A Decision Theoretic Framework for Message Prioritization via Machine Learning Algorithm in Vehicular Networks

Abstract

Connected and Autonomous Vehicles (CAVs) in the network require to process messages received from neighboring vehicles depending on various applications. Due to limits of hardware, the On-board Unit (OBU) on a vehicle cannot deal with such massive received messages in a certain time frame, especially in high-density road networks. That means only certain messages can be timely computed, making it a challenging work in concern of urgent messages response in the vehicular network. Current work focuses on prioritizing messages by a single attribute or manually setup, omitting an efficient framework for deciding how to prioritize messages for multiple attributes in the dynamic vehicular environment. In this paper, we propose a decision-theoretic framework enabling deterministic real-time support for prioritizing messages based on their contents. Specifically, we utilize the computation and storage resources of Roadside Units (RSU) to generate utility functions. Firstly the RSUs profile received messages to decide attributes, and generate utility functions which represent the current traffic status. These functions are then broadcast to the nearby vehicles, helping them to prioritize received messages using decision theory. We show the high efficiency and effectiveness of our framework through extensive simulations in different scenarios.