| Summary: | Digital connectivity plays a major role in enabling sustainable transport solutions. A key ingredient that ensures this connectivity is the cellular network coverage, whose information is typically available from the service providers. However, from a practical point of view, the actual connectivity experienced by the vehicles may vary significantly. In this study, we implement machine learning methods to build a network connectivity map for connected vehicles using minimal ingredients such as data from Scania vehicles, cellular tower locations and geographical features. We compare the performance of four machine learning models -Random Forest, XGBoost, AdaBoost and Neural Networks, to predict the delay time in receiving the messages from vehicles. Our results show that Random Forest is the best model for predicting connectivity with a Root Median Squared Error of 1.38sec. We demonstrate the results by building a connectivity map for Northern Sweden.
|
|---|