Abstract:
A critical issue on the telecommunication systems relies on serving different demands of service specification in an on-demand fashion. This approach has gained more and ...Show MoreMetadata
Abstract:
A critical issue on the telecommunication systems relies on serving different demands of service specification in an on-demand fashion. This approach has gained more and more attention with the Internet of Things (IoT) popularity. One possibility is the co-existence of plenty of massive machine-type communication (mMTC) users' equipment (UEs) and mission-critical ultra-reliable low latency (URLLC) UEs. The first is characterized by massive connectivity, while the latter is constrained to minimum latency and the highest reliability. Serving both services simultaneously lies in a demanding network coordination and resource allocation problem, especially on random access (RA). In this paper, we propose a novel fully RA procedure specially designed for the network traffic employed in mixed URLLC-mMTC scenarios. We characterize the access patterns from both URLLC and mMTC use modes, as well as their packet size based on 3GPP standards in which both mMTC and URLLC use modes present a long-term traffic regularity. Taking into account such characteristics, we employ a long short-term memory (LSTM) neural network (NN) as a traffic forecast framework. The predicted traffic is the key enabler for a novel resource slicing (RS) scheme able to fully optimize the time and frequency (T-F) grid. The RS deals with the available spectrum and time slots to efficiently assign channels to the upcoming traffic at every frame. The access procedure is performed by a modified access class barring (ACB)-based collision mitigation protocol. We present results for synthetic and real-world traffic patterns, showing the scheme is capable of fulfilling Ultra Reliable Low Latency Communications (URLLC) requirements with high probability while keeping massive machine-type communications (mMTC) UEs latency limited to 18 frames (180 ms).
Published in: IEEE Transactions on Network Science and Engineering ( Volume: 11, Issue: 3, May-June 2024)