Machine learning applications for heterogeneous networks

Next-generation (NG) cellular networks, such as beyond fifth-generation (B5G), promise data rates in the range of Tbits/s, ultralow latency, and most importantly mass connectivity. 5G systems are already being rolled out and are not expected to meet all these requirements, and therefore, 6G is expected to address the shortfall. One of the most important aspects of the sixth-generation (6G) system is seamless global connectivity with significantly improved energy and spectrum efficiency. 6G is expected to generate a massive volume of heterogeneous data that needs to be analyzed to support various services. Since data traffic is a major constituent of cellular traffic, a low-cost and effective solution for handling this traffic is the deployment of small cells. The deployment of small cells improves cellular coverage and capacity. The interoperation of these small cells with macro-cells forms a heterogeneous network (HetNet). Initial studies indicate that HetNets will be an integral part of 6G. This chapter presents an overview of HetNets, discusses research challenges, and outlines how a machine learning (ML) approach can address some of the challenges.