Multi-Layer Agent Learning for Real-Time IoT Data Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing communication networks face challenges in managing and processing the increasing volume of data from IoT devices, particularly in handling real-time and non-real-time data efficiently, which conventional methods struggle to address effectively.

Innovation Solution

A multi-layered autonomous distributed system is implemented, comprising cloud, fog, and edge computing layers, with agents at each layer performing cooperative learning to process and manage data efficiently, ensuring real-time, intermediate, and long-term analysis and control, thereby stabilizing the network.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a multi-layered autonomous distributed system is implemented with cloud, fog, and edge computing layers, then data processing efficiency and real-time responsiveness are improved, but device complexity increases

Engineering Contradiction:
Improvedata processing efficiencyVSAvoidsystem structure complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system divides data processing into three distinct layers: edge computing layer (third layer) for real-time processing of time-sensitive data, fog computing layer (second layer) for intermediate processing and analysis, and cloud computing layer (first layer) for long-term storage and non-real-time analysis. Each layer handles specific types of data processing tasks, enabling efficient data management while maintaining a structured and manageable system architecture.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If agents at each layer perform cooperative learning to process and manage data, then adaptability to environmental changes is improved, but difficulty of detecting and measuring increases

Engineering Contradiction:
Improveadaptability to environmental changesVSAvoidlearning process monitoring
Core Design Contradiction:
Adaptability or versatilityVSDifficulty of detecting and measuring

Solution Approach 1:

Agents at each layer execute learning through cooperative interactions and provide feedback to adjacent layers. The edge layer agents feed processed information to fog layer agents, which in turn provide feedback to cloud layer agents. This multi-layer feedback mechanism enables continuous adaptation to environmental changes while maintaining measurable performance metrics at each layer through standardized information exchange protocols.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If the system processes large volumes of data from IoT devices across multiple layers, then measurement precision is improved, but loss of time in data transmission increases

Engineering Contradiction:
Improvedata analysis accuracyVSAvoiddata transmission time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The edge computing layer performs preliminary processing and filtering of data from IoT devices before transmitting to higher layers. Time-sensitive processing is executed at the edge layer, while less time-critical analysis is performed at fog and cloud layers. This preliminary action at the edge reduces transmission time for critical data while maintaining high measurement precision through distributed processing.

Inventive Principle:
Principle #10Preliminary action

4Loss of information

If cloud computing is used for long-term data storage and analysis, then loss of information is reduced, but use of energy increases

Engineering Contradiction:
Improvedata storage retentionVSAvoidcomputational energy consumption
Core Design Contradiction:
Loss of informationVSUse of energy by moving object

Solution Approach 1:

The system segments data storage and processing tasks across three layers: edge layer for immediate local storage, fog layer for regional storage and intermediate analysis, and cloud layer for long-term centralized storage. This segmentation reduces the energy burden on any single layer by distributing computational and storage loads, while ensuring comprehensive data retention through multi-layer redundancy.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4502879B1system
Publication Date: 2026.02.04 SOFTBANK CORPORATION
  • EP4502879B1 patent drawingFigure 1
  • EP4502879B1 patent drawingFigure 2
  • EP4502879B1 patent drawingFigure 3

AI summary

There is provided a system including: a first layer agent of a first layer, and a plurality of second layer agents of a second layer which is lower than the first layer, the first layer agent and the plurality of second layer agents being arranged in a network, in which each of the plurality of second layer agents uses collected information to execute learning in cooperation with another second layer agent, and transmit, to the first layer agent, information generated by using a learning result, and the first layer agent executes learning by using the information received from the plurality of second layer agents.