AI Model Synchronization in Radio Access Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI/machine learning microservices in radio access networks (RANs) face challenges in synchronizing, sharing, and automating AI models across different RAN instances, leading to inefficiencies and the need for manual intervention to maintain network performance.
Innovation Solution
A system that enables synchronization and sharing of pre-trained AI models between RANs through an AI model synchronization system, using an AI microservice adapter to adapt models for different protocols, and a configuration file to standardize interfaces and deployment mechanisms, allowing for autonomous network management and resource optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If AI models are manually deployed in each RAN instance, then deployment control is precise, but operational complexity and time consumption increase
Solution Approach 1:
The system enables autonomous AI model synchronization where RAN instances automatically discover, select, and deploy models without manual intervention. The synchronization mechanism self-manages the deployment process by comparing local and remote models, transferring only necessary updates, and applying changes automatically, thereby eliminating manual operations and reducing time loss.
Solution Approach 2:
The system implements feedback mechanisms where each RAN instance reports its current model status and performance metrics. This feedback enables the synchronization system to intelligently determine which models need updating, how to optimize transfer timing, and verify deployment success, thereby automating the deployment process and reducing manual intervention requirements.
2Adaptability or versatility
If AI models are customized for each RAN instance, then local optimization is achieved, but synchronization and sharing capability deteriorates
Solution Approach 1:
The system segments AI models into modular components that can be independently customized for local RAN optimization while maintaining a standardized core structure. This allows each RAN instance to adapt models to local conditions without losing the ability to share and synchronize customized versions with other instances, preserving both adaptability and information sharing capabilities.
Solution Approach 2:
The system implements dynamic model synchronization where AI models can evolve independently in each RAN instance based on local conditions, while automatically synchronizing updates across the network. This dynamic approach enables continuous adaptation to local requirements while maintaining shared knowledge, preventing information loss through selective synchronization of only changed model components.
3Reliability
If comprehensive AI model synchronization is implemented, then network performance improves, but system complexity increases
Solution Approach 1:
The system employs a universal synchronization framework that handles multiple functions including model discovery, version comparison, data transfer, error handling, and deployment verification through a single integrated mechanism. This multi-functional approach improves network performance through comprehensive synchronization while avoiding the complexity of multiple separate systems by consolidating functions into one unified platform.
Solution Approach 2:
The system performs preliminary actions by pre-configuring synchronization parameters, establishing communication protocols, and preparing model packaging structures before actual synchronization occurs. This preparation simplifies the real-time synchronization process and reduces operational complexity while ensuring reliable model deployment across RAN instances, as the heavy lifting is done in advance during system initialization.
Data Source
AI summary
Aspects of the subject disclosure may include, for example, receiving network-related information associated with a first RAN that includes a first RIC, obtaining, from an artificial intelligence (AI) model synchronization system associated with a second RAN, data relating to an AI model deployed by a second RIC of the second RAN, determining, based on the data relating to the AI model and the network-related information associated with the first RAN, that the AI model can be leveraged by the first RAN to improve network performance of the first RAN, performing synchronization with the AI model synchronization system to obtain the AI model, responsive to the determining that the AI model can be leveraged by the first RAN to improve the network performance of the first RAN, and causing the first RIC to deploy the AI model in the first RAN after the performing the synchronization. Other embodiments are disclosed.