Link Adaptation in MIMO Systems by Using Machine Learning

Abstract

Optimizing the performance of link adaption in practice has proven challenging. The core of this problem is how to deal with the channel state information (CSI), modulation and coding scheme (MCS).Furtherly, it is exploring the relationship between CSI and Channel Quality Indicator (CQI). In order to get CQI, prior research struggled to build a single valued signal-to-noise ratio (SNR) by CSI. However, the results of these methods are not good enough. We consider using the machine learning method to explore the relationship between CSI and CQI. It is not a simple classification problem, because the premise of link adaptation is to satisfy the reliability of the system, and the target is maximizing the spectral efficiency. In this paper, present a novel link adaption scheme in the MIMO systems by using an improved kNN algorithm, in the system setting of maximizing the spectral effectivenss. And the improved kNN algorithm is named as E cost kNN. The main emphasis is placed on the problem of how to build a highly individualized cost function for it. With the aid of using E cost kNN, the link adaption of MIMO systems can get better performance. The simulation result reflects the proposed novel method significantly outperforms the existing kNN algorithm in the link adaption technology.