O-RAN Radio Resource Management Using Reinforcement Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing radio resource management (RRM) systems in O-RAN networks face challenges in efficiently allocating resources to meet diverse QoS requirements and network slicing demands, leading to suboptimal performance and interference issues.

Innovation Solution

Implementing machine learning techniques, specifically reinforcement learning, to optimize RRM policies by using Q-learning and Deep Q Neural Networks (DQN) to dynamically allocate radio resources based on real-time feedback and network conditions, incorporating QoS metrics and slice-aware scheduling to enhance resource utilization and interference management.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional RRM algorithms are used for resource allocation, then the system is simple to implement, but the network performance and resource utilization are suboptimal

Engineering Contradiction:
Improvenetwork performanceVSAvoidRRM algorithm complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent replaces traditional mechanical RRM algorithms (MAC scheduler, interference coordination methods) with machine learning-based algorithms. The ML model learns optimal resource allocation strategies from historical data and network conditions, substituting rule-based mechanical decision-making with intelligent adaptive algorithms that improve network performance while managing complexity through automated learning.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent dynamically adjusts RRM parameters (such as scheduling weights, power allocation, resource block assignment) based on machine learning predictions and real-time network conditions. Instead of fixed parameter values, the system continuously optimizes parameters to maximize network performance, allowing adaptive response to changing traffic patterns and interference conditions.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If machine learning techniques are implemented to optimize RRM, then resource allocation efficiency improves, but system complexity increases

Engineering Contradiction:
Improveresource allocation efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The machine learning-based RRM system performs self-optimization by automatically learning from network data and adjusting resource allocation without manual intervention. The system serves itself by continuously training models on historical performance data and adapting to new conditions, reducing the need for complex manual configuration and optimization while improving resource allocation efficiency autonomously.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent implements feedback loops where network performance metrics are continuously collected, analyzed by the ML model, and used to adjust resource allocation decisions. The system monitors outcomes of previous allocations and uses this feedback to refine future decisions, creating a closed-loop control system that improves efficiency while managing complexity through data-driven adaptation.

Inventive Principle:
Principle #23Feedback

3Reliability

If conventional RRM methods are used, then the system is easier to operate, but QoS requirements and interference management are not adequately met

Engineering Contradiction:
ImproveQoS fulfillmentVSAvoidsystem operation simplicity
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The patent transitions from static RRM configurations to dynamic adaptive resource allocation using machine learning. The system continuously adjusts resource allocation based on real-time network conditions, traffic patterns, and QoS requirements. This dynamic approach enables the system to automatically adapt to changing demands and interference conditions, improving QoS fulfillment without requiring manual reconfiguration.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The machine learning model performs preliminary analysis of network conditions and traffic patterns to predict future resource allocation needs. By anticipating QoS requirements and potential interference issues before they occur, the system can proactively adjust resource allocation to prevent QoS violations, improving reliability while maintaining operational simplicity through automated prediction and prevention.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4648463A1Optimizing radio resource management in o-ran networks using machine learning techniques
Publication Date: 2025.11.12 MAVENIR SYST INC
  • EP4648463A1 patent drawingFigure 1A
  • EP4648463A1 patent drawingFigure 1B
  • EP4648463A1 patent drawingFigure 2

AI summary

A system and method for dynamically determining optimal values of various radio resource management (RRM) parameters used for RRM to meet various performance objectives such that RRM parameters are selected and dynamically adapted using a Radio Resource Management - MultiObjective (RRM-MO) optimization module adapted to optimize and dynamically adjust the RRM parameters.