Machine learning for improved path loss prediction in urban vehicle-to-infrastructure communication systems

Area of Science:

• Wireless Communications
• Machine Learning
• Radio Propagation

Background:

• Reliable vehicle-to-infrastructure (V2I) communications are essential for intelligent transportation systems.
• Accurate path loss prediction is critical for V2I system design and performance optimization.
• Existing empirical models may struggle in complex urban radio environments.

Purpose of the Study:

• To investigate and compare the performance of machine learning models against traditional empirical models for V2I path loss prediction.
• To develop an enhanced path loss prediction model incorporating an environmental classification system.
• To identify key factors influencing path loss in urban V2I scenarios.

Main Methods:

• Empirical path loss measurements were collected at 5.9 GHz from eight Road Side Unit (RSU) sites.
• Machine learning models, including Extreme Gradient Boosting (XGBoost) and Multilayer Perceptron (MLP), were trained and evaluated.
• Performance was benchmarked against traditional Dual Slope and 3rd Generation Partnership Project (3GPP) models across open, suburban, and densely urbanized environments.
• A novel environmental classification system was developed based on building density, street geometry, and transmitter position.

Main Results:

• Machine learning models, particularly XGBoost, demonstrated superior performance with lower Root Mean Square Error (RMSE) compared to traditional models, especially in complex urban settings.
• The proposed environmental classification system improved prediction robustness.
• Feature importance analysis identified distance, environmental class, and transmitter height as critical predictors of path loss.

Conclusions:

• Machine learning techniques, especially XGBoost, offer a significant advancement for V2I path loss prediction accuracy in diverse urban environments.
• The developed environmental classification system enhances prediction reliability.
• Findings provide valuable insights for designing adaptive and reliable V2I communication systems.