SMO Framework Optimizing O-RAN Energy Efficiency via AI/ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In open radio access networks (O-RAN), the trade-off between system performance and energy conservation leads to inferior overall network energy efficiency when deciding to switch off or on carriers and cells, as local energy-saving measures can increase overall network energy consumption.
Innovation Solution
A service management and orchestration (SMO) framework with a non-real-time radio intelligent controller (NRT-RIC) uses AI/ML techniques to collect and analyze data for predicting traffic, user mobility, and resource usage, allowing flexible configuration of carrier and cell switch off/on parameters to optimize overall network energy efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by stationary object
If local energy-saving measures (switching off carriers or cells) are implemented, then local energy consumption is reduced, but overall network energy efficiency deteriorates
Solution Approach 1:
The system implements a feedback mechanism where the nRT-RIC continuously monitors network traffic patterns, user mobility, and resource usage, then uses this information to dynamically adjust carrier and cell switching decisions. The AI/ML models process this feedback data to predict future network conditions and optimize energy-saving strategies at the network level rather than locally, resolving the contradiction between local energy savings and overall network efficiency
Solution Approach 2:
The nRT-RIC acts as an intermediary between local RAN elements (gNBs, carriers, cells) and the network-wide optimization goal. It collects data from multiple sources, processes it through AI/ML models, and generates coordinated switching decisions that consider overall network energy efficiency while enabling local energy-saving actions
2Loss of energy
If AI/ML models are used to predict traffic and resource usage, then carrier and cell switch off/on decisions are optimized, but system complexity increases
Solution Approach 1:
The system segments the complex AI/ML processing into distinct functional modules: data collection from RAN elements, data processing and feature extraction, AI/ML model inference for prediction, and decision generation for carrier/cell switching. This segmentation allows each module to be optimized independently and deployed in a distributed architecture, reducing overall system complexity while maintaining optimization capabilities
Solution Approach 2:
The nRT-RIC serves as an intermediary layer that handles the complexity of AI/ML model management, training, and inference. By centralizing these complex functions in the nRT-RIC rather than distributing them across multiple RAN elements, the system reduces device complexity at the network edge while maintaining network-wide optimization
Data Source
AI summary
Systems and methods for implementing an optimization of a carrier and/or cell switch off/on in an O-RAN by a SMO-framework, the method includes: collecting O1-related data providing O1 configurations required to perform cell and/or carrier switch off/on; based on the collected O1-related data, re-training of AI/ML model, deploying and activating one re-trained AI/ML model for inferring data providing O1 configurations required to perform the cell and/or carrier switch off/on within the O-RAN; monitoring the O1-related data providing O1 configurations required to perform the cell and/or carrier switch off/on; evaluating the O1-related data providing O1 configurations required to perform cell and/or carrier switch off/on; determining to generate O1 configuration data to prepare and execute the cell and/or carrier switch off/on and sending the O1 configuration data to prepare and execute the cell and/or carrier switch off/on to an E2 node; implementing the cell and/or carrier switch off/on within the O-RAN.