Wireless Terminal Policy Sharing for Stable Distributed Transmission Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Distributed resource management in wireless networks faces challenges such as stability and convergence issues, especially when packet delivery with low delay is critical, and existing AI-based solutions struggle with large state spaces and inter-agent behavior prediction.

Innovation Solution

A wireless terminal adjusts a local transmission policy using reinforcement learning, exchanging information with neighboring terminals via D2D links to stabilize learning and achieve efficient resource allocation through a common global policy, leveraging neural networks for decentralized decision-making.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If distributed resource management is implemented where each device decides its own transmission parameters based on local observations, then the complexity of centralized control is reduced and scalability is improved, but stability and convergence issues arise leading to poor system performance

Engineering Contradiction:
Improvecentralized control complexityVSAvoidsystem performance stability
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent combines centralized policy coordination with distributed execution. Multiple wireless devices share a common global transmission policy that is updated through collaborative learning, merging individual device decisions into a coordinated system-wide policy that ensures stability while maintaining distributed operation

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent implements feedback mechanisms where devices exchange information about their local policy adjustments and observations with neighboring devices. This feedback loop enables the global policy to adapt and converge towards optimal performance while maintaining distributed decision-making

Inventive Principle:
Principle #23Feedback

2Adaptability or versatility

If distributed training is implemented on each wireless device, then autonomy and scalability are improved, but stability issues arise and lead to unfair medium utilization

Engineering Contradiction:
Improvedevice autonomyVSAvoidlearning stability
Core Design Contradiction:
Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent creates equipotentiality by having all devices operate under the same global transmission policy. This ensures fair medium utilization as all devices follow identical policy rules, preventing any single device from gaining unfair advantage while maintaining distributed autonomy

Inventive Principle:
Principle #12Equipotentiality

Solution Approach 2:

The patent segments the learning process into local policy adjustment at each device and global policy coordination through information exchange. This segmentation allows devices to maintain autonomy in local decisions while ensuring overall stability through shared global policy updates

Inventive Principle:
Principle #1Segmentation

3Adaptability or versatility

If reinforcement learning is used for distributed resource management, then adaptability to dynamic environment is improved, but the curse of dimensionality limits application to models with small number of states

Engineering Contradiction:
Improveenvironment adaptabilityVSAvoidstate space complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent combines reinforcement learning with a shared global policy framework. By merging individual device learning experiences into a common policy, the system can handle large state spaces that would be intractable for single devices, enabling adaptability to complex dynamic environments

Inventive Principle:
Principle #5Merging (Combining)

4Ease of operation

If devices learn to predict decisions of other agents in distributed RL, then strategic decision-making is improved, but convergence becomes problematic

Engineering Contradiction:
Improvestrategic decision capabilityVSAvoidconvergence stability
Core Design Contradiction:
Ease of operationVSStability of the object's composition

Solution Approach 1:

The patent introduces a global transmission policy as an intermediary that mediates between individual device decisions. This intermediary coordinates strategic interactions by providing a common framework that all devices follow, enabling strategic decision-making while ensuring convergence through centralized policy updates

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4305823B1Devices and methods for collaborative learning of a transmission policy in wireless networks
Publication Date: 2026.04.08 HUAWEI TECH CO LTD
  • EP4305823B1 patent drawingFigure 1
  • EP4305823B1 patent drawingFigure 2
  • EP4305823B1 patent drawingFigure 3

AI summary

A wireless terminal (120) comprises a communication interface (123) adapted to transmit data to an access point (110) and a processing circuitry (121) adapted to select for a plurality of data transmissions to the access point (110) a respective data transmission configuration based on an global transmission policy and information about a respective first transmission state of the wireless terminal (120). The processing circuitry (121) is further adapted to determine for the respective data transmission a reward for the data transmission configuration based on the data transmission configuration and information about a second transmission state of the wireless terminal (120), and to adjust a local transmission policy based on one or more of the plurality of selected data transmission configurations and the respective rewards. The communication interface (123) is adapted to exchange information about the adjusted local transmission policy with further wireless terminals (130) for adjusting the global transmission policy.