Enabling mMTC and URLLC in 5G : Initial Access, Traffic Prediction, and User Availability

The 5th generation (5G) mobile communication networks focus on three main technology pillars as enhanced mobile broadband (eMBB), ultra-reliable and low latency communications (URLLC), and massive machine type communications (mMTC). Among them, URLLC and mMTC introduce many novel challenging tasks for system and protocol design. This is a result of the stringent service requirements associated with URLLC and the massive number of devices that constitute mMTC. In this dissertation, we investigate three topics towards enabling mMTC and URLLC in 5G networks as stated below. The first topic which is the main focus of this dissertation lies in proposing initial access approaches in URLLC and mMTC scenarios. Initial access plays an essential role for end device to base station communications. In fourth generation, the grant based (GB) long term evolution-advanced (LTE-A) random access procedure was adopted for initial access. However, when a large number of devices compete for scarce radio resources, the LTE-A random access procedure causes higher device collisions resulting in thereby long latency. The proliferation of devices triggered by massive Internet of things (mIoT)/mMTC deployments and the stringent access requirements of futuristic applications further exaggerates this problem. Moreover, mMTC devices typically generate small packets making the initial control message exchange in LTE-A random access a major burden. As an e ort to address these shortcomings in existing initial access procedures, this dissertation proposes both grant based and grant-free access schemes. The proposed GB schemes enable both device and resource grouping to give priority access for mMTC devices with URLLC requirements considering a large device population. Meanwhile, the proposed GF access scheme enables priority access for high priority devices based on a heterogeneous traffic arrival scenario. A two dimensional Markov model that aggregates both high and low priority traffic through a pseudo-aggregated process is developed. Furthermore, prediction of traffic arrivals at a base station is of paramount importance for successful detection of congestion and to initiate proactive measures to minimize collisions. In reality, a base station only has information about the number of detections which does not naturally reflect the number of arrivals that caused it. Therefore, as the second topic, this dissertation presents a machine learning based traffic arrival prediction model for