LAUSR

Since Internet of Things (IoT) traffic in cellular networks tends to be explosively concentrated on uplink for a short period of time, the random access (RA) channel may be frequently congested. The 3GPP uses a $p$ -persistent access scheme called access class barring (ACB), which mitigates congestion by redistributing bursty access attempts over time. The ACB scheme can effectively control the congestion problem but treats equally all user equipments (UEs) in initial access. Thus, when a variety of UEs with different service requirements (e.g., for mission-critical applications and for delay-tolerant applications) coexist, the Quality of Service (QoS) of some UEs may not be guaranteed. In this article, we propose a new RA scheme that effectively alleviates a congestion situation of the RA channel, while providing the differentiated RA service to UEs in accordance with their QoS requirements. The proposed RA scheme consists of the load estimation and the optimization of an access probability based on the estimated load. Considering that the base station estimates the system loads by observing the status of received preambles, we formulate the proposed RA scheme as a partially observable Markov decision process (POMDP) framework. By using simulations, we demonstrate that the proposed scheme guarantees a short access delay requirement of high-priority UEs without any additional energy consumption of low-priority UEs during the access procedure.