AI MIMO Detection with QR Preprocessing for Lower Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
MIMO detectors face high computational complexity due to extensive processing requirements, especially in massive MIMO systems with numerous antennas, necessitating improved algorithms and hardware resources.
Innovation Solution
Implementing a QR decomposition preprocessing of the estimated channel matrix (H) followed by a deep neural network (DNN) to generate LLR values, reducing the network size and processing complexity through AI MIMO detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MIMO detectors search over a full set of possible outcomes to detect transmitted data streams, then detection accuracy is improved, but computational complexity increases significantly
Solution Approach 1:
The patent segments the detection process into two stages: first applying QR decomposition to transform the MIMO detection problem into a simpler form, then using a deep neural network to handle the remaining detection task. This segmentation reduces the search space from a full set of possible outcomes to a more manageable subset, thereby reducing computational complexity while maintaining detection accuracy.
Solution Approach 2:
The patent introduces QR decomposition as an intermediary step between the received signal and the deep neural network detector. The QR transformation pre-processes the input data to eliminate interference terms, making the subsequent neural network detection easier and more efficient. This intermediary transformation reduces the computational burden on the neural network while preserving detection performance.
2Quantity of substance
If massive MIMO systems with numerous antennas are deployed to improve channel capacity, then MIMO channel capacity is improved, but processing complexity increases extensively
Solution Approach 1:
For massive MIMO systems, the patent applies QR decomposition to segment and simplify the large-scale MIMO detection problem. By transforming the channel matrix into upper triangular form, the complex multi-antenna detection problem is broken down into simpler sequential detections, reducing the processing complexity that would otherwise scale extensively with the number of antennas.
Solution Approach 2:
The patent replaces traditional mechanical/computational MIMO detection algorithms with a deep neural network-based approach. The neural network learns optimal detection strategies during training and can perform detection much faster than conventional algorithms, especially for massive MIMO systems where the number of antennas creates extensive processing requirements.
3Device complexity
If deep neural network is used for MIMO detection to reduce computational complexity, then processing complexity is reduced, but training time and data requirements increase
Solution Approach 1:
The patent applies QR decomposition as a preliminary action before feeding data to the deep neural network. This pre-processing step transforms the input data into a form that is more suitable for neural network detection, reducing the training time required. By eliminating interference terms through QR transformation beforehand, the neural network needs to learn fewer complex patterns, thus reducing training time and data requirements.
Data Source
AI summary
The subject matter discloses an AI (Artificial intelligent) MIMO (Multiple INPUT Multiple Output) radio channel data detector which includes QR decomposition preprocessing of the channel matrix (H) and providing related LLR results to enable the close to optimal data detection.