AI/ML Beam Failure Prediction for Reduced Reference Measurements

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing communication systems face inefficiencies in beam failure detection, leading to unnecessary beam re-establishment procedures due to reliance on traditional measurement methods, which can be resource-intensive and delay-sensitive.

Innovation Solution

Implementing an AI/ML model on user equipment (UE) for predicting beam failure, allowing the UE to skip reporting beam failure detection when the model predicts a high probability of failure, thereby enabling the selection of alternative beams without requiring continuous reference signal measurements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional reference signal measurement methods are used for beam failure detection, then measurement accuracy is maintained, but resource consumption increases and detection delays occur

Engineering Contradiction:

Improvebeam failure detection accuracyVSAvoidresource consumption

Core Design Contradiction:

Measurement precisionVSLoss of energy

Solution Approach 1:

The AI/ML model performs preliminary prediction of beam failure probability before actual beam failure occurs. By analyzing historical measurement data and current channel conditions, the model predicts potential beam failures in advance, allowing the system to proactively switch to alternative beams before communication quality degrades, thereby avoiding the need for continuous reference signal measurements while maintaining detection accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The AI/ML model is deployed locally on the user equipment, enabling self-service beam failure prediction without requiring continuous network-side reference signals. The model uses locally available data (historical measurements, channel state information) to autonomously predict beam failures, reducing dependency on network resources and minimizing energy consumption while maintaining detection capability

Inventive Principle:

Principle #25Self-service

2Reliability

If continuous reference signal measurements are performed for beam failure detection, then detection reliability is improved, but time delays increase

Engineering Contradiction:

Improvebeam failure detection reliabilityVSAvoiddetection delay

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The AI/ML model continuously learns from historical data and performs real-time prediction of beam failure probability. This preliminary action enables the system to identify impending beam failures before they actually occur, allowing proactive beam switching that eliminates detection delays while maintaining high reliability through the model's predictive capability rather than reactive measurement

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces the traditional mechanical measurement-based detection system with an AI/ML-based predictive system. Instead of relying on periodic reference signal measurements and threshold comparisons, the system uses machine learning models to predict beam failure probability, substituting computational intelligence for traditional signal processing mechanisms, thereby achieving both high reliability and real-time response without measurement delays

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Loss of energy

If AI/ML model prediction is used to skip beam failure reporting, then resource usage is optimized, but measurement precision may be reduced

Engineering Contradiction:

Improveresource usage efficiencyVSAvoidbeam failure detection precision

Core Design Contradiction:

Loss of energyVSMeasurement precision

Solution Approach 1:

The AI/ML model transforms multiple input parameters (historical RSRP measurements, channel state information, mobility patterns, interference conditions) into a single predictive output (beam failure probability). By changing the detection parameter from direct RSRP threshold comparison to predicted failure probability, the system achieves higher resource efficiency while maintaining or improving detection precision through the model's ability to synthesize multiple information sources

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The AI/ML model acts as an intermediary between raw measurement data and beam failure detection decisions. Instead of directly comparing RSRP values against thresholds, the model processes historical and current data through learning algorithms to generate predictive insights, serving as an intelligent mediator that enhances detection accuracy while reducing the need for continuous measurements, thereby optimizing resource usage without sacrificing precision

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20250234221A1Beam failure prediction detection based on ai/ml

Publication Date: 2025.07.17 NOKIA TECHNOLOGIES OY

AI summary

An apparatus including at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: determine an output of a model used to predict failure of a current beam used with the apparatus; and perform one of: determine to skip reporting of beam failure detection, in response to the output of the model used to predict failure of the current beam used with the apparatus being greater than or equal to a beam failure detection threshold, or determine to skip reporting of beam failure detection, in response to the output of the model used to predict failure of the current beam used with the apparatus being equal to one.