| Summary: | This thesis investigates the performance of 802.11p-based V2V communication in real-life scenarios, and explores potential practical applications such as GNSS correction data broadcasting to improve the positioning accuracy of nearby vehicles, and enhancing communication robustness by preemptively predicting potential disruptions with the assistance of Machine Learning (ML) models. A custom V2V On-board Unit (OBU) hardware platform was developed, and real- world multi-vehicle outdoor experiments were planned and carried out. The collected data was examined and used to train a number of ML models, and their performance was compared.
The experiments revealed that the custom OBU was fully functional, and signal quality and communication range were observed to be affected by real-world imperfections. The GNSS correction data broadcasting was shown to notably increase the positioning accuracy of nearby vehicles, and the ML models trained from Key Performance Indicators (KPIs) demonstrated excellent prediction accuracy, allowing pre-emptive actions to be taken to reduce the downtime from communication disruption.
|
|---|