MCS Selection for Throughput Improvement using Machine Learning in Downlink LTE

Abstract

In Long-Term Evolution (LTE) networks, the desire for high data rates and an improved user experience have driven the demand for effective modulation and coding schemes (MCS) selection techniques in the downlink direction. This thesis investigates the use of machine learning (ML) algorithms to enhance MCS selection and increase throughput in LTE systems. First, a thorough analysis of the current MCS selection procedures is provided along with a list of their shortcomings. The difficulties brought on by the wireless channel's dynamic nature and the requirement for swift decision-making are highlighted. A innovative ML based technique is next suggested as an approach to these problems. The suggested method makes use of system parameters as well as prior channel state data to train ML models, facilitating thoughtful MCS selection options. In terms of throughput performance and complexity, various ML algorithms, including decision trees, support vector machines, and deep neural networks, are compared and evaluated. The efficiency of the ML-based MCS selection strategy in delivering substantial throughput gains compared to conventional methods is demonstrated by extensive simulations utilizing realistic network configurations. The outcomes open the door for additional research into using ML for next-generation wireless systems and validate the potential of ML for enhancing downlink performance in LTE networks.