Macro Base Station Coordinating Micro Stations for Federated Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In heterogeneous wireless communication networks, the efficiency of federated learning is hindered by poor channel quality due to geographical distance, differing data structures among terminals, reliance on core networks for data interaction, and inadequate handling of terminal mobility and model training data alignment.

Innovation Solution

The method involves a macro base station coordinating micro base stations to perform model training, aligning model results across different terminals, and managing terminal switching, ensuring continuous training participation and data alignment within the network, thereby improving network efficiency and model accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If terminals perform federated learning directly through the core network, then model training can be performed across the network, but channel quality deteriorates due to geographical distance and network reliance

Engineering Contradiction:

Improvefederated learning efficiencyVSAvoidchannel quality

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent introduces base stations as intermediary nodes between terminals and the core network. Terminals perform federated learning through base stations rather than directly through the core network, which improves channel quality by reducing transmission distance and avoiding core network reliance. The base stations aggregate model parameters and coordinate training processes locally, maintaining federated learning functionality while improving communication reliability.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If model training is performed across distributed terminals, then network coverage and participation are improved, but data structure differences among terminals cause model alignment difficulties

Engineering Contradiction:

Improveterminal participation coverageVSAvoidmodel training consistency

Core Design Contradiction:

Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The patent employs parameter transformation techniques to handle data structure differences among terminals. Different terminal data types (categorical, numerical, sequential) are converted to unified parameter representations that can be processed consistently. This allows diverse terminal data to be aligned to a common model framework, maintaining training consistency while accepting diverse terminal participation.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent creates a universal model training framework that can handle multiple data types and terminal configurations. The system design allows the same federated learning protocol to work across terminals with different data structures by implementing data-type-agnostic processing mechanisms, enabling broad terminal participation without sacrificing model alignment.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Ease of operation

If terminals move within the network coverage area, then network mobility and flexibility are improved, but terminal mobility management and training continuity become inadequate

Engineering Contradiction:

Improveterminal mobilityVSAvoidtraining participation continuity

Core Design Contradiction:

Ease of operationVSStability of the object's composition

Solution Approach 1:

The patent implements feedback mechanisms where terminals report their movement status and base stations track terminal locations continuously. This feedback loop enables the system to detect terminal mobility events and trigger appropriate responses such as reassigning base stations or adjusting training schedules, thereby maintaining training continuity despite terminal movement.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent introduces dynamic base station assignment and flexible training scheduling that adapts to terminal mobility. Instead of fixed assignments, the system dynamically adjusts which base station serves which terminal based on current location and training progress, allowing terminals to move freely while maintaining stable training participation through adaptive reconfiguration.

Inventive Principle:

Principle #15Dynamics

4Productivity

If micro base stations are distributed within macro base station coverage, then network efficiency and spectral efficiency are improved, but coordinating model learning across multiple base stations increases system complexity

Engineering Contradiction:

Improvenetwork efficiencyVSAvoidmodel learning coordination complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the federated learning coordination function into hierarchical layers: macro base stations handle high-level coordination and global model aggregation, while micro base stations handle local terminal management and parameter collection. This segmentation distributes the coordination complexity across multiple levels, allowing efficient model learning across distributed base stations without overwhelming any single node.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20240235954A1Model learning method, model learning apparatus, and storage medium

Publication Date: 2024.07.11 BEIJING XIAOMI MOBILE SOFTWARE CO LTD

AI summary

A method for model learning, the method includes in response to reception of a model training request sent by an operation administration and maintenance (OAM) entity, sending the model training request to a first number of micro base stations, where a communication coverage range of the first number of micro base stations being within a communication coverage range of the macro base station.