Prediction of Packet Inter-Reception Time for Platooning using Conditional Exponential Distribution

We present a conditional exponential distribution model for the prediction of the packet inter-reception time. This value is our objective in terms of quality of service prediction for a high-density platooning application. This application indeed requires the exchange of information with low latency and high reliability. The prediction of the packet inter-reception time is therefore of high interest for this scenario. We extend our discrete kernel density estimation model to a continuous model based on the parametrization of an exponential distribution. In this model, the rate of the distribution is a function of our predictors, the inter-antenna distance and the spatial distribution of the surrounding communicating vehicles. We approach this function using a multi-layer perceptron regressor for different granularity values of our spatial model. Based on simulation data obtained with a combination of ns-3 and SUMO, we show that our model performs better than our previous model. This performance, combined with the advantages of being continuous and being able to be trained on-line, is promising for a further deployment and test on other scenarios. We present further results on the influence of the interaction between inter-antenna distance and spatial distribution of the communicating vehicles. Finally, we show a successful example of on-line instantaneous prediction of the packet inter-reception time along with the observed values of the predictors and the experienced packet inter-reception time.