Supervised Learning Approaches to Link Adaptation in Wireless Communication Systems

Current wireless communication systems require link adaptation method to provide consumers with reliable and efficient services. Adaptive modulation and coding (AMC) based on channel state information is the most common way of implementing link adaptation. Traditional implementation of AMC attempts to solve an optimization problem with the goal of maximizing the channel throughput with packet error rate (PER) constraints. Our project here considers the use of machine learning approaches for adaptive modulation and coding. We employ three supervised learning techniques, namely k-nearest neighbor (k-NN), support vector machine (SVM) and random forest (RF), to enable the transmitter to perform AMC based on the knowledge of post-processing SNRs at the receiver (through feedback). Simulation results based on IEEE 802.11n standard show that AMC with machine learning would be promising in real systems.