Search

CN-122003844-A - Method for model packet signaling

CN122003844ACN 122003844 ACN122003844 ACN 122003844ACN-122003844-A

Abstract

A method of pre-configured AI/ML (artificial intelligence/machine learning) -based group model ID allocation is used in a wireless mobile communication system including a base station (e.g., gNB) and a mobile station (e.g., UE). If the AI/ML model is applied to a radio access network, signaling for model ID information exchange may be severely congested based on different model execution environments with different models. Thus, a model operation may be established between the network and the UE by assigning a group model ID.

Inventors

  • Jin Huchen
  • R. SHAH
  • A. ANDRE

Assignees

  • 欧摩威德国有限公司

Dates

Publication Date
20260508
Application Date
20241014
Priority Date
20231027

Claims (20)

  1. 1. A method of configuring group model identification for a two-sided model operation in a network comprising a plurality of Artificial Intelligence (AI)/Machine Learning (ML) models deployed on at least one User Equipment (UE) for use in the two-sided model operation, the method comprising the steps of: The packet standard is received and the packet data is received, Assigning a group model Identifier (ID) to each of the plurality of AI/ML models to generate a group model ID, wherein each group model ID specifically corresponds to an element of the group criteria, an The two-sided model operation is activated by activating models with the same group model ID via L1/L2/L3 signaling.
  2. 2. The method of claim 1, wherein the plurality of AI/ML models are deployed on a single UE of the at least one UE.
  3. 3. The method of claim 1 or 2, wherein the plurality of AI/ML models are distributed deployed on different ones of the at least one UE.
  4. 4. Method according to one of the preceding claims, characterized in that a group model ID of a model is assigned to a network side model, which is an AI/ML model co-operating with the model during operation of the two side model.
  5. 5. The method according to one of the preceding claims, characterized in that the grouping criterion is based on the following similarity: Preconfigured measurements of different combinations of model properties of the AI/ML model, and/or The type, structure, class, application, functionality or performance of the AI/ML model, and/or Attributes or metadata of the AI/ML model, Use cases or scenarios for ML signaling purposes.
  6. 6. The method according to one of the preceding claims, characterized in that a preconfigured mapping relation with index information between the AI/ML model and the group model ID is generated, which mapping relation is signaled via RRC signaling or system information to activate the two-sided model operation.
  7. 7. The method according to one of the preceding claims, characterized in that a plurality of UEs with ML encoders perform LCM operation in conjunction with a network-side ML decoder, wherein the plurality of UEs have been assigned the same group model ID during LCM operation or AI/ML models on different UEs have been assigned the same group model ID during LCM operation.
  8. 8. The method according to one of the preceding claims, characterized in that a single UE with multiple UEs is assigned the same group model ID.
  9. 9. The method according to the preceding claims 1 to 5, wherein there are a plurality of UE groups such that each UE group can be individually assigned a group model ID, UEs belonging to the same group having the same group model ID.
  10. 10. The method according to one of the preceding claims, characterized in that the group model ID is determined to be assigned at the network side such that the UE can use the assigned group model ID for one or more models.
  11. 11. The method according to one of the preceding claims, wherein the UE assigns a group model ID to a model on the UE, the assigning being based on a group model ID assignment configuration sent over the network, the UE further sending an acknowledgement message.
  12. 12. The method according to one of claims 1 to 10, characterized in that the network side assigns the group model ID and the UE identifies candidate models applicable to the group model ID, the UE further sending an indication message to the network confirming activation of the model based on the group model ID.
  13. 13. The method according to one of the preceding claims, characterized in that the model configuration update is performed using common/dedicated control signaling or data channel signaling via a lifecycle management (LCM) operation based on a group model ID.
  14. 14. The method according to one of the preceding claims, characterized in that LCM operation based on group model ID is applied to dynamic/semi-persistent scheduling or other scheduling types depending on different conditions, wherein: Activating/deactivating LCM operation based on group model ID is based on configured trigger metrics, and The network or the UE side decides the scheduling type.
  15. 15. A method of forming a model ID type structure comprising hierarchical model ID types for classifying different models assigned model IDs, wherein: assigning static model IDs to any specific model, or Assigning a semi-static model ID to any specific model based on different applications and/or configurations, or Assigning dynamic model IDs for and/or during ML/LCM operation, or Assigning a group common model ID to a plurality of models based on common properties configured by the network side or the UE side, or Assigning device-specific model IDs to any specific local device model, or Classified model ID types, which map each model to a different type of model ID depending on any condition or configuration when assigning a grouped model ID to a UE or group of UEs.
  16. 16. Group model identification configuration equipment for configuring the same group model ID for multiple models within a UE and/or across a group of UEs in a wireless communication system, the equipment comprising a wireless transceiver, a processor coupled with a memory in which computer program instructions are stored, the instructions being configured to implement the steps of the method according to claims 1 to 15, and the equipment being designed for use in a base station (gNB).
  17. 17. Group model identification configuration equipment for configuring the same group model ID for multiple models within a UE and/or across a group of UEs in a wireless communication system, the equipment comprising a wireless transceiver, a processor coupled with a memory in which computer program instructions are stored, the instructions being configured to implement the steps of the method of claims 1 to 15, and the equipment being designed for use in a User Equipment (UE).
  18. 18. A base station (gNB) comprising the apparatus of claim 16.
  19. 19. A user equipment comprising the apparatus of claim 15.
  20. 20. A wireless communication system comprising at least one base station (gNB) according to claim 16 and at least one User Equipment (UE) according to claim 17.

Description

Method for model packet signaling Technical Field The present disclosure relates to AI/ML-based model grouping identification, in which techniques are presented for pre-configuring and signaling specific information about model groupings using associations between models and their properties. Background In 3GPP (third Generation partnership project), one of the selected research projects as an approved version 18 package is AI/ML (Artificial Intelligence/machine learning), as described in the related document (RP-213599) submitted in 3GPP TSG (technical Specification group) RAN (radio Access network) conference #94 e. The formal name of this AI/ML study item is "AI/ML study of the NR air interface", and currently RANs WG1 (working group 1) and WG2 are actively formulating specifications. The goal of the research project is to identify a common AI/ML framework and a field of use of AI/ML-based techniques and use cases to obtain revenue. According to 3GPP, the main goal of this research project is to exploit the target use case to study the AI/ML framework for the air interface by taking into account performance, complexity and potential specification effects. In particular, the AI/ML model, terms and descriptions used to identify common and specific characteristics of the framework will be a key working scope. With respect to the AI/ML framework, surveys are being conducted considering various aspects, and one of the key items is lifecycle management (LCM) with respect to AI/ML models, wherein multiple phases are forcibly included for model training, model deployment, model reasoning, model monitoring, model updating, etc. Earlier, in release 17 3GPP TR 37.817, entitled "Study on enhancement for Data Collection for NR and EN-DC (research on NR and EN-DC data collection enhancement)", UE (user Equipment) mobility was also considered one of the AI/ML cases, and one of the scenarios of model training/reasoning is that both functions are located within the RAN node. Subsequently, a new work item "ARTIFICIAL INTELLIGENCE (AI)/MACHINE LEARNING (ML) for NG-RAN (artificial intelligence (AI)/Machine Learning (ML) for NG-RAN))" was initiated in release 18 to specify data collection enhancements and signaling support within existing NG-RAN interfaces and architectures. For the above-described aggressive standardization work, model identification (e.g., model ID) for supporting RAN-based AI/ML models is considered to be important for both the network and the UE to meet any desired model operation (e.g., model training, reasoning, selection, handover, update, monitoring, etc.). Model ID information may be signaled to pair both network-side and UE-side models for various lifecycle management operations. However, if all available physical models of both parties are individually identified with their own model IDs, the signaling overhead to indicate model ID information may be very high, especially when handling model ID-based LCMs between a base station (BS/gNB) and multiple UEs. Currently, there is no specification defined for signaling methods and network-UE behavior when all models are identified with their own model ID information to support RAN-based model operations. Thus, new mechanisms for base station (BS/gNB) -UE behavior and procedures are needed to avoid the heavy signaling overhead in model operation using multiple specific AI/ML models and/or operation across different UEs. When traffic is congested due to the signaling overhead of exchanging single model ID information, it is necessary to investigate any canonical effects by considering model operations. The disclosure of US 2022 400 A1 provides techniques and equipment for determining Neural Network Functions (NNFs) and for configuring and using corresponding Machine Learning (ML) models to execute one or more ML-based wireless communication hypervisors. An example method performed by a user equipment includes transmitting, to a Base Station (BS), UE capability information indicating at least one radio capability of the UE and at least one Machine Learning (ML) capability of the UE, and receiving, from the BS, ML configuration information indicating at least one Neural Network Function (NNF) and at least one ML model corresponding to the at least one NNF based on the UE capability information. US 2022,108,214 A1 discloses a Machine Learning (ML) model management method for a network data analysis function (NWDAF) device. NWDAF the device performs at least one of an analysis logic function (AnLF) and an ML Model Training Logic Function (MTLF) of the network data. US 2022 337 487 A1 discloses a network entity determining at least one model parameter of a model to digitally analyze input data in dependence of the at least one model parameter of the model, the network entity being configured to receive a model request from a requesting entity over a communication network, the model requesting the at least one model parameter of the model to obtain the requested