AI Wireless Channel Prediction for Non-SRS Slot Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current channel estimation methods in 5G systems fail to utilize side information from RF and user characteristics, leading to inadequate performance in dynamic and time-varying environments due to channel aging and under-sampling, especially in mobile scenarios with high doppler frequencies.
Innovation Solution
An AI-assisted wireless channel prediction system that uses a trained neural network to generate accurate channel estimates and predictions by leveraging historical low-resolution two-dimensional image representations of channel estimates over time and frequency, incorporating frequency correlation and user characteristics, and employing convolutional and recurrent neural networks to interpolate and predict channel responses for non-SRS slots.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional channel estimation methods are used in 5G systems, then the system structure remains simple and easy to implement, but the channel estimation accuracy deteriorates in dynamic and time-varying environments due to channel aging and under-sampling
Solution Approach 1:
The patent introduces an AI model as an intermediary component between the SRS receiver and the channel estimation output. This AI model processes the SRS measurements and generates enhanced channel estimates by learning from historical low-resolution two-dimensional image representations (TDIR) of channel estimates over time and frequency, thereby improving accuracy without requiring fundamental changes to the existing 5G system architecture
Solution Approach 2:
The patent applies preliminary action by pre-training the AI model offline using historical TDIR data before deployment. The model learns channel characteristics and patterns in advance, enabling it to make accurate predictions during real-time operation without adding computational complexity to the live estimation process
2Measurement precision
If SRS slots are transmitted frequently to improve channel estimation accuracy, then the measurement precision improves, but the loss of time and system resources increases
Solution Approach 1:
The patent uses copying by creating virtual channel estimates for non-SRS slots based on the AI model's predictions. Instead of transmitting actual SRS signals in every slot, the system generates copied channel information using the trained AI model, which has learned from historical TDIR data, thereby maintaining accuracy while reducing the frequency of actual SRS transmissions
Solution Approach 2:
The AI model serves multiple functions: it processes SRS measurements, predicts channel states for non-SRS slots, and generates predictions based on historical patterns. This multi-functionality allows the system to maintain accurate channel estimation without requiring dedicated SRS transmissions in every time slot
3Reliability
If SRS slots are transmitted frequently to capture channel variations, then the reliability of channel estimation improves, but the productivity of the system deteriorates due to reduced data transmission capacity
Solution Approach 1:
The system creates virtual channel estimates for non-SRS slots using the AI model's predictions, allowing reliable channel information to be available without actual SRS transmissions in every slot. This copying approach maintains reliability while freeing up time resources for data transmission, thereby preserving system throughput
4Loss of substance
If the sampling rate of SRS slots is reduced to save resources, then the loss of substance decreases, but the measurement precision deteriorates due to under-sampling in mobile scenarios
Solution Approach 1:
The AI model acts as an intermediary that processes the limited SRS measurements and compensates for the under-sampling effect. By learning from historical TDIR data, the model infers channel characteristics that would otherwise be missed due to reduced sampling, thereby maintaining accuracy despite lower resource consumption
Solution Approach 2:
The system generates virtual channel estimates for non-SRS slots by copying and extrapolating channel information using the AI model's predictions, effectively reconstructing the full channel behavior from sparse SRS measurements and reducing the need for frequent sampling
Data Source
AI summary
This disclosure relates to methods, systems, and devices for AI/ML assisted wireless channel fingerprinting, estimation, and prediction. In one example embodiment, a method of combined AI/ML assisted wireless channel fingerprinting and channel prediction is disclosed. The method includes using a trained neural network to fingerprint the channel with the channel fingerprinting results advantageously being leveraged to improve the channel prediction.