Traffic-Prediction Routing Selection for Reusable Network Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for network management are mechanism-specific and difficult to reuse, lacking flexibility in adapting to diverse network control mechanisms, and require specialized methods for each network function, such as routing and resource management.
Innovation Solution
A routing decision method and system based on traffic prediction, utilizing a machine learning orchestrator to train a routing selection model through data collection, feature extraction, and classification, enabling plug-and-play integration with existing networks for optimized routing and load balancing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are embedded into specific network control mechanisms, then accurate traffic prediction and optimized routing decisions are achieved, but the system becomes mechanism-specific and difficult to reuse across different network functions
Solution Approach 1:
The patent applies universality by designing a machine learning orchestrator that can handle multiple network control mechanisms (routing, mobility management, resource management, VNF orchestration) through a unified architecture. The orchestrator uses a common feature extraction framework and model training pipeline that can be applied across different network functions, making the system reusable while maintaining specialized performance for each function.
Solution Approach 2:
The patent segments the machine learning system into modular components: data collection module, feature extraction module, model training module, and deployment module. This segmentation allows the same modular framework to be reused across different network functions while each function can have its own specialized data and models, resolving the contradiction between specialization and reusability.
2Reliability
If specialized machine learning methods are built for each network control mechanism, then optimal routing decisions are achieved, but the device complexity and implementation cost increase
Solution Approach 1:
The patent merges multiple specialized machine learning methods into a unified machine learning orchestrator that handles routing, mobility management, resource management, and VNF orchestration through a single system. This consolidation reduces implementation complexity while maintaining the specialized capabilities needed for each network function through a common framework.
Solution Approach 2:
The machine learning orchestrator acts as an intermediary layer between raw network data and specialized control mechanisms. It provides a standardized interface for data collection, feature extraction, and model deployment, simplifying the implementation process while enabling specialized routing decisions through its mediation capabilities.
3Adaptability or versatility
If a flexible machine learning system with plug-and-play integration is implemented, then adaptability to various network scenarios is improved, but network scheduling delays and implementation costs are reduced
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models offline using historical network data before deployment. The models are prepared in advance and can be quickly deployed to production environments, reducing real-time scheduling delays while maintaining high adaptability to various network scenarios through pre-computed predictions.
Solution Approach 2:
The patent uses copying by creating reusable model templates and feature extraction pipelines that can be replicated across different network functions and scenarios. Once a model is trained for one function, the same framework can be copied and adapted for other functions, reducing implementation time and enabling rapid deployment without sacrificing adaptability.
Data Source
AI summary
Disclosed are a routing decision method and system based on traffic prediction. The method is used for obtaining each data link to be analyzed between a starting point and an ending point of a direct connection or an indirect connection for forwarding data packets to be forwarded in a target network area, and each node respectively included in each data link to be analyzed; when the target network area receives the data packets to be forwarded, training and obtaining a routing selection model corresponding to said packets for each data link to be analyzed for the direct connection or the indirect connection between the starting point and the ending point in the target network area, respectively; and applying the routing selection model to obtain an optimal forwarding path corresponding to said data packets, and forwarding said data packets.