Efficient Last-Mile Link Adaptation in Next-Gen Wireless Heterogeneous Networks

The forthcoming generations of wireless networks have a high demand for reliable and enhanced data rate transmissions which are crucial for emerging smart cities. Thus, there is a need to develop efficient Link Adaptation (LA) for the last hop between end-users and the base stations in order to mitigate the severe interference resulting from the dense heterogeneous networks. We develop a Machine Learning (ML) based last-mile link adaptation method for a 5G wireless communication network. Dynamic selection of MCS for Resource Block (RB) allocation is efficient in terms of better network throughput and reduced BER which is verified through simulations for 5G New Radio (NR). We assume perfect channel estimation in our analysis. We have used a Deep Neural Network (DNN) model that dynamically selects the appropriate Modulation and Coding Scheme (MCS) ensuring 10 percent Bit Error Rate (BER) and maximizes the system spectral efficiency. Further, we evaluate the Signal to Interference and Noise Ratio (SINR) corresponding to varied channel states for different frequencies of operation and the DNN model selects the Channel Quality Indicator (CQI) corresponding to the optimal MCS available at the corresponding base stations for the end -users. This results in seamless connectivity for mobile users adapting to the last-mile link efficiently and achieving a higher downlink network throughput.