Predicting CAM generation times through machine learning for cellular V2X communication

Abstract

In the narrow Intelligent Transportation System (ITS) band, avoiding wireless channel congestion is essential. Reporting vehicle kinematics in the Cooperative Awareness Message (CAM) only when there are notable changes in vehicle dynamics is a standardized approach to reducing bandwidth usage of periodic CAM messages that are exchanged between vehicles, and is called the CAM generation rule. However, in cellular vehicle-to-everything (V2X) communication, aperiodicity due to frequent omissions of periodic CAM raises problems of resource waste and instability in resource scheduling. The problem can be solved by reserving a resource only for actual CAM transmission times in the future. This article demonstrates that a neural network-based scheme can predict the next CAM generation times at an average accuracy of over 94%, which can be utilized for resource reservation under the CAM generation rule.

Keywords

• V2X
• Cooperative Awareness Message
• Congestion control
• Machine learning
• Prediction