US-12628019-B2 - Retrieval of trained ML model from UE

Abstract

It is provided an apparatus comprising: one or more processors, and memory storing instructions that, when executed by the one or more processors, cause the apparatus to: monitor if a trained model is available at a terminal; inform a network that the trained model is available at the terminal if the trained model is available at the terminal.

Inventors

• Malgorzata Tomala
• Cinzia Sartori
• Hakon Helmers
• Anna Pantelidou

Assignees

• NOKIA TECHNOLOGIES OY

Dates

Publication Date

20260512

Application Date

20200803

Claims (4)

1. 1 . An apparatus comprising: one or more processors, and memory storing instructions that, when executed by the one or more processors, cause the apparatus to: monitor if an information is received that a trained model is available at a terminal; request the terminal to provide the trained model if the information is received from the terminal; check whether the information received from the terminal comprises an actual value of a first parameter regarding the training of the trained model is received; decide on a time when to request the terminal to provide the one or more parameters of the trained model based on the actual value of the first parameter; inhibit the requesting the terminal to provide the trained model earlier than at the decided time; check whether the information received from the terminal comprises an actual value of a second parameter regarding the training of the trained model; decide a priority for retrieving the trained model from the terminal based on the actual value of the second parameter; and configure a radio bearer for the terminal to provide the trained model according to the priority.
2. 2 . The apparatus according to claim 1 , wherein the first parameter and the second parameter, respectively, comprises at least one of a maturity level of the trained model and a status of the trained model.
3. 3 . The apparatus according to claim 1 , wherein the instructions, when executed by the one or more processors, further cause the apparatus to: provide a target value of the first parameter and a target value of the second parameter, respectively, to the terminal via a radio resource control command.
4. 4 . The apparatus according to claim 3 , wherein the target value of the first parameter and the target value of the second parameter, respectively, is received from an operation and maintenance center.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application is a national phase entry of International Application No. PCT/EP2020/071757, filed Aug. 3, 2020, the contents of which are hereby incorporated by reference in their entirety. FIELD OF THE INVENTION The present disclosure relates to retrieval of one or more trained ML models from UE. ABBREVIATIONS 3GPP 3rd Generation Partnership Project3G/4G/5G 3rd/4th/5th GenerationAI Artificial IntelligenceAMF Access and Mobility Management FunctionCN Core NetworkgNB 5G base stationIE Information ElementLTE Long-Term EvolutionMDT Minimization of Drive TestsML Machine LearningMR-DC Multi RAT— Dual ConnectivityMTC Machine-Type CommunicationNF Network FunctionNG-RAN Next Generation RANNR New RadioO&M Operation & MaintenanceRAN Radio Access NetworkRAT Radio Access TechnologyRel ReleaseRRC Radio Resource ControlRRM Radio Resource ManagementSA System ArchitectureSON Self-Optimizing NetworksSRB Signaling Radio BearerTS Technical SpecificationUE User EquipmentUL Uplink BACKGROUND OF THE INVENTION 5G networks are expected to meet the challenges of joint optimizations of an ever-increasing number of performance measures. In addition, 5G brings a complicated system design and optimization issues introduced by NR architecture/features including MR-DC, beamforming, etc., which render traditional human-machine interaction slow, error-prone, expensive, and cumbersome to handle these challenges. 5G evolution drives the need to study use cases and to propose potential service requirements for 5G system support of Artificial Intelligence (AI)/Machine Learning (ML). The agreed 3GPP SA1 Study Item in S1-193606 describes targeted objectives and emphasizes that ML and AI will engage concrete 5G network entities and infrastructure. The way of developing machine learning processes and models already assumes that the 5G traffic and end-user's device will take part in ML model training. The book by Shai Shalev-Shwartz and Shai Ben-David, “Understanding Machine Learning: From Theory to Algorithms”, Cambridge University Press, 2014, describes ML as follows: “As an interdisciplinary field, machine learning shares common threads with the mathematical fields of statistics, information theory, game theory, and optimization. It is naturally a subfield of computer science, as our goal is to program machines so that they will learn. In a sense, machine learning can be viewed as a branch of AI (Artificial Intelligence), since, after all, the ability to turn experience into expertise or to detect meaningful patterns in complex sensory data is a cornerstone of human (and animal) intelligence.”. Also, in this book, Machine Learning (ML) is defined as part of automated learning through which computers are programmed so that they can “learn” from input available to them. Learning is defined to be the process of converting experience into expertise or knowledge. The input to a learning algorithm is training data, representing experience, and the output is some expertise, which usually takes the form of another computer program that can perform some task. ML execution in 5G may require the end user to train an ML model (i.e. UE learns from the provided data in a training process). Meanwhile, 3GPP RAN discusses use cases that may be enabled via AI. They include energy saving, traffic steering, mobility optimization, load balancing, physical layer configuration optimization, etc. 3GPP Rel-16 defined 5G features under RAN-centric Data Collection mechanisms that addressed a number of network optimization related use cases. Under the Rel.16 RAN Data Collection umbrella, 5G standard supports SON and MDT reports [3GPP TS 37.320], providing the means for the operators to monitor and optimise performance of any kind of 5G deployments and related issues. MDT reports, by default, can be categorized per optimization use case (e.g. for radio coverage, the UE logs radio signal levels with positioning data; for mobility robustness, the UE records radio signal levels when radio link failure is detected, etc). Overall, MDT and SON enable data collection for both user performance and network performance. Training of ML models requires big amounts of data. Thus, SON/MDT feature constitutes an efficient method for collecting UE data. SON/MDT records of measurements, taken by an end user, are very good candidates for monitoring the predictions of mobile traffic/users with the use of ML models. A UE may have several trained ML models locally available. ML models training may be performed per use case or optimization problem. MDT is an existing framework allowing radio measurements retrieval. The data collection in Rel.16 for 5G inherits (from LTE) two types of MDT: Immediate MDT methods to deliver real-time measurements (Immediate MDT when e.g. results of measurements performed for typical RRM operations in RRC CONNECTED), and Logged MDT methods to deliver in non-real time, measurements results taken during the time the UE contex