ML-Assisted Sphere Precoding for Lower-Complexity MIMO
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MIMO systems face challenges in adapting to increasing data rates and complexity in signal processing, with existing precoding techniques offering suboptimal performance and high computational complexity.
Innovation Solution
A precoder that uses a machine learning algorithm to determine a search sphere radius and applies a sphere search-based sequential algorithm to find a perturbation vector, reducing computational complexity while maintaining optimal performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If non-linear precoding techniques (e.g., Sphere Encoder) are used to achieve optimal performance and full diversity order, then precoding performance is improved, but computational complexity increases exponentially with the number of antennas
Solution Approach 1:
The patent segments the precoding problem into two parts: a low-complexity linear precoding stage (using MRT, ZF, or MMSE) followed by a simplified non-linear optimization stage. This segmentation allows the system to achieve near-optimal performance while avoiding the exponential complexity of traditional Sphere Encoding by only performing the complex search on the residual interference cancellation rather than the entire precoding process
Solution Approach 2:
The patent changes the search space parameter by constraining the perturbation vector to have integer components within a bounded range. This parameter transformation converts the continuous optimization problem into a discrete search problem with finite candidates, enabling fixed-complexity implementation while maintaining near-optimal performance
2Productivity
If MIMO systems are designed to meet increasing data rate demands, then system capacity is improved, but signal processing complexity and energy consumption increase
Solution Approach 1:
The patent introduces dynamic adaptability by allowing the system to switch between different linear precoding techniques (MRT, ZF, MMSE) based on channel conditions and system requirements. This dynamic selection enables the system to optimize between capacity and complexity trade-offs in real-time, maintaining high system capacity while adapting processing complexity to actual needs
Solution Approach 2:
The patent applies preliminary linear precoding to pre-shape the transmitted signal and cancel dominant interference components before the final detection stage. This preliminary action reduces the complexity of subsequent processing by pre-processing the signal in a computationally efficient manner, allowing the system to handle higher data rates without proportionally increasing overall complexity
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A precoder (200) for precoding a vector of information symbols. The precoder (200) comprises: - a radius determination unit (201) configured to determine a search sphere radius, the determination of the search sphere radius comprising applying a machine learning algorithm to input data dependent on the vector of information symbols and on a predefined precoding matrix; - a sphere encoding unit (203) configured to determine a perturbation vector from lattice points found inside a spherical region by applying a sphere search-based sequential algorithm, the spherical region having as a radius the search sphere radius, and - a precoding unit (205) configured to precode the vector of information symbols using the perturbation vector and a precoding matrix.