Turning Adversity Into Advantage: Robust MCS Selection Utilizing the Uncertainty of Channel Prediction Neural Networks

Abstract

In wireless communication systems, the selection of suitable modulation and/or coding schemes (MCS) based on predicted channel quality is vital. However, the dynamic nature of wireless channels poses a challenge, leading to persistent inaccuracies in prediction. This letter tackles this challenge by introducing a robust communication approach for MCS selection, rooted in new paradigm of prediction uncertainty quantification. The proposed method involves establishing an uncertainty pool to quantify potential prediction errors. This approach enables the assessment of the accuracy of predictions, facilitating the informed selection of MCSs. Through simulations utilizing real-world data, the integration of our proposed design into various channel prediction neural networks enhances the performance of MCS selection. The simulations show our design enhances communication robustness without compromising throughput. The enhancement of up to 53% in communication success rates, with an average improvement of around 15%.