Downlink Channel Quality Prediction for NTN MCS Adaptation

The patent replaces probabilistic models with machine-learning-based prediction models that utilize ego-sensor information and sidelink data. This substitution enables more accurate prediction of future downlink channel quality by leveraging actual sensor measurements and communication data from the radio terminal, thereby resolving the contradiction between model simplicity and prediction accuracy in non-terrestrial networks

The patent introduces an intermediary prediction mechanism that uses machine-learning functions as a mediator between raw sensor/communication data and channel quality predictions. This intermediary layer processes ego-sensor information and sidelink data to generate accurate future channel quality indicators, enabling high-precision predictions without requiring overly complex direct modeling approaches

The patent extracts only the essential prediction information (future downlink channel quality indicator and estimation quality indicator) from the terminal's extensive sensor and communication data. By taking out only the necessary prediction components rather than transmitting all raw data, the system achieves improved spectral efficiency through better modulation and coding adaptation while minimizing information transmission overhead

The patent applies partial action by transmitting only the specific prediction indicators needed for channel quality assessment rather than all possible terminal data. This selective transmission approach provides sufficient information for accurate MCS adaptation without the excessive overhead of transmitting complete raw data sets, thereby optimizing the balance between productivity and information loss