Enhancing IoT Connectivity in NTN: A Machine Learning Approach for Packet Detection in Grant-Free Access

Non-terrestrial networks (NTN) are key to the 6G-Internet of Things (IoT) ecosystem, providing broader coverage and reliable connectivity. As IoT adoption grows, the demand for scalable communication solutions increases. IoT communications, relying on grant-free random access protocols, face challenges in packet detection due to uncoordinated transmissions. Traditional detection methods struggle in NTN environments, especially due to interference and Doppler shifts. This paper proposes a convolutional neural network for efficient packet detection in IoT-NTN scenarios. The method outperforms a traditional correlator-based approach, showing superior detection performance under harsh conditions. Our results highlight the potential of machine learning for enhancing IoT connectivity over NTN.