Wireless Device Machine Learning for Adaptive Handover Decisions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing wireless communication networks face inefficiencies in managing handovers and cell reselections due to the use of static event criteria, leading to unnecessary reporting, increased processing costs, and potential radio link failures, particularly in 5G networks with diverse mobility patterns and varying signal strengths.

Innovation Solution

Implementing a machine learning model, such as a neural network, trained on historical data from wireless communication devices to dynamically adjust handover and cell reselection processes, replacing static thresholds with intelligent decision-making at the device level.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If static event criteria are used for handover decisions, then the network can maintain simple control logic, but unnecessary handovers occur and battery consumption increases

Engineering Contradiction:
Improvecontrol logic complexityVSAvoidbattery consumption
Core Design Contradiction:
Device complexityVSUse of energy by moving object

Solution Approach 1:

The patent applies dynamics by transitioning from static handover thresholds to dynamic machine learning models that adapt to varying mobility patterns. The ML model continuously learns from historical measurement data and adjusts handover decisions based on real-time conditions, thereby reducing unnecessary handovers and optimizing battery consumption without oversimplifying the control logic.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The invention changes the parameters used for handover decisions from fixed static thresholds to dynamic parameters generated by machine learning models. These parameters evolve based on learned mobility patterns and current network conditions, enabling more accurate handover predictions that reduce unnecessary measurements and signaling, thus conserving battery energy.

Inventive Principle:
Principle #35Parameter changes

2Ease of operation

If static thresholds are used for handover events, then the system operation is simple, but handover accuracy deteriorates leading to radio link failures

Engineering Contradiction:
Improvesystem operation simplicityVSAvoidhandover decision accuracy
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The patent implements self-service by enabling the system to automatically learn and optimize handover parameters through machine learning. The ML model autonomously analyzes historical measurement data, identifies mobility patterns, and generates optimized handover decisions without requiring manual configuration or intervention, thereby maintaining operational simplicity while significantly improving handover accuracy and reliability.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The invention incorporates feedback mechanisms where the ML model continuously receives measurement reports and handover outcomes, learns from these results, and refines its predictions. This closed-loop feedback system enables the network to adapt to changing conditions and improve handover accuracy over time while keeping the operational interface simple.

Inventive Principle:
Principle #23Feedback

3Reliability

If frequent measurements are performed to ensure accurate handover decisions, then handover reliability is improved, but network resource usage and processing costs increase

Engineering Contradiction:
Improvehandover decision accuracyVSAvoidnetwork resource efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies partial action by performing measurements selectively rather than continuously. The ML model predicts when measurements are likely to lead to handovers based on learned mobility patterns, and only triggers measurements in those specific scenarios. This partial measurement approach maintains handover reliability by focusing resources on critical decision points while reducing overall network resource consumption.

Inventive Principle:
Principle #16Partial or excessive action

4Measurement precision

If the network performs all handover evaluations centrally, then control accuracy is maintained, but network processing capacity is overwhelmed

Engineering Contradiction:
Improvehandover evaluation accuracyVSAvoidnetwork processing capacity
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent segments the handover evaluation process into two parts: the ML model runs locally on the UE to perform preliminary evaluations and filter candidates, while the network performs final verification and decision-making. This segmentation reduces the processing burden on the network by pre-filtering handover candidates at the device level, maintaining evaluation accuracy through collaborative processing.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3763148B1Managing communication in a wireless communications network
Publication Date: 2025.10.15 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3763148B1 patent drawingFigure 1
  • EP3763148B1 patent drawingFigure 2
  • EP3763148B1 patent drawingFigure 3

AI summary

Some embodiments herein relate to a method performed by a wireless communication device (10) for managing communication in a wireless communications network (1). The wireless communication device (10) obtains an indicator indicating a 5 model and one or more trained model parameters for the model, wherein the model is related to an event being one of the following events: a handover procedure, a cell reselection procedure, and a beam reselection procedure. The wireless communication device (10) further selects the model based on the obtained indicator. The wireless communication device (10) executes the selected model using the obtained one or more 10 trained model parameters; and triggers a process, being associated with the event, based on an output of the executed model.