ABSTRACT
The massive machine-to-machine communication (mM2M) has led to several research challenges in 5G cellular Internet-of-Things (IoT). Here, radio access network congestion is the main challenging task due to the presence of sporadic mM2M traffic, huge signaling overhead, and quality of service (QoS) provisioning. The mM2M devices should connect to the base station using a random access channel (RACH) mechanism. However, the mM2M devices increase network congestion for RACH. To control the number of RACH accesses, the 3rd Generation Partnership Project (3GPP) has proposed extended access barring mechanism (EAB). Moreover, several schemes are proposed in the literature to enhance the performance of 3GPP-EAB. However, these mechanisms need to solve highly complex and long-term optimization problems to configure the network parameters for future transmissions based on the traffic in the current transmissions. Machine learning (ML)-based approaches solve these high-complexity optimization problems efficiently in comparison to non-ML-based approaches. Even though these ML approaches give promising results, their computational complexity increases exponentially. This results in a large carbon footprint during training and testing when deployed in real-time. Thus, in this chapter, green machine learning approaches are presented which are efficient for mM2M in 5G cellular IoT.
