AMC Channel Adjustment Using CRC Probability Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing adaptive modulation and coding (AMC) schemes face challenges in accurately reflecting channel state information (CSI), leading to erroneous judgments and delayed feedback, which affects the selection of modulation and coding schemes (MCS), thereby weakening system performance.
Innovation Solution
A model training method is employed to collect historical samples, including scheduling information and CRC results, to train a prediction model that predicts the probability of cyclic redundancy check (CRC) success or failure, allowing for more accurate channel adjustment based on predicted probability values rather than real-time CRC results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real-time CRC results are used for channel adjustment, then the feedback is immediate, but the judgment accuracy is low leading to erroneous decisions
Solution Approach 1:
The patent applies preliminary action by training a prediction model in advance using historical scheduling information and CRC results. This pre-trained model can then predict CRC outcomes for current scheduling scenarios without requiring real-time CRC computation, thus providing accurate channel state information ahead of time and eliminating feedback delay while maintaining high judgment accuracy.
Solution Approach 2:
The patent uses copying by creating a prediction model that replicates the relationship between scheduling information and CRC results based on historical data. Instead of directly measuring current channel state, the system copies the learned patterns from historical samples to predict current CRC outcomes, achieving both speed and accuracy in channel adjustment decisions.
2Measurement precision
If historical samples are collected and model training is performed, then the prediction accuracy of CRC results is improved, but the system complexity increases
Solution Approach 1:
The patent applies self-service by having the system automatically collect historical samples, train the prediction model, and use it for channel adjustment without requiring external intervention. The model continuously learns from historical data and self-updates, reducing the need for manual configuration and complex external control mechanisms while improving prediction accuracy.
Solution Approach 2:
The prediction model acts as an intermediary between historical CRC results and current channel adjustment decisions. Instead of directly complex interactions between multiple system components, the model mediates by translating historical patterns into predictive insights, simplifying the overall system architecture while maintaining high prediction accuracy.
3Productivity
If the prediction model is used to predict CRC probability, then the MCS selection accuracy is improved, but the computational overhead increases
Solution Approach 1:
The patent applies preliminary action by pre-training the prediction model offline using historical data. Once trained, the model can be deployed for real-time inference with minimal computational overhead. This separates the computationally intensive training phase from the real-time operation phase, allowing accurate MCS selection during actual data transmission without excessive energy consumption during throughput-critical operations.
Data Source
AI summary
The present application provides a model training method, a channel adjustment method, an electronic device, and a computer readable storage medium, the model training method includes: collecting historical samples, with the historical samples including first scheduling information and first information corresponding to a historical data transmission, the first information representing a result of cyclic redundancy check, and the first scheduling information including first intermediate variable information in an adaptive modulation and coding process; and performing model training according to the historical samples to obtain a first prediction model, and during the model training, the first scheduling information is used as an input of the first prediction model, the first information is converted into second information corresponding to the historical data transmission to be used as an output of the first prediction model, and the second information represents a probability value of the result of the cyclic redundancy check.