Research on the Channel Information Forecast and Adaptive Coding and Modulation with Reinforcement Learning

In GEO satellite communications, effectiveness of the channel estimated information is insufficient due to the large time-delay in transmissions and mode selection mechanism in existing adaptive coding and modulation(ACM) is also not flexible and efficient. In view of above-mentioned problems, we propose a channel information forecast method based on corrected time-series to improve the effectiveness of estimated channel information. We propose a more efficient ACM modulo based on the channel information forecast and reinforcement learning with Q-learning. In this paper, we fit bit error rate and spectral efficiency into one criteria to ensure the fairness between performance and quality. After simulation, we get the conclusion that the error of proposed channel information forecast method is 0.792db and proposed ACM modulo not only lower the bit error rate by 43.5% but also optimize the utilization of spectrum by 42.3%.