AI-Based RAN QoE Management for Predictive Resource Allocation

An AI model is introduced as an intermediary component between the RAN node and QoE management decisions. The AI model processes limited RAN parameters and transforms them into comprehensive future scenario predictions, enabling adaptability without directly expanding the core RAN node structure. This mediator approach allows the system to handle complex future scenarios using current limited parameters.

The system performs preliminary QoE parameter collection and AI model training using historical data before actual future scenarios occur. By pre-training the AI model on past QoE data and patterns, the system prepares predictive capabilities in advance, enabling it to handle future scenarios adaptively without requiring real-time complex data collection during those future events.

The patent replaces traditional mechanical/algorithmic QoE checking methods with an AI-based predictive system. Instead of using conventional threshold-based or rule-based approaches to evaluate service fulfillment, the system substitutes these with machine learning models that can infer service quality from limited parameters, significantly improving reliability while abstracting away the complexity through the AI layer.

The system implements feedback mechanisms where QoE measurement results are continuously fed back into the AI model for retraining and refinement. This feedback loop allows the model to progressively improve its predictions based on actual observed QoE data, enhancing the effectiveness of service fulfillment checking over time while managing complexity through iterative improvement rather than requiring complex initial designs.

The AI model is trained in advance using historical QoE data collected during network operation. This preliminary training action is performed before the network needs to handle future scenarios, allowing the model to be ready for predictive tasks without requiring real-time training during critical future events. The training phase is completed beforehand, eliminating the need for time-consuming on-demand training when future scenarios arise.

The system utilizes its own historical operational data and QoE measurements to train the AI model autonomously. By self-training on its past performance data, the network system can improve its predictive capabilities without requiring external intervention or additional time for model development. The system serves its own training needs by leveraging existing internal data resources, reducing the need for separate time-consuming training operations.