Low Complexity ML Detector for MIMO Spatial Streams
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Maximal likelihood (ML) detection in MIMO systems faces high computational complexity, especially with high-order modulation schemes like 256-QAM, making real-time implementation challenging due to the NP-hard nature of finding nearest lattice coordinates, and existing methods like sphere decoding and list detectors suffer from performance loss due to incomplete candidate sets.
Innovation Solution
A low complexity ML detection method for 2n-QAM signals in MIMO spatial streams that estimates an initial solution and iteratively determines search centers, either as the initial solution or its mirror constellation point, to generate a candidate search set, allowing for reduced processing throughput while maintaining performance by including non-adjacent points and using a Look-Up Table for efficient candidate selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If ML detection is used to reduce SNR requirement, then detection performance is improved, but computational complexity increases to O(ML)
Solution Approach 1:
The patent segments the exhaustive ML search space into multiple smaller candidate sets by introducing a first parameter (e.g., phase information) to divide the constellation points. This allows the detector to search through multiple reduced subsets rather than the entire constellation, significantly lowering computational complexity from O(ML) to O(M×L/K) where K is the number of segments, while still maintaining near-optimal detection performance.
2Device complexity
If sphere decoding is used to reduce complexity, then computational complexity is reduced, but candidate set completeness is lost causing performance loss
Solution Approach 1:
The patent employs dynamic candidate set generation where the search parameters and candidate sets are adaptively adjusted based on the specific channel conditions and received signal characteristics. This dynamic approach ensures that the candidate sets remain complete and representative of the true signal space, preventing performance loss while maintaining reduced complexity through selective searching.
3Productivity
If 256-QAM modulation is adopted to increase throughput, then spectral efficiency is improved, but ML detector complexity increases dramatically
Solution Approach 1:
For high-order modulations like 256-QAM, the patent applies segmentation by dividing the large constellation into multiple smaller subsets based on a first parameter (such as phase or amplitude groups). This reduces the effective search space from 256 points to manageable subsets, making real-time implementation feasible while maintaining the high spectral efficiency benefits of 256-QAM.
Solution Approach 2:
The patent performs preliminary classification of constellation points into candidate sets before the actual detection process. By pre-organizing the 256-QAM constellation points into structured subsets based on their properties, the system prepares the search space in advance, enabling faster and more efficient detection without sacrificing the high data rate capabilities of 256-QAM modulation.
4Speed
If candidate sets are defined in pre-transferred space, then search efficiency is improved, but complementary hypotheses are excluded causing performance loss
Solution Approach 1:
The patent introduces an additional dimension (first parameter) to the candidate set definition, creating a multi-dimensional search space instead of relying solely on the traditional pre-transferred space. This dimensional extension allows the candidate sets to include complementary hypotheses that would otherwise be excluded, improving detection performance while maintaining search efficiency through the structured organization of the expanded search space.
Data Source
AI summary
Methods, apparatuses, and systems are provided for generating a candidate search set for ML detection of 2n-QAM signals transmitted on two or more MIMO spatial streams. A method includes estimating an initial solution yq for a received 2n-QAM symbol value b0b1 . . . bn-1, wherein all possible 2n-QAM symbol values are Gray-mapped constellation points; and performing an iteration for each hypothetical value of each bit position i of the initial solution yq, wherein each iteration comprises: determining a search center as: if ith bit of the initial solution equals the hypothetical value assumed for the current iteration, the initial solution yq; or if ith bit of the initial solution does not equal the hypothetical value assumed for the current iteration, a mirror constellation point yqc to the initial solution yq; and searching outward from the determined search center for candidate constellation points.