Neural Network based Non Orthogonal Random Access for 6G NTN-IoT

Internet of Things (IoT) devices have become increasingly more pervasive and distributed. To provide connectivity to the massive amount of IoT devices and to satisfy the need of an ubiquitous and resilient coverage, Non-Terrestrial Network (NTN) will be pivotal to assist and complement terrestrial systems. In particular, due to the fact that IoT communication is mainly characterized by sporadic uplink data reports, non-continuous satellite coverage, provided by cost efficient incomplete LEO constellations, is a baseline approach for most of the foreseen IoT-NTN architectures. In such configurations, all the terminals within a satellite beam must be served during the short satellite visibility window, thus generating congestion because of IoT devices simultaneously trying to access the same resources. When the number of colliding terminals increases, the number of successful access decreases, and the average time to complete the access increases. A possible countermeasure to this problem is represented by Non-Orthogonal Multiple Access scheme, which requires the knowledge of the number of users transmitting on the same resources. In this paper, we address this problem by proposing a Neural Network (NN) algorithm to cope with the uncoordinated random access performed by a massive number of Narrowband-IoT devices. Our proposed method classifies the number of colliding users and for each of them estimates the Time of Arrival (ToA). The performance assessment, under Line of Sight (LoS) and Non LoS conditions in sub-urban environments with two different satellite configurations, shows significant benefits of the proposed NN algorithm with respect to traditional methods for the ToA estimation.