ML-Based Scheduling Weight Optimization in O-RAN RIC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems face challenges in determining optimal weights for scheduling priority in cellular networks, particularly in policy-based performance management, due to competing considerations between maximizing cell throughput and meeting diverse QoS requirements, which are difficult to balance especially under varying traffic conditions and channel conditions.
Innovation Solution
Implementing a machine-learning-based method to dynamically determine suitable weights for scheduling priority at the Distributed Unit (DU)/Centralized Unit (CU) or RAN Intelligent Controller (RIC), incorporating parameters like GBR, packet delay budget, and proportional fairness to improve radio resource management and policy-based performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional static weight-based scheduling is used to manage radio resources, then system complexity is low and ease of operation is maintained, but the ability to adapt to varying traffic conditions and channel conditions is poor, resulting in suboptimal network performance
Solution Approach 1:
The patent implements dynamic weight adjustment for scheduling priorities based on real-time network conditions, traffic types, and QoS requirements. The system transitions from static pre-configured weights to dynamically computed weights that adapt to changing conditions, resolving the contradiction between adaptability and complexity through controlled dynamic behavior.
Solution Approach 2:
The system changes the parameters of scheduling weights based on multiple factors including traffic type (e.g., eMBB, uRLLC, mMTC), channel conditions, QoS parameters, and network load. By adjusting these parameters dynamically, the system achieves adaptability to varying conditions while maintaining manageable complexity through structured parameter management.
2Reliability
If multiple QoS parameters are considered simultaneously for scheduling decisions, then QoS satisfaction for diverse applications is improved, but the difficulty of detecting and measuring optimal weights increases
Solution Approach 1:
The patent introduces a machine learning model as an intermediary between the multiple QoS parameters and the scheduling weight determination. The ML model processes complex inputs including GBR, packet delay budget, proportional fairness requirements, and channel conditions to output optimized weights, thereby reducing the difficulty of detecting and measuring optimal weights while maintaining high QoS satisfaction.
Solution Approach 2:
The system implements feedback mechanisms where scheduling performance metrics are continuously monitored and used to refine weight optimization. The ML model learns from historical scheduling outcomes and adjusts weight recommendations accordingly, making the detection and measurement of optimal weights more reliable through iterative feedback loops.
3Productivity
If machine-learning-based dynamic weight determination is implemented, then network performance under varying conditions is improved and QoS for diverse applications is enhanced, but the device complexity and computational requirements increase
Solution Approach 1:
The patent segments the scheduling system into distinct functional components: a machine learning model for weight determination, a traditional scheduler for resource allocation, and a performance monitoring module. This segmentation allows the ML component to be developed and optimized independently, improving network performance while managing overall system complexity through modular architecture.
Solution Approach 2:
The machine learning model serves multiple functions simultaneously: it determines scheduling weights, adapts to different traffic types (eMBB, uRLLC, mMTC), satisfies diverse QoS requirements, and optimizes for multiple objectives (throughput, delay, fairness). This multi-functionality improves network performance across various scenarios while avoiding the need for separate specialized systems for each function.
Data Source
AI summary
A method for enhanced radio resource management of radio access network (RAN) based on machine-learning-based technique includes: deploying a trained machine-learning-based model trained on a plurality of RAN-related parameters for a selected user equipment (UE) on the RAN, including i) plurality of trained weights for determining scheduling priority of the selected UE, ii) a network operator policy influencing scheduling priority of the selected UE, and iii) at least one of Packet Delay Budget (PDB), a target guaranteed bit rate (GBR), and proportional fair (PF) metric of the selected UE; computing, at the RIC, a difference between two consecutive overall error functions for PDB, GBR, and PF calculated based on corresponding observed values at two consecutive sampling time points; and updating, at the RIC, the plurality of weights based on the difference between the first and second overall error functions.