Beam Confidence Reporting for AI-Based Wireless Resource Management

The confidence information is segmented into different reporting formats and levels based on the network's needs and the terminal's capabilities. The patent defines specific reporting configurations where terminals can report confidence information at different granularities (e.g., per beam, per cell, or aggregated), allowing the system to manage complexity by selecting appropriate segmentation levels rather than always reporting all details.

The reporting mechanism is made dynamic through configurable parameters that allow the network to adjust the reporting requirements based on current system conditions. The patent describes how the network can dynamically configure which beams require confidence reporting, the level of detail required, and when reporting should occur, enabling the system to adapt complexity levels to actual needs rather than maintaining fixed high complexity.

The patent implements feedback mechanisms where the terminal reports confidence information about AI/ML predictions to the network, and the network provides feedback configurations to the terminal. This feedback loop allows the system to monitor prediction quality and adjust beam selection strategies accordingly, mitigating the impact of prediction errors by continuously learning from actual system performance.

The system changes parameters related to beam selection based on confidence levels. When confidence information indicates high prediction accuracy, the system can rely more heavily on AI/ML outputs for beam selection. When confidence levels are low, the system can switch to alternative selection criteria or request additional information, thereby adapting reliability to the specific prediction context rather than maintaining fixed reliability expectations.