Real-time prediction of communication link quality for V2V applications

In this paper we address prediction of the communication link quality for Vehicle-to-Vehicle (V2V) applications. We focus on the prediction at the receiver vehicle and suggest two novel frameworks, which allow real-time and short-term prediction whether a predefined application-specific QoS will be maintained in the near future. First framework makes use of machine learning approach, and the second one is realized with model-based estimation. Both frameworks are developed based on the measurement data which was gathered over the 5,5 month of the field trials in the sim^TD project. In our paper we also suggest an optimization method to increase prediction accuracy and validate both frameworks through an additional real-world measurement campaign. The main advantages of suggested algorithms comparing to the existing work is their completely generic nature and low to no memory requirements. We demonstrate that in low network density scenarios prediction accuracy can reach up to 97%.