Enhancing Expressway Traffic State Perception: A Novel BAS-Optimized PSO-BP Fusion Model with Tensor Completion

Area of Science:

• Intelligent Transportation Systems
• Data Fusion
• Machine Learning

Background:

• Traditional single-source traffic data lacks sufficient spatial-temporal coverage and accuracy for intelligent expressways.
• Existing data preprocessing methods struggle to capture global spatiotemporal features.
• Conventional Particle Swarm Optimization-Backpropagation (PSO-BP) neural networks are susceptible to local optima.

Purpose of the Study:

• To develop a robust multi-source traffic data fusion model for enhanced intelligent expressway operation.
• To address limitations in data accuracy and spatiotemporal feature extraction.
• To improve the global search capability and convergence stability of traffic data fusion models.

Main Methods:

• Utilized fusion of Electronic Toll Collection-Dedicated Short Range Communication (ETC-DSRC) and RTMS microwave data.
• Employed the HaLRTC tensor completion algorithm for data repair and spatiotemporal correlation mining.
• Introduced the Beetle Antennae Search (BAS) mechanism into PSO to optimize a PSO-BP neural network (BSO-BP) for data fusion.

Main Results:

• The proposed BSO-BP model demonstrated significantly higher accuracy in predicting average road speed compared to single-source data and other benchmark models (BP, PSO-BP, GA-PSO-BP).
• The fusion model effectively captured spatiotemporal correlation characteristics of traffic flow.
• The BAS optimization improved the global search capability and convergence stability of the fusion model.

Conclusions:

• The BSO-BP model offers a superior approach for multi-source traffic data fusion in intelligent expressways.
• This method enhances the accuracy and reliability of traffic state estimation.
• The findings support more refined operation and management of intelligent expressway networks.