GNN-RMNet: Leveraging graph neural networks and GPS analytics for driver behavior and route optimization in logistics

Area of Science:

• Artificial Intelligence
• Machine Learning
• Deep Learning

Background:

• Complex logistics networks require real-time vehicle data for monitoring.
• Traditional methods fail to capture dynamic spatiotemporal driver behaviors and route anomalies.
• Intelligent systems are needed for efficient fleet management and safety.

Purpose of the Study:

• To introduce GNN-RMNet, a hybrid deep learning system for driver behavior profiling and route anomaly detection.
• To enable real-time analysis of spatiotemporal GPS trajectories and vehicle sensor data.
• To improve the interpretability, scalability, and efficiency of logistics monitoring.

Main Methods:

• GNN-RMNet combines Graph Neural Networks (GNN), ResNet, and MobileNet.
• Utilizes spatiotemporal GPS trajectories and vehicle sensor streams for contextual pattern learning.
• Employs a modular design for edge and on-vehicle inference.

Main Results:

• Achieved 98% accuracy, 97% recall, and 97.5% F1-score on a real-world dataset.
• Demonstrated high Anomaly Detection Precision (96%) and Route Deviation Sensitivity (95%).
• Reduced inference latency to 32 ms with a modular design.

Conclusions:

• GNN-RMNet offers accuracy, efficiency, and generalization advantages over baseline models.
• The framework supports real-time fleet safety, secure logistics, and intelligent transportation systems.
• Future work includes addressing cybersecurity, data privacy, and multimodal sensor integration.