Using Time Series Analysis to Generate Synthetic Traffic Data for LoRa Network Simulation

Modelling radio signal attenuation in Low Power Wide Area Networks (LPWANs) is a complex challenge due to the numerous factors influencing signal propagation and attenuation. Recent advancements in artificial intelligence have enabled the use of neural network-based time series analysis to develop radio attenuation models suitable for radio simulators. In this paper, we evaluate the feasibility of such AI-driven models for discrete event simulations. To train and validate these models, we utilize data collected from a large-scale LoRaWAN network that has been operational for over six years. The trained model is then applied to generate synthetic data for radio network simulations. We compare the AI-generated data with conventional distribution-based generators, assessing their effectiveness in modelling real-world events and analysing the statistical distribution of the generated data. Finally, we discuss the benefits and limitations of using AI-based models for simulation purposes. While traditional generators typically produce data distributions that closely mirror real data, AI algorithms are more effective at capturing specific patterns found in actual datasets what is challenging with conventional methods.