, incorporating bus priority rules and a flexible right-of-way lane concept.

Main Results:

• The proposed GRU model demonstrated higher accuracy in identifying bus lane-changing maneuvers compared to other models.
• The model effectively uses processed sensor data and bus priority rules for decision-making.

Conclusions:

• The developed deep learning model provides a robust decision basis for optimizing bus operations.
• This approach is crucial for refining urban bus management and mitigating traffic congestion.