A path selection method based on rule prediction in non-terrestrial networks

Abstract

Non-terrestrial networks (NTN) is becoming an attractive approach in the beyond 5G/6G era to enable ubiquitous connectivity, particularly in areas that are currently uncovered or underserved. NTN provides extensive coverage from the sky by utilizing satellites and unmanned aerial vehicles (UAVs) as mobile network nodes, such as base stations and routers. However, the mobility of these nodes in NTN leads to dynamic changes in network topology, which in turn reduces the opportunities and duration of NTN-ground communication. Additionally, variations in the communication environment, such as weather conditions, cause fluctuations in link quality and availability. Consequently, NTN faces challenges in maintaining a high packet delivery rate due to its dynamic topology and communication environment. This paper proposes a path selection method that uses link information-based path selection rule prediction in NTN. The proposed method selects paths based on rules predicted by a link information-based rule prediction model using machine learning (ML). The rule prediction model is trained using a dataset obtained through simulations of various NTN training scenarios. Simulation results over four evaluation scenarios show that the proposed method outperforms the existing methods in terms of packet delivery rate and its stability, even under severe weather conditions. The results further indicate that each path selection rule contributes to packet delivery, with the selective use of multiple path selection rules enabling the proposed method to adapt to various situations.