Wireless Base Station Neural Network Splitting for 5G Delay Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing 5G wireless communication systems face challenges in minimizing network delay, especially when handling AI applications that require efficient processing and low latency.
Innovation Solution
A wireless base station is designed to split neural network models into multiple layers and distribute the arithmetic processing across different communication devices and the base station itself, based on acquired profile information and quality of service parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If neural network model processing is centralized in the wireless base station or terminal, then processing capability is sufficient, but network delay increases
Solution Approach 1:
The neural network model is segmented into multiple layers, with the splitting unit dividing the model into first neural network model layers processed by the terminal and second neural network model layers processed by the wireless base station. This segmentation enables distributed processing that reduces network delay by performing computations locally rather than centrally, while managing complexity through automated splitting point determination based on profile information.
2Loss of time
If AI processing is performed locally at the terminal, then network delay is reduced, but processing capability and accuracy may be insufficient
Solution Approach 1:
The solution moves from a single-location processing model to a distributed multi-location processing model across the terminal and base station. By determining optimal splitting points based on profile information containing layer characteristics, the system achieves both low latency through local processing and high accuracy through base station processing of complex layers, effectively adding a spatial dimension to the processing architecture.
3Reliability
If the entire neural network model is processed at the wireless base station, then processing accuracy is maintained, but network load and delay increase
Solution Approach 1:
The terminal extracts and processes specific layers of the neural network model locally, taking out the first neural network model layers from the complete model and processing them at the edge. This extraction reduces network load and delay by eliminating the need to transmit all intermediate computation data to the base station, while the base station retains processing of the second layers to maintain overall accuracy.
4Loss of time
If neural network model layers are distributed for processing, then network delay is reduced, but determining optimal splitting points becomes complex
Solution Approach 1:
Profile information containing characteristics of each neural network model layer is prepared in advance before the actual processing occurs. The splitting unit uses this pre-prepared profile information to quickly determine optimal splitting points without complex real-time analysis, thereby reducing the difficulty of splitting point determination while achieving low-latency distributed processing.
Data Source
AI summary
An object is to improve network delay. A wireless base station according to an embodiment of the present disclosure is a wireless base station capable of communicating with a first communication device and a second communication device, and includes: a splitting unit configured to acquire first profile information corresponding to one or more neural network models, and determine a splitting point for split of multiple layers constituting the neural network model on the basis of the first profile information; and a control unit configured to set, in the first communication device, arithmetic processing of a first neural network model generated by splitting the neural network model at the splitting point, and set, in the second communication device or the wireless base station, arithmetic processing of a second neural network model generated by splitting the neural network model at the splitting point.