Distributed MIMO Power Allocation via Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed multiple input multiple output (MIMO) systems face challenges in reducing fronthaul overhead and latency due to the need for frequent updates of short-term channel state information, leading to performance degradation in beamforming and power allocation.
Innovation Solution
A power cooperative learning-based method that utilizes deep neural networks to normalize and transmit long-term local channel state information, enabling decentralized power allocation and reducing fronthaul overhead by using preconfigured DNNs for power determination and message generation, allowing for accurate beamforming vector calculation in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If instantaneous local channel information is transmitted over the fronthaul, then the CPU can compute optimal beamforming vectors, but fronthaul overhead and transmission latency increase significantly
Solution Approach 1:
The patent extracts only the essential statistical channel information (channel covariance matrix) from the complete instantaneous channel state information, transmitting only this extracted component over the fronthaul. This reduces the quantity of transmitted data while retaining sufficient information for beamforming computation, directly resolving the contradiction between information accuracy and fronthaul overhead.
Solution Approach 2:
The patent performs preliminary computation at each dAP to calculate local channel covariance matrices before transmission to the CPU. By pre-processing the channel information locally and extracting only the statistical characteristics, the system reduces the data volume requiring fronthaul transmission while maintaining the capability for optimal beamforming computation at the CPU.
2Reliability
If complete local channel information is transmitted frequently, then beamforming performance is maintained, but transmission latency increases
Solution Approach 1:
The patent extracts statistical channel characteristics (covariance matrices) that change more slowly than instantaneous channel states. By transmitting these extracted statistical features at lower rates, the system maintains beamforming performance while reducing transmission latency, as the covariance matrices can be updated less frequently without significantly degrading system performance.
Solution Approach 2:
The patent implements periodic updates of channel covariance matrices at dAPs and selective transmission to the CPU based on change detection. Instead of continuous transmission, the system periodically computes and transmits updated covariance information only when significant changes occur, thereby reducing transmission latency while maintaining beamforming reliability.
3Productivity
If the CPU performs complex power allocation optimization calculations, then system performance is optimized, but computation time and fronthaul capacity requirements increase
Solution Approach 1:
The patent segments the power allocation computation into two parts: (1) dAPs independently compute local channel covariance matrices and transmit them to the CPU, and (2) the CPU performs centralized power allocation optimization using these covariance matrices. This segmentation allows distributed preprocessing to reduce fronthaul overhead while maintaining centralized optimization capability, balancing computation complexity and system performance.
Solution Approach 2:
The patent transforms the complex instantaneous channel state information into simplified statistical parameters (covariance matrices) that capture the essential channel characteristics. By changing the representation from full channel matrices to covariance parameters, the computation complexity at the CPU is reduced while still enabling effective power allocation optimization, as the covariance matrices contain sufficient information for beamforming design.
Data Source
AI summary
A method for power allocation in a dAP may comprise: when a change cycle of a transmit power determination vector arrives, generating an uplink message including long-term CSI, the uplink message being normalized such that the long-term local CSI becomes a value within a preconfigured limit range; transmitting the uplink message to a central processing unit through a fronthaul; receiving a downlink message vector for power allocation from the central processing unit through the fronthaul; generating decentralized determination information using the downlink message vector; and extracting a transmit power determination vector based on the decentralized determination information.