Edinburgh Napier University

  The in-flight connectivity (IFC) turns to a crucial need from luxury with technological advances. The WiFi-enabled IFC (W-IFC) meets most of this need by deploying access points within the aircraft. These access points can allow Internet connectivity through various core network links air-to-ground (A2G), air-to-satellite (A2S), and air-to-air (A2A) at different times during the flight. More specifically, the core network of W-IFC should be selected from these links according to their availabilities throughout the aircraft's flight. However, the ultra-dynamic characteristic of aeronautical networks caused by aircraft's high speed reduces W-IFC's core network selection efficiency. The problems on the core network selection of W-IFC increase the core network selection delay with higher packet losses. Additionally , the core network selection of W-IFC becomes more complex when user traffic heterogeneity is added to this connection availability. These complexities necessitate the continuous monitoring of the aircraft environment while dealing with multiple data entries. At that point, the digital twin (DT) technology enables us a continuous monitoring and management opportunity in a virtual manner for the ultra-dynamic aeronautical environment. By considering this, in this article, we aim to introduce a proof-of-concept (PoC) about the utilization of digital twin technology in WiFi core network selection for IFC. Our proposed DT module executes the hybrid combination by utilizing the connectivity and traffic-based core network selection models simultaneously. Here, the connectivity-based core network selection focuses on determining aircraft's possible core network links, while the traffic-based selection considers heterogeneous traffic flows of passengers. Results reveal that the proposed DT-controlled model reduces the WiFi core network selection delay 36 percent with 25 percent packet delivery improvement. Also, we prove the feasibility of the PoC W-IFC model through twinning rate with near-real-time measurements. And we show the decision performance of DT with false positive and false negative rates.