Hierarchical Federated Learning Handover for Stable Model Aggregation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In conventional federated learning networks, mobility and changes in wireless channels lead to handover challenges, causing disruptions and inefficiencies due to limited base station coverage, communication resource constraints, and variable service quality, which affect model aggregation and convergence.

Innovation Solution

A hierarchical federated learning network structure with intermediate nodes and a global node, where handover decisions are made based on predefined conditions involving model aggregation and service time, ensuring seamless transitions and reduced resource load by optimizing handover timing and resource allocation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If handover is performed frequently to maintain connection quality, then communication reliability is improved, but model aggregation convergence deteriorates due to interruptions

Engineering Contradiction:

Improvecommunication reliabilityVSAvoidmodel aggregation convergence

Core Design Contradiction:

ReliabilityVSStability of the object's composition

Solution Approach 1:

The system performs preliminary evaluation of handover conditions by comparing remaining service time against model aggregation time before executing handover. This preliminary action prevents premature handovers that would interrupt model aggregation, thereby maintaining convergence while ensuring communication reliability when conditions permit.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The handover decision mechanism dynamically adapts to the current federated learning state by adjusting handover timing based on real-time comparison of service time and aggregation time parameters. This dynamic approach allows the system to optimize between communication reliability and model convergence under varying conditions.

Inventive Principle:

Principle #15Dynamics

2Productivity

If base station coverage is expanded to serve more user equipment, then system capacity is improved, but resource load and complexity increase

Engineering Contradiction:

Improvesystem capacityVSAvoidnetwork complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The network is segmented into multiple base stations with divided coverage areas, allowing the system to serve more user equipment through distributed architecture. Each base station handles a subset of UEs, reducing individual device complexity while maintaining overall system capacity through coordinated operation.

Inventive Principle:

Principle #1Segmentation

3Stability of the object's composition

If handover decisions are made based on traditional signal quality metrics, then connection stability is improved, but model aggregation efficiency deteriorates due to timing mismatches

Engineering Contradiction:

Improveconnection stabilityVSAvoidmodel aggregation efficiency

Core Design Contradiction:

Stability of the object's compositionVSProductivity

Solution Approach 1:

The handover decision parameters are changed from traditional signal quality metrics to a composite evaluation including remaining service time and model aggregation time. This parameter change aligns handover timing with federated learning phases, improving both connection stability and aggregation efficiency by preventing handovers during critical model transfer periods.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20260032535A1Device, method and medium for handover in a hierarchical federated learning network

Publication Date: 2026.01.29 SONY GROUP CORP

AI summary

The present disclosure relates to device, method and medium for handover in a hierarchical federated learning network. An electronic device for federated learning at a network, comprising processing circuitry configured to: determine a model aggregation time and a remaining service time for a user equipment, wherein the user equipment is directly connected to the network or indirectly connected to the network via an intermediate node; make a handover decision for the user equipment in a case where the model aggregation time and the remaining service time meet a predefined condition; and transmit the handover decision for the user equipment.