A Method for ACM on Q/V-Band Satellite Links Based on Artificial Intelligence

This paper compares classical algorithms for adaptive coding and modulation (ACM) with an approach based on artificial intelligence (AI). The proposed approach utilizes an online random regression forest (ORRF) to predict time series of signal to noise ratio (SNR) values aiding the ACM switching decisions. The evaluation of the ACM algorithms is based on two years of Q/V-band channel data recorded at the ground station in Graz using the Alphasat experimental Q/V-band payload. The results indicate that the ORRF based approach could outperform the classical approaches in terms of spectral efficiency, and parameterization of the ORRF is simpler and needs less knowledge of the channel properties as the discussed classical ACM approaches.

I. Introduction

In Graz, Austria, communication experiments have been performed since 2014 on the Q/V -band Aldo Paraboni payload hosted on Alphasat [1]. The goal of these communication experiments is to verify and optimize data transmission via DVB-S2 [2] over an experimental Q/V -band transponder (38/48GHz). Due to the significant fade dynamics of the Q/V -band channel, there are different fade-mitigation-techniques applied. A significant part of the experiments has been devoted to test and optimize adaptive coding and modulation (ACM).