Learning Based Access Class Barring for Massive Machine Type Communication Random Access Congestion Control in LTE-A Networks

Massive Machine Type Communication (mMTC) is an enabling technology for cellular network, like LTE-A, scalability and densification. However, LTE-A network is mainly designed for Human Type Communications (HTCs) and hence the inclusion of large number of MTC devises (MTCDs) results in several challenges including Random Access (RA) congestion. With the availability of limited RA resources per LTE-A cell, a massive bursty access request from MTCDs results in RA network overload. This is a severe problem in that it degrades the RA throughput at large which results in access delay, increased energy consumption (by the MTCDs), and hence degraded Quality of Service (QoS). To this end, 3GPP has included Access Class Barring (ACB) scheme in the LTE-A specifications to control the congestion. The scheme controls the congestion by spreading the devices access requests in time using its parameters namely, barring factor and barring time. Thus, literatures show that the scheme can effectively manage Human Type Communications (HTCs) and MTCs' sporadic access requests. However, its performance (in terms of access success probability, access delay, and energy efficiency) degrades when the RA network gets continuously congested. Therefore, learning techniques are used by different researchers to adaptively vary one or both parameters of the scheme along with the access request intensity. In this paper we investigate learning based ACB solutions. For ease of study, we grouped the contributions on learning based ACB solutions into two classes namely; adaptive barring factor ACB solutions and adaptive barring factor and barring time ACB solutions. The achieved performance improvements and the limitations of each class of schemes are identified in the paper.