Sidelink Model Sharing for Reliable UE Peer-to-Peer Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The distribution of machine learning models to user equipment (UEs) using volunteer terminals leads to issues such as increased radio resource usage, server load, delay, battery consumption, and system instability due to volunteer terminal movement or failure.
Innovation Solution
A radio communication system enables peer-to-peer networking among terminals using sidelink communication to share machine learning models, allowing volunteer terminals to distribute models directly to other UEs, reducing load and ensuring continuity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If volunteer terminals are used to distribute models via sidelink communication, then server load and radio resource usage are reduced, but system reliability deteriorates due to volunteer terminal movement or failure
Solution Approach 1:
The base station pre-notifies other terminals about the volunteer terminal's model distribution capability before actual distribution occurs. This advance notification ensures that if the volunteer terminal becomes unavailable, alternative terminals are already identified and ready to take over, preventing system failure and maintaining continuity.
Solution Approach 2:
The base station acts as an intermediary that coordinates between volunteer terminals and other terminals. It manages the notification process and facilitates the selection of alternative volunteers, ensuring reliable model distribution even when original volunteer terminals move or fail.
2Loss of time
If volunteer terminals distribute models to all other terminals, then individual terminal download time is reduced, but volunteer terminal battery consumption increases
Solution Approach 1:
Instead of requiring volunteer terminals to distribute models to all terminals uniformly, the system allows different terminals to acquire models through different paths: some receive from volunteer terminals via sidelink, others receive directly from the server. This localized optimization reduces the burden on volunteer terminals while maintaining fast distribution for those who benefit most.
3Quantity of substance
If volunteer terminals are selected for model distribution, then radio resource usage between base station and terminals is reduced, but discovery and selection complexity increases
Solution Approach 1:
Terminals autonomously determine their own volunteer status and distribution capabilities without requiring complex centralized discovery protocols. The base station simply notifies terminals of volunteer identities, and the system self-organizes the distribution process, reducing both radio resource usage and operational complexity.
Data Source
Figure 1
Figure 2
Figure 3(A)~3(B)
AI summary
A radio communication system (100) includes a plurality of terminals (110A to 110E), a base station (130) configured to perform radio communication with at least one volunteer terminal (110A, 110B) included in the plurality of terminals (110A to 110E), and an application server (150) configured to perform communication with the base station (130). The at least one volunteer terminal (110A, 110B) is configured to acquire a model to be used for machine learning from the application server (150) via the base station (130), and notify other terminals (110A to 110E) of the model. The plurality of terminals (110A to 110E) are configured to construct a peer-to-peer network in response to the notification by using inter-terminal direct communication through sidelink communication to share the model via the peer-to-peer network.