WO2025065720A1 - 无线通信系统中的ai/ml模型的控制方法 - Google Patents

无线通信系统中的ai/ml模型的控制方法 Download PDF

Info

Publication number
WO2025065720A1
WO2025065720A1 PCT/CN2023/123047 CN2023123047W WO2025065720A1 WO 2025065720 A1 WO2025065720 A1 WO 2025065720A1 CN 2023123047 W CN2023123047 W CN 2023123047W WO 2025065720 A1 WO2025065720 A1 WO 2025065720A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
information
base station
message
terminal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2023/123047
Other languages
English (en)
French (fr)
Inventor
曲淼
陈喆
张银成
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen TCL New Technology Co Ltd
Original Assignee
Shenzhen TCL New Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen TCL New Technology Co Ltd filed Critical Shenzhen TCL New Technology Co Ltd
Priority to PCT/CN2023/123047 priority Critical patent/WO2025065720A1/zh
Publication of WO2025065720A1 publication Critical patent/WO2025065720A1/zh
Anticipated expiration legal-status Critical
Pending legal-status Critical Current

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/16Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence

Definitions

  • the present disclosure relates to the field of communication systems, and more specifically, to a control method for an artificial intelligence/machine learning (AI/ML) model in a wireless communication system, a chip thereof, a computer-readable storage medium, and a computer program product, a wireless communication method for a user equipment (terminal), a chip thereof, a computer-readable storage medium, and a computer program product, and a wireless communication method for a first base station, a chip thereof, a computer-readable storage medium, and a computer program product.
  • AI/ML artificial intelligence/machine learning
  • AI artificial intelligence
  • ML machine learning
  • the channel may become worse due to the mobility of the user equipment (terminal), which may cause a handover to occur.
  • the terminal moves to a new serving base station (such as a gNB)
  • the system can use traditional methods or AI/ML-based methods in certain use cases.
  • model control e.g., model selection, switching, updating, etc.
  • current wireless communication networks do not have the ability to support such model control.
  • the present disclosure provides a control method for an artificial intelligence/machine learning (AI/ML) model in a wireless communication system, which can ensure the normal operation of the model control during the switching process, reduce signaling overhead, and reduce the delay in reporting dynamic information.
  • AI/ML artificial intelligence/machine learning
  • One aspect of the present disclosure provides a control method for an artificial intelligence/machine learning AI/ML model in a wireless communication system, including:
  • the wireless communication node obtains AI/ML model area information, and the wireless communication node obtains AI/ML model area information based on the AI/ML model area information.
  • the information determines the model control of the AI/ML model, and the wireless communication node is a base station or a terminal.
  • the model control of the AI/ML model includes at least one of the following:
  • AI/ML model activation AI/ML model deactivation
  • AI/ML model selection AI/ML model selection
  • AI/ML model switching AI/ML model switching
  • AI/ML model rollback AI/ML model update
  • AI/ML model training AI/ML model retraining
  • AI/ML model fine-tuning AI/ML model fine-tuning.
  • the AI/ML model region information includes at least one of the following:
  • the location information includes at least one of the following:
  • Cell information AI/ML model area identification ID, terminal location information, base station location information, scene information, control unit CU ID, data unit DU ID.
  • the scene information includes at least one of the following:
  • Transmission environment information For example, speed information, angle spread information, delay spread information, signal-to-noise ratio, channel received power, and/or interference information.
  • the wireless communication node capability information includes at least one of the following:
  • AI/ML model supports AI/ML usage scenarios and/or AI/ML model supports AI/ML features.
  • some specific computing capacity, some specific storage capacity, AI/ML model supports some specific capacity, some specific load conditions, some specific transmission capacity, AI/ML model supports some specific usage scenarios and/or supports some specific features.
  • Wireless network capability node capability information indicates the capabilities of a wireless network node, such as the model operation capabilities of the AI/ML models supported by the node, the node's hardware capability information (such as storage capacity, computing power, battery capacity, etc.), the functionality/features of the AI/ML models supported by the node, the data of the AI/ML models supported by the node, the scenarios of the AI/ML models supported by the node, etc.
  • the AI/ML model information includes at least one of the following:
  • AI/ML model ID information information, and/or AI/ML model meta information.
  • the AI/ML model region information is transmitted through at least one of the following:
  • Non-access layer NAS message radio resource control RRC message, medium access control element MAC-CE message, Downlink control information DCI message, uplink channel, downlink channel, system message, N2 message and/or Xn message
  • determining, by the wireless communication node, model control of the AI/ML model based on the AI/ML model area information includes at least one of the following:
  • the wireless communication node does not perform model control of the AI/ML model
  • the wireless communication node If the location information of the wireless communication node does not belong to the location information indicated in the AI/ML model area information, the wireless communication node performs model control of the AI/ML model;
  • the wireless communication node does not perform model control of the AI/ML model
  • the wireless communication node If the AI/ML model information of the wireless communication node does not belong to the AI/ML model information indicated in the AI/ML model area information, the wireless communication node performs model control of the AI/ML model.
  • the model control of executing the AI/ML model includes at least one of the following:
  • the model control of executing the AI/ML model includes at least one of the following:
  • AI/ML model activation AI/ML model deactivation
  • AI/ML model selection AI/ML model selection
  • AI/ML model switching AI/ML model switching
  • AI/ML model rollback AI/ML model update
  • AI/ML model training AI/ML model retraining
  • AI/ML model fine-tuning reporting of results of model control of the AI/ML model, and/or response of model control of the AI/ML model.
  • the terminal when the wireless communication node is a terminal, the terminal obtains the AI/ML model area configuration information, and the terminal performs model control of the AI/ML model according to the AI/ML model area information.
  • the terminal acquires the AI/ML model area configuration information including at least one of the following:
  • AI/ML model information reasons, indicators, terminal ID, and/or location information.
  • the base station when the wireless communication node is a base station, the base station obtains the AI/ML model area configuration information, and the base station performs model control of the AI/ML model according to the AI/ML model area information.
  • the base station acquires the AI/ML model area configuration information including at least one of the following:
  • AI/ML model information reasons, indicators, base station ID, and/or location information.
  • the AI/ML model area information request message and/or response message sent by the first wireless communication node to the second wireless communication node includes at least one of the following:
  • Wireless communication node capability information and/or
  • the wireless communication node sends a request message for the AI/ML model area information, including periodic transmission or event triggering.
  • the AI/ML model area information is updated in response to a periodic configuration, wherein the periodic configuration includes a duration, a start time, and/or an end time of the update.
  • the AI/ML model area information is updated in response to the event triggered configuration.
  • the periodic transmission and/or the event triggering corresponding periodic configuration and/or event triggering configuration are preconfigured, and/or configured, and/or fixed.
  • the AI/ML model region information request message and/or response message is carried by at least one of the following:
  • Non-access stratum NAS message radio resource control RRC message, medium access control element MAC-CE message, downlink control information DCI message, uplink channel, downlink channel, system message, N2 message and/or Xn message.
  • One aspect of the present disclosure provides a wireless communication method for a terminal, including:
  • the terminal acquires AI/ML model area information from the first base station, where the AI/ML model area information is used by the terminal to determine model control of the AI/ML model,
  • the terminal determines whether to execute model control of the AI/ML model based on the AI/ML model area information.
  • the present invention further includes:
  • the terminal obtains an AI/ML model area configuration message from the first base station.
  • the terminal determines to perform model control of the AI/ML model.
  • the model control of executing the AI/ML model includes: The AI/ML model area information is updated between the first base stations.
  • the present invention further includes:
  • the terminal sends a report message for the AI/ML model area configuration message and a result of executing the model control of the AI/ML model to the first base station, and the first base station sends the obtained result of executing the model control of the AI/ML model to the second base station.
  • One aspect of the present disclosure provides a wireless communication method for a first base station, including:
  • the first base station obtains a model control reporting message of the AI/ML model from the terminal, where the model control reporting message of the AI/ML model includes a result of the model control of the AI/ML model;
  • the first base station sends the cell switching request of the second base station and the result of the model control of the AI/ML model to the second base station.
  • One aspect of the present disclosure provides a wireless communication method for a first base station, including:
  • the first base station determines, according to a model control message of the AI/ML model fed back by the terminal, whether the second base station is in the AI/ML model area information;
  • the first base station performs model control of the AI/ML model.
  • the model control of the AI/ML model includes updating the AI/ML model area information between the first base station and the second base station.
  • the present invention further includes:
  • the first base station sends the cell switching request of the second base station and the result of the model control to the second base station.
  • the present invention further includes:
  • the first base station sends the AI/ML model information contained in the AI/ML model area information and a cell switching request to the second base station to the second base station.
  • One aspect of the present disclosure provides a chip, including:
  • the processor is configured to call and run the computer program stored in the memory so that the device in which the chip is installed performs any one of the methods in the example embodiments.
  • One aspect of the present disclosure provides a chip, comprising: a processor configured to call and run a computer program stored in a memory, so that a device in which the chip is installed executes a method of an embodiment of any aspect of the present disclosure.
  • One aspect of the present disclosure provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program enables a computer to execute a method of an embodiment of any aspect of the present disclosure.
  • One aspect of the present disclosure provides a computer program product, including a computer program, wherein the computer program enables a computer to execute a method according to any aspect of the present disclosure.
  • FIG1 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG2 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG3 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG4 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG5 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG6 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG. 7 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG8 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • FIG9 illustrates a signaling diagram of a fast model control procedure during a handover procedure according to some embodiments of the present disclosure.
  • FIG. 10 illustrates a signaling diagram of a fast model control procedure during a handover procedure according to some embodiments of the present disclosure.
  • FIG. 11 illustrates a signaling diagram of a fast model control procedure during a handover procedure according to some embodiments of the present disclosure.
  • FIG. 12 is a block diagram of an example system for wireless communications according to an embodiment of the disclosure.
  • a or B may mean “only A”, “only B”, or “both A and B”.
  • a or B may be interpreted as “A and/or B”.
  • A, B or C may mean “only A”, “only B”, “only C” or "any combination of A, B, C”.
  • a slash (/) or a comma used in the present disclosure may mean “and/or”.
  • A/B may mean “A and/or B”.
  • A/B may mean “only A”, “only B”, or “both A and B”.
  • A, B, C may mean “A, B, or C”.
  • At least one of A and B may mean “only A”, “only B”, or “both A and B”.
  • the expression “at least one of A or B” or “at least one of A and/or B” may be interpreted as "at least one of A and B”.
  • At least one of A, B, and C may mean “only A”, “only B”, “only C”, or “any combination of A, B, and C”.
  • at least one of A, B, or C or “at least one of A, B and/or C” may mean “at least one of A, B, and C”.
  • first and second are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as “first” and “second” may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of “plurality” is two or more, unless otherwise clearly and specifically defined.
  • the channel may become worse due to terminal mobility, which may cause handover to occur.
  • a new serving base station such as a gNB
  • the system may use traditional methods or AI/ML-based methods in certain use cases.
  • model control of the AI/ML model may need to be performed during the handover process (for example, AI/ML model activation, AI/ML model deactivation, AI/ML model selection, AI/ML model switching, AI/ML model fallback, AI/ML model update, AI/ML model training, AI/ML model retraining, or AI/ML model fine-tuning, etc., the present disclosure does not impose any restrictions on this).
  • AI/ML model activation for example, AI/ML model activation, AI/ML model deactivation, AI/ML model selection, AI/ML model switching, AI/ML model fallback, AI/ML model update, AI/ML model training, AI/ML model retraining, or AI/ML model fine-tuning, etc.
  • AI/ML model activation for example, AI/ML model activation, AI/ML model deactivation, AI/ML model selection, AI/ML model switching, AI/ML model fallback, AI/ML model update, AI/ML model training, AI/ML model retraining, or AI/ML model fine-tun
  • model control also known as model operation, model management, etc.
  • model control of the AI/ML model such as model activation, deactivation, selection, switching, fallback, update, training, and retraining is a decision made based on the monitoring results after the model starts monitoring.
  • the model monitoring solution is used to determine whether the current AI/ML model is suitable for the second base station
  • the second base station needs to configure model monitoring and monitoring reporting.
  • this configuration process usually requires the terminal to access the second base station, that is, after the switching is completed, before the model monitoring judgment can be made.
  • the terminal when the model monitoring is on the terminal side and the monitoring decision is on the network side (for example, but not limited to the base station (such as gNB), etc.), the terminal needs to report the monitoring results to assist the network side in selecting a new model, and the network side then sends the newly selected model to the terminal.
  • the terminal When the model monitoring is on the terminal side and the monitoring decision is also on the terminal side, the terminal needs to reselect the model based on the decision.
  • Another optional solution that does not use model monitoring is that when a cell switching occurs, in order to adapt the terminal's AI/ML model to the new network side (such as the second base station), the model control of the AI model can be performed directly.
  • the newly switched/selected/updated AI model can be directly transmitted to the terminal after being selected by the second base station, or the relevant information of the new model configured by the second base station (such as the model identifier (ID), model meta-information, etc.) is sent to the terminal to assist the terminal side in selecting/switching the new AI model.
  • ID model identifier
  • this method will result in the need for direct interaction of the AI model or the required AI model information between the base station and the terminal, resulting in a large signaling overhead.
  • the entire system will face a large delay and cannot meet the requirements for service continuity.
  • the present disclosure proposes a control method for an artificial intelligence/machine learning (AI/ML) model in a wireless communication system.
  • the method can be used to determine whether model control needs to be performed during a switching process, and can be used to assist nodes, for example, to quickly select/switch/update one or more suitable AI/ML models when model control occurs.
  • the delay caused by model monitoring and model control of AI/ML models can be effectively reduced; and the signaling interaction overhead when performing model control can be reduced.
  • the terminal may be various user devices such as a user station, a mobile station, a mobile station, a wireless communication device, a user agent, etc.
  • some examples of the terminal may be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile Internet device, a wearable device, a virtual reality device, an augmented reality device, etc., and the present disclosure does not impose any limitation on this.
  • a base station may include or may be referred to as a base station transceiver, a wireless base station, an access point, a wireless transceiver, a node B, an eNode B (eNB), a next generation node B or a gigabit node B (any of which may be referred to as a gNB), a home node B, a home eNode B or other suitable terms.
  • a wireless communication system may include different types of base stations (e.g., a macro cell base station or a small cell base station, etc.).
  • a terminal may communicate with various base stations and/or network devices, including macro eNBs, small cell eNBs, gNBs, relay base stations, etc.
  • AI/ML model specific area information which can be used to assist nodes (such as terminals, gNBs, etc.) in performing AI/ML model control.
  • an AI/ML model can be used for one or more locations in a specific AI/ML area.
  • the AI/ML model area information may include at least one of the following: an AI/ML model area identification ID, location information, scene information, wireless communication node capability information, and/or AI/ML model information.
  • the AI/ML model area information may be configured to a node, such as a terminal and/or a base station, via an RRC message, a NAS message, an N2 message, an Xn message, a DCI, a MAC CE, a system message, an uplink channel, a downlink channel, etc.
  • AI/ML model area information refers to an AI-specific area (AI-specific area), which may also be interchangeably referred to as an AI/ML model-specific area in the present disclosure, and the AI/ML model area information corresponds to or is associated with an AI/ML model;
  • AI/ML model region indicator used to identify AI-specific regions
  • AI-specific area code of AI-specific area AI-specific area code that can be used to indicate AI-specific area
  • AI/ML model information information that can be used to indicate the AI/ML model, such as model ID, model meta-information, etc.
  • the AI/ML model information can be information about one AI/ML model or multiple AI/ML models;
  • Location information cell information, AI/ML model area identification ID, terminal location information, base station location information, scene information, control unit CU ID, data unit DU ID, etc.; Among them,
  • the wireless communication node capability information includes at least one of the following:
  • AI/ML model supports AI/ML usage scenarios and/or AI/ML model supports AI/ML features.
  • some specific computing capacity, some specific storage capacity, AI/ML model supports some specific capacity, some specific load conditions, some specific transmission capacity, AI/ML model supports some specific usage scenarios and/or supports some specific features.
  • Wireless network capability node capability information indicates the capabilities of a wireless network node, such as the model operation capabilities of the AI/ML models supported by the node, the node's hardware capability information (such as storage capacity, computing power, battery capacity, etc.), the The functionality/features of the AI/ML model, the data of the AI/ML model supported by the node, the scenarios of the AI/ML model supported by the node, etc.
  • Base station information may indicate cell ID, transmission reception point (TRP) ID, central unit (CU) ID and/or distribution unit (DU) ID, etc.;
  • Scene information one or more parameters reflecting channel characteristics such as environment information, speed information, angle spread information, delay spread information, signal-to-noise ratio, channel receiving power, and/or interference information may be transmitted;
  • Terminal location information may include, for example, the terminal ID, the cell ID associated with the terminal, the physical location area of the terminal, longitude, latitude, altitude, etc.
  • Model ID used to identify the AI/ML model, including the global model ID and/or the local model ID, where:
  • the global AI/ML model ID is used to identify the AI/ML model, which can be a globally unique ID, a PLMN-specific unique ID, an operator-specific unique ID, or an AI/ML management platform-specific unique ID.
  • the local AI/ML model ID is used to identify the AI/ML model used in the network and has a certain mapping relationship with the global ID.
  • the logical ID can be a globally unique ID, a PLMN-specific unique ID, an operator-specific unique ID, a cell-specific ID, a link-specific ID, a tracking area (TA)-specific ID, a CU-specific ID, a DU-specific ID, a user plane function (UPF)-specific ID, an access and mobility management function (AMF)-specific ID, a radio resource control (RRC)-specific ID, or a network slice-specific ID.
  • TA tracking area
  • CU-specific ID a DU-specific ID
  • UPF user plane function
  • AMF access and mobility management function
  • RRC radio resource control
  • cell-specific means that each AI/ML model has a unique ID within a specific cell.
  • link-specific means that each AI/ML model has a unique ID within a specific cell.
  • link-specific means that each AI/ML model has a unique ID within a specific cell.
  • TA-specific means that each AI/ML model has a unique ID within a specific cell.
  • CU-specific means that each AI/ML model has a unique ID within a specific cell.
  • DPF-specific DU-specific
  • UPF-specific Ultra-specific
  • AMF-specific AMF-specific
  • network slice-specific refer to the unique ID of an AI/ML model in a specific context in the network.
  • the model ID can also be indicated implicitly, such as RRC ID, DCI ID, etc.
  • Model meta-information Information used to describe AI/ML models, which may include model functionality/features, model version information, model format information, vendor information, computational complexity, model complexity, model performance, applicable conditions, etc.
  • Model functionality/features Indicates the functionality of a model, e.g. the model is used for CSI compression, beam prediction, positioning, etc. The indication should be at a sub-use case level or even finer.
  • Model version information indicates the version information of the AI/ML model.
  • Model format information Indicates the format information of the AI/ML model, such as open formats (such as ONNX), private formats, etc.
  • Vendor Information Information indicating the AI/ML model training vendor.
  • Computational complexity used to indicate the computational complexity of model training, such as FLOP (floating point operations), pre-processing level of processing/post-processing.
  • Model complexity used to indicate the complexity of model training, including the number of real-valued model parameters, the number of real values, etc.
  • Model performance used to indicate the performance of the model, which can be model accuracy, model bias, model variance, etc.
  • Applicable conditions used to indicate the applicable conditions of AI/ML models or functionality, such as scenarios, configurations or sites, terminal internal capabilities, etc.
  • Transmission environment information used to describe the channel transmission environment in which the current wireless communication node is located, such as indoor/urban/urban suburbs/rural/tunnel, etc., or channel classification information, such as classification information based on UMA/UMi/InH/RMA, etc.
  • Speed information used to describe the current moving speed of the wireless communication node.
  • the moving speed can be measured by the terminal or the base station.
  • Angle expansion information used to describe the angle expansion of the channel emitted by the current wireless communication node, that is, the real-time information or statistical information of the channel angle expansion, or other information related to the angle expansion.
  • Delay spread information used to describe the delay spread of the channel emitted by the current wireless communication node, that is, the real-time information or statistical information of the channel delay spread, or other information related to the delay spread.
  • the wireless communication node capability information includes at least one of the following:
  • Wireless network capability node capability information indicates the capabilities of a wireless network node, such as the model operation capabilities of the AI/ML models supported by the node, the node's hardware capability information (such as storage capacity, computing power, battery capacity, etc.), the functionality/features of the AI/ML models supported by the node, the data of the AI/ML models supported by the node, the scenarios of the AI/ML models supported by the node, etc.
  • the AI/ML model information includes at least one of the following:
  • AI/ML model ID information information, and/or AI/ML model meta information.
  • model control may also be referred to as model operation, model management, etc.
  • AI-specific area information is introduced to assist nodes (such as terminals, accounting, etc.) to correctly perform AI/ML model control, and/or to help nodes, for example, select/switch/update one or more appropriate AI/ML models when model control occurs.
  • AI/ML model region information can be constructed in the following ways.
  • composition of AI/ML model area information includes at least one of the following: AI/ML model area identification ID, location information, scene information, wireless communication node capability information, and/or AI/ML model information.
  • AI/ML model region information can consist of single or multiple location information
  • AI/ML model region information can consist of single or multiple location information and some AI/ML model information
  • the AI/ML model area information can consist of single or multiple location information, some AI/ML model information, and some scene information.
  • AI/ML model information includes at least one of the following: AI/ML model ID information, and/or AI/ML model meta information.
  • the mapping between the AI/ML model region information and the AI/ML model information can be a one-to-one, one-to-many, or many-to-one relationship. It should be noted that there can be a mapping relationship between any one or more elements in the AI/ML model region information and any one or more elements in the AI/ML model information, and the present disclosure does not impose any restrictions on this.
  • an AI/ML model region indicator may be used to represent a mapping between location information and AI/ML model information.
  • Table 1 shows a possible mapping relationship between AI/ML model region information and AI/ML model information, where the location information is a cell ID and the model information is a model ID;
  • Table 2 shows another possible mapping relationship between AI/ML model region information and AI/ML model information, where the location information is a cell ID and the model information is model features and/or functionality, model format.
  • Table 1 Mapping relationship between AI/ML model region information and AI/ML model information
  • Table 2 Mapping relationship between AI/ML model region information and AI/ML model information
  • a possible mapping relationship between AI/ML model region information and AI/ML model information is shown in Table 3 below, where the location information is a cell ID and the model information is a model ID:
  • Table 3 Mapping relationship between AI/ML model region information and AI/ML model information
  • AI/ML model region information a possible mapping relationship between AI/ML model region information and AI/ML model information is shown in Table 4 below, where the location information is the cell ID, and the model information is the model ID and AI features:
  • Table 4 Mapping relationship between AI/ML model region information and AI/ML model information
  • an index may be introduced, as shown in the following Table 5, where the location information is the cell ID and the model information is the model ID:
  • Table 5 Mapping relationship between AI/ML model region information and AI/ML model information
  • a terminal (such as a UE) in an RRC activated, inactive, or idle state may be configured with AI/ML model area information or an AI/ML model area information list, which may be configured through the following options:
  • the core network element (CN, such as AMF) sends AI/ML model area information or AI/ML model area information list to the terminal through, for example, a NAS message.
  • CN core network element
  • AMF AI/ML model area information
  • NAS message a NAS message
  • the base station (such as gNB) sends AI/ML model area information or AI/ML model area information list to the terminal through, for example, an RRC message, where the RRC message includes, for example, an RRC reconfiguration message, an RRC release message, or other dedicated RRC messages.
  • RRC message includes, for example, an RRC reconfiguration message, an RRC release message, or other dedicated RRC messages.
  • the present disclosure does not impose any restrictions on the type of message used, and does not exclude the use of any other suitable message other than the RRC message;
  • the base station (such as gNB) sends AI/ML model area information or AI/ML model area information list to the terminal through a system information message, including SIB1 or other newly defined SIBs.
  • a system information message including SIB1 or other newly defined SIBs.
  • the present disclosure does not impose any restrictions on the type of message used, and does not exclude the use of any other suitable message other than the system information message.
  • the base station (such as gNB) sends AI/ML model area information or AI/ML model area information list to the terminal through, for example, a MAC CE message; the present disclosure does not impose any restrictions on the type of message used, and does not exclude the use of any other suitable message other than the MAC CE message;
  • the base station (such as gNB) sends AI/ML model area information or AI/ML model area information list to the terminal through, for example, a DCI message; the present disclosure does not impose any restrictions on the type of message used, and does not exclude the use of any other suitable message other than the DCI message;
  • a base station (such as a gNB) can be configured with AI/ML model region information or a list of AI/ML model region information, which can be configured through the following options:
  • the core network element (such as AMF) sends AI/ML model area information or AI/ML model area information list to the base station (such as gNB) through N2 message.
  • AMF AI/ML model area information
  • gNB base station
  • a base station (such as a gNB) sends AI/ML model area information or an AI/ML model area information list to another base station (such as a gNB) via an Xn message.
  • a base station such as a gNB
  • a base station sends AI/ML model area information or an AI/ML model area information list to another base station (such as a gNB) via an Xn message.
  • the present disclosure does not impose any restrictions on the type of message used, and does not exclude the use of any other suitable message other than the Xn message.
  • the AI/ML model area information update cycle can be configured through, for example, RRC messages, NAS messages, N2 messages or system messages, Xn messages, MAC CE, DCI, etc.
  • the update cycle configuration content includes the duration, start time and end time of at least one of the following update cycles, for example, 5 minutes to 720 minutes.
  • Event-triggered request update Any of the following events may trigger an AI/ML model region information update:
  • Event 1 The model corresponding to the AI/ML model region information is not in the configured AI/ML model region information, which will trigger the AI/ML model region information update; for example, if the model changes (including model selection, model switching, model update, etc.), the corresponding AI/ML model region information does not belong to the configured AI/ML model region information.
  • Event 2 The location information of the candidate base station and/or cell is not in the configured AI/ML model area information, which will trigger the update of the AI/ML model area information.
  • Event 3 When the cell reselection process of the terminal selects a base station and/or cell that does not belong to the configured AI/ML model area information, the AI/ML model area information update will be triggered.
  • the following describes a possible process for updating the AI/ML model region information.
  • FIG1 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the process of updating the regional information of the AI/ML model triggered by the terminal is illustrated.
  • the process can be triggered by an event or a period.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include a terminal (such as a terminal) sending a request message for updating AI/ML model area information to a base station (such as a gNB).
  • the message includes at least the following information:
  • -Model information used to indicate the current AI model on the terminal side, which can be an activated AI model or an inactivated AI model.
  • - Reason Used to indicate the reason for updating AI/ML model area information.
  • the reason is the current AI Model and AI/ML model region information do not belong to the configured AI/ML model region information list.
  • -Terminal ID used to identify a specific terminal.
  • the message may be a dedicated message for requesting an update of the AI/ML model area information.
  • the base station such as a gNB
  • the base station knows that the message is used to request an update of the AI/ML model area information.
  • the message may be a traditional message, which may include an indicator to indicate that the message is for an AI/ML model area information update request.
  • the message may also be a traditional message, which may include a cause to indicate the reason for initiating the dedicated message.
  • step 2 may include the terminal receiving updated AI/ML model area information.
  • the updated AI/ML model area information may include an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information.
  • the terminal is able to update its AI/ML model area information and obtain the latest AI model information and related location information and mapping rules. In this way, the terminal can better adapt to the current network environment and needs and provide higher quality services. Therefore, the embodiments of the present disclosure provide an effective AI/ML model area information update method that can improve network performance and user experience.
  • FIG2 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by the terminal is illustrated.
  • the process can be triggered by an event or periodically.
  • the first base station may forward the request to other base stations (for example, the second base station), as shown in FIG2.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include the terminal sending a request message for updating AI/ML model area information to the first base station, wherein the information is the same as step 1 in the example shown in FIG. 1 .
  • step 2 may include the first base station forwarding the request message received in step 1 to (e.g., multiple) other base stations.
  • the other base stations will send updated AI/ML model area information, more specifically, the updated AI/ML model area information may include AI/ML model area identification ID, location information, scene information, wireless communication node capability information, and/or AI/ML model information.
  • step 3 may include the terminal receiving a response message regarding the AI/ML model region information update request, wherein The information in is the same as the information sent by other base stations to the first base station in step 2.
  • the terminal can request an AI/ML model area information update from a base station, such as the first base station and then other base stations.
  • a base station such as the first base station and then other base stations.
  • the terminal can ensure that it has the most relevant AI model, improve performance and adapt to changing network conditions. Therefore, the embodiments of the present disclosure provide an effective terminal-triggered AI/ML model area information update process that can improve network performance and user experience.
  • FIG3 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by a base station (such as a gNB) is illustrated.
  • the process can be triggered by an event or periodically.
  • the first base station if the first base station (s-gNB) does not have AI/ML model area information (or AI/ML model area information list) corresponding to other candidate gNBs, the first base station triggers the AI/ML model area information update process.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include the first base station triggering an AI/ML model area information update process.
  • the first base station may send a request message for updating the AI/ML model area information to other base stations, and the information may include model information and a reason.
  • step 2 may include other base stations sending a response message regarding the AI/ML model area information update request to the first base station B.
  • the response message may indicate that the AI/ML model area information update request is granted and includes updated AI/ML model area information.
  • the base station (such as gNB) can trigger the update of AI/ML model area information to ensure the validity of the configured AI/ML model area information (list), for example, including AI/ML model area information (list) corresponding to other candidate base stations. Therefore, the embodiments of the present disclosure provide an AI/ML model area information update process triggered by a base station (such as gNB), which can improve network performance and efficiency.
  • a base station such as gNB
  • FIG4 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by the terminal is illustrated.
  • the process can be triggered by an event or periodically.
  • a method for updating AI/ML model region information is provided. The following are the steps of the method: Steps/Signaling:
  • step 1 may include the terminal sending a request message for updating AI/ML model area information to a base station (such as a gNB), where the message includes at least the following contents:
  • Location information used to indicate the location information of base stations (such as gNB), including candidate base stations.
  • Cause used to indicate the reason for the AI/ML model area information update.
  • the cause may be that the cell of the candidate base station is not in the configured AI/ML model area information, or that a model (re)selection is performed and the new resident cell does not belong to the configured AI/ML model area information.
  • Terminal ID used to indicate the identity of the terminal.
  • the message may be a dedicated message for requesting an AI/ML model area information update. Once a base station (such as a gNB) receives the message, it knows that it is used to request an AI/ML model area information update. Alternatively, the message may also be a traditional message, which may include an indicator to indicate that the message is for an AI-TA update request. Alternatively, the message may also be a traditional message, which may include a reason to indicate why a dedicated message is being initiated.
  • step 2 may include the terminal receiving updated AI/ML model area information.
  • the updated AI/ML model area information may include any one of the following: an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information.
  • the terminal is able to send an AI/ML model area information update request to a base station (such as a gNB) and receive updated AI/ML model area information to improve network performance and efficiency.
  • a base station such as a gNB
  • FIG5 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by the terminal is illustrated.
  • the process can be triggered by an event or periodically.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include the terminal sending a request message for updating the AI/ML model area information to a base station (such as a gNB), the message including at least one of the following: location information, reason, terminal identification, etc. More specifically, for the above event 2, the information of the candidate base station needs to be included in the message to indicate the requested AI/ML model area information (list).
  • a base station such as a gNB
  • step 2 may include the first base station forwarding the request message in step 1 to (e.g., multiple) other base stations, and the other base stations sending updated AI/ML model area information.
  • the updated AI/ML model area information to The data element shall include at least one of the following: an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information.
  • step 3 may include the terminal receiving updated AI/ML model area information from the first base station, more specifically, the updated AI/ML model area information includes at least one of the following: an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information.
  • the updated AI/ML model area information includes at least one of the following: an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information.
  • the terminal can send an AI/ML model area information update request to a base station, such as the first base station and then other base stations, and receive updated AI/ML model area information from the first base station to improve network performance and efficiency.
  • a base station such as the first base station and then other base stations
  • FIG6 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by a base station (such as a gNB) is illustrated.
  • the process can be triggered by an event or periodically.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include the first base station sending a request message for updating the AI/ML model area information to other base stations, the message including at least one of the following: location information, reason, terminal identification, etc. More specifically, for event 2, the cell information of the candidate base station needs to be included in the message to indicate the requested AI/ML model area information (list).
  • step 2 may include other base stations responding to the message in step 1, and the other base stations sending updated AI/ML model area information.
  • the updated AI/ML model area information may be an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information.
  • the first base station is able to send AI/ML model area information update requests to other base stations and receive updated AI/ML model area information from other base stations to improve network performance and efficiency.
  • FIG. 7 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by the terminal is illustrated.
  • the process can be triggered by an event or periodically.
  • the terminal triggers the AI/ML model area information update process, and it will send a message to the base station (such as gNB).
  • the base station such as gNB
  • the base station forwards the message to the CN node, and the CN will respond to the terminal by updating the AI/ML model area information.
  • the AI/ML model area information update may occur during the registration process.
  • the AI/ML model area information is sent by the AMF to the terminal.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include the terminal sending a request message for updating the AI/ML model area information to the base station (such as gNB) through a registration request message, and the message includes at least one of the following: location information, reason, terminal identification, AI model information, etc. More specifically, for event 1, the AI model information needs to be included in the message; for event 2, the cell information of the candidate base station needs to be included in the message to indicate the requested AI/ML model area information (list).
  • the base station such as gNB
  • the message includes at least one of the following: location information, reason, terminal identification, AI model information, etc. More specifically, for event 1, the AI model information needs to be included in the message; for event 2, the cell information of the candidate base station needs to be included in the message to indicate the requested AI/ML model area information (list).
  • step 2 may include a base station (such as a gNB) forwarding the AI/ML model area information update request message to the AMF via, for example, an N2 message.
  • a base station such as a gNB
  • step 3 may include if the registration request of the terminal is accepted, the terminal receiving updated AI/ML model area information via a registration acceptance message from the AMF.
  • the terminal is able to trigger the AI/ML model area information update process and receive updated AI/ML model area information from the AMF to improve network performance and efficiency.
  • FIG8 illustrates a signaling diagram of an AI/ML model region information update process according to some embodiments of the present disclosure.
  • the AI/ML model area information update process triggered by a base station (such as a gNB) is illustrated.
  • the process can be triggered by an event or periodically.
  • the base station (such as gNB) triggers the AI/ML model area information update process, and it will send a request message to the CN node (such as AMF, Network Data Analysis Function (NWDAF), etc.).
  • the CN node such as AMF, Network Data Analysis Function (NWDAF), etc.
  • a method for updating AI/ML model region information is provided. The following are the steps/signaling of the method:
  • step 1 may include a base station (such as a gNB) sending a request message for updating the AI/ML model area information to the AMF through, for example, an N2 message.
  • the request message includes at least one of the following: location information, cause, terminal identification, AI model information, etc. More specifically, for event 1, the AI model information needs to be included in the message; for event 2, the cell information of the candidate base station needs to be included in the message to indicate the requested AI/ML model area information (list).
  • step 2 may include the base station receiving updated AI/ML model area information from the AMF via, for example, an N2 message.
  • the base station (such as gNB) can send a request message for updating the AI/ML model area information to the AMF through the N2 message, and can receive the updated AI/ML model area information from the AMF through the N2 message to achieve more efficient network communication and optimization.
  • the terminal determines the process of fast model control during the switching process
  • FIG9 illustrates a signaling diagram of a fast model control procedure during a handover procedure according to some embodiments of the present disclosure.
  • model control (such as model selection, model switching, model update, etc.) is located at the terminal end.
  • step 0 The terminal receives a configuration message about AI/ML model area information, which is used to help the terminal quickly determine model control during the switching process.
  • the message may include an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information, and the message may be transmitted in, for example, RRC messages, system information, MAC CE, DCI, NAS, etc.
  • step 1 the terminal receives a traditional measurement configuration, and the measurement configuration is provided through dedicated signaling (such as RRCReconfiguration or RRCResume, etc.).
  • dedicated signaling such as RRCReconfiguration or RRCResume, etc.
  • step 2 the terminal checks the configured AI/ML model area information. If the candidate second base station (such as Target gNB, T-gNB) is within the configured AI/ML model area information list, there is no need to perform model control, and then the terminal (such as the terminal) executes step 4; if the candidate second base station (such as Target gNB, T-gNB) is not within the configured AI/ML model area information list, the terminal will execute step 3 model control, and then the terminal executes step 3' and step 4. The details of step 3' are described above.
  • the candidate second base station such as Target gNB, T-gNB
  • step 4 based on the inspection result of step 2, the terminal sends a measurement reporting message to the first base station, and the message may carry the model control result, and the message includes at least one of the following: model information, indication or reason.
  • Indication Used to indicate whether the current in-progress model has been changed.
  • Model information used to indicate the current AI/ML model.
  • step 5 the first base station initiates a handover (HO) decision.
  • HO handover
  • step 6 the first base station sends a switching request message to the second base station, where the message may carry model information, where the model information indicates an AI/ML model applicable to the second base station.
  • This step may perform model identification in advance between the terminal and the second base station.
  • step 7 the second base station performs admission control.
  • step 8 the second base station sends a confirmation message about the handover request to the original base station, which carries information about the second base station and resource information of the terminal.
  • step 9 the first base station sends a message with a handover command (such as an RRCReconfiguration message, etc.) to the terminal, which carries information for accessing the second base station, such as a cell ID.
  • a handover command such as an RRCReconfiguration message, etc.
  • step 3' can occur before or after step 4. In some examples, if step 3' is sent after step 4, steps 4 and 6 do not carry AI model information.
  • step 9 carries activation/deactivation configuration information, including at least one of the following indicating the activation/deactivation start time, bias, duration and end time of the model.
  • step 9 only carries model activation/deactivation indication information. When the terminal receives the indication information, it activates/deactivates the model by itself.
  • a base station such as a gNB determines fast model control during handover
  • FIG. 10 illustrates a signaling diagram of a fast model control procedure during a handover procedure according to some embodiments of the present disclosure.
  • model control (such as model selection, model switching, model update, etc.) is located on the terminal side.
  • step 1 the terminal receives a traditional measurement configuration, which is provided via dedicated signaling (eg, RRCReconfiguration or RRCResume, etc.).
  • dedicated signaling eg, RRCReconfiguration or RRCResume, etc.
  • step 2 the terminal sends a measurement report message to the first base station (such as Source gNB, S-gNB) according to the measurement configuration in step 1.
  • the first base station such as Source gNB, S-gNB
  • step 3 based on the reporting message in step 2, the first base station initiates a switching (eg, handover, HO) decision.
  • a switching eg, handover, HO
  • step 4 The first base station checks the configured AI/ML model area information list. If the candidate second base station (such as Target gNB, T-gNB) is within the configured AI/ML model area information list, there is no need to perform model control. At this time, the first base station sends the AI/ML model information to the second base station and can carry it in the handover request message of step 8. If the candidate second base station is not within the configured AI/ML model area information list, the first base station executes steps 5, 6, and 7.
  • the candidate second base station such as Target gNB, T-gNB
  • step 5 the terminal receives a message for model control, the message being used to indicate that the current AI/ML model of the terminal needs to be model controlled.
  • the message includes at least one of the following:
  • Location information used to indicate the current location information of the second base station
  • AI/ML model area information used to indicate the AI/ML model area information corresponding to the second base station, such as an AI/ML model area indicator, an AI/ML model area indicator list, AI model information, location information, or a mapping rule/relationship between AI/ML model area information and model information; it can also be used to help the terminal select/switch an AI/ML model;
  • Indicators used to indicate that model controls need to be implemented and/or that the current AI/ML model is no longer applicable
  • step 6 the terminal performs model control according to step 5.
  • step 7 the terminal sends a message with a model control result to the first base station, such as AI/ML model information; or an indicator for indicating that the AI/ML model has been changed;
  • step 8 the first base station sends a handover request message to the second base station, the message may carry AI/ML model information, and the AI/ML model information indicates an AI model applicable to the second base station.
  • This step can perform model identification in advance between the terminal and the second base station.
  • step 9 the second base station performs admission model control.
  • step 10 the second base station sends a confirmation message about the handover request to the first base station, which carries information about the second base station and resource information of the terminal.
  • step 11 the first base station sends a message with a handover command (such as an RRCReconfiguration message, etc.) to the terminal, which carries information for accessing the second base station, such as a cell ID.
  • a handover command such as an RRCReconfiguration message, etc.
  • steps 4-7 can be performed simultaneously with steps 3, 8, and 9-11.
  • the first base station checks the configured AI/ML model area information list. If the candidate second base station is within the configured AI/ML model area information list, indicating that the terminal-side AI/ML is applicable to the second base station, there is no need to perform model control. At this time, the first base station may also send an indication message to the terminal, which is used to indicate to the terminal that the terminal-side AI/ML is applicable to the second base station and model control does not need to be performed, or the first base station may also send a traditional message to the terminal, which includes an indicator, which is used to indicate to the terminal that the terminal-side AI/ML is applicable to the second base station and model control does not need to be performed.
  • FIG. 11 illustrates a signaling diagram of a fast model control procedure during a handover procedure according to some embodiments of the present disclosure.
  • the model control e.g., model selection, model switching, model update, etc.
  • the base station e.g., gNB
  • step 1 the terminal receives a traditional measurement configuration, where the traditional measurement configuration is provided via dedicated signaling (eg, RRCReconfiguration, RRCResume, etc.).
  • dedicated signaling eg, RRCReconfiguration, RRCResume, etc.
  • step 2 the terminal sends a measurement report message to the first base station (such as Source gNB, S-gNB) according to the measurement configuration in step 1.
  • the first base station such as Source gNB, S-gNB
  • step 3 based on the report message in step 2, the first base station initiates a handover HO decision.
  • step 4 the first base station checks the configured AI/ML model area information list. If the candidate second base station (such as Target gNB, T-gNB) is within the configured AI/ML model area information list, there is no need to perform model control. The first base station executes step 6. At this time, the first base station can send the AI/ML model information to the second base station and carry it in the handover request message of step 6. If the candidate second base station is not within the configured AI/ML model area information list, the first base station executes step 4', step 5.
  • the candidate second base station such as Target gNB, T-gNB
  • step 4' Alternatively, the first base station initiates an AI/ML model area information update process, the specific details of which are shown in Example 2.
  • step 5 the terminal performs model control according to step 4', such as selecting/switching one or more suitable AIML models.
  • step 6 the first base station sends a handover request message to the second base station, the message may carry model information, and the model information indicates an AI/ML model applicable to the second base station.
  • This step can be performed in advance between the terminal and the second base station.
  • step 7 the second base station performs admission model control.
  • step 8 the second base station sends a confirmation message about the handover request to the first base station, which carries information about the second base station and resource information of the terminal.
  • step 9 the first base station sends a message with a handover command (such as an RRCReconfiguration message, etc.) to the terminal, which carries information for accessing the second base station, such as a cell ID.
  • a handover command such as an RRCReconfiguration message, etc.
  • step 7 carries activation/deactivation configuration information, including indicating the activation/deactivation start time, bias, duration and end time of the model.
  • step 7 carries activation/deactivation configuration information, including indicating the activation/deactivation start time, bias, duration and end time of the model.
  • the model activation/deactivation indication information is carried.
  • model information can be transmitted through the message in step 9 (such as RRCReconfiguration message, etc.) It can be sent through a dedicated RRC message, MAC CE or DCI after step 4.
  • the present disclosure describes an example of communication between a terminal and a network element component in a network architecture in the above embodiments, which is mainly for illustrative purposes and not restrictive.
  • any of the components, modules, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuits), manual processing, or any combination thereof.
  • Some operations of the example methods can be described in the general context of executable instructions stored on a computer-readable memory locally and/or remotely of a computer processing system, and implementations can include software applications, programs, functions, and the like.
  • any function described herein can be performed at least in part by one or more hardware logic components, such as, but not limited to, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), an application specific standard product (ASSP), a system on a chip (SoC), a complex programmable logic device (CPLD), and the like.
  • FPGA field programmable gate array
  • ASIC application specific integrated circuit
  • ASSP application specific standard product
  • SoC system on a chip
  • CPLD complex programmable logic device
  • the signaling transmission described in the embodiments of the present disclosure can be implemented in any manner known in the art.
  • the signaling transmission can be explicit and/or implicit.
  • the illustrated steps (signaling/frames) are only for illustrative purposes and are not intended to limit the present application.
  • a control method of an artificial intelligence/machine learning AI/ML model in a wireless communication system comprising:
  • the wireless communication node obtains AI/ML model area information, and the wireless communication node determines model control of the AI/ML model based on the AI/ML model area information.
  • the wireless communication node is a base station or a terminal.
  • the model control of the AI/ML model includes at least one of the following:
  • AI/ML model activation AI/ML model deactivation
  • AI/ML model selection AI/ML model selection
  • AI/ML model switching AI/ML model switching
  • AI/ML model rollback AI/ML model update
  • AI/ML model training AI/ML model retraining
  • AI/ML model fine-tuning AI/ML model fine-tuning.
  • the AI/ML model region information includes at least one of the following:
  • the location information includes at least one of the following:
  • Cell information AI/ML model area identification ID, terminal location information, base station location information, scene information, control unit CU ID, data unit DU ID.
  • the scene information includes at least one of the following:
  • Transmission environment information For example, speed information, angle spread information, delay spread information, signal-to-noise ratio, channel received power, and/or interference information.
  • the wireless communication node capability information includes at least one of the following:
  • AI/ML model supports AI/ML usage scenarios and/or AI/ML model supports AI/ML features.
  • some specific computing capacity, some specific storage capacity, AI/ML model supports some specific capacity, some specific load conditions, some specific transmission capacity, AI/ML model supports some specific usage scenarios and/or supports some specific features.
  • Wireless network capability node capability information indicates the capabilities of a wireless network node, such as the model operation capabilities of the AI/ML models supported by the node, the node's hardware capability information (such as storage capacity, computing power, battery capacity, etc.), the functionality/features of the AI/ML models supported by the node, the data of the AI/ML models supported by the node, the scenarios of the AI/ML models supported by the node, etc.
  • the AI/ML model information includes at least one of the following:
  • AI/ML model ID information information, and/or AI/ML model meta information.
  • the AI/ML model information includes at least one of the following:
  • AI/ML model ID information information, and/or AI/ML model meta information.
  • the AI/ML model region information is transmitted by at least one of:
  • Non-access stratum NAS message radio resource control RRC message, medium access control element MAC-CE message, downlink control information DCI message, uplink channel, downlink channel, system message, N2 message and/or Xn message.
  • determining, by the wireless communication node, model control of the AI/ML model based on the AI/ML model region information comprises at least one of the following:
  • the wireless communication node does not perform model control of the AI/ML model
  • the wireless communication node If the location information of the wireless communication node does not belong to the location information indicated in the AI/ML model area information, the wireless communication node performs model control of the AI/ML model;
  • the wireless communication node does not perform model control of the AI/ML model
  • the wireless communication node If the AI/ML model information of the wireless communication node does not belong to the AI/ML model information indicated in the AI/ML model area information, the wireless communication node performs model control of the AI/ML model;
  • the model control of executing the AI/ML model includes at least one of the following:
  • AI/ML model activation AI/ML model deactivation
  • AI/ML model selection AI/ML model selection
  • AI/ML model switching AI/ML model switching
  • AI/ML model rollback AI/ML model update
  • AI/ML model training AI/ML model retraining
  • AI/ML model fine-tuning reporting of results of model control of the AI/ML model, and/or response of model control of the AI/ML model.
  • the terminal acquires the AI/ML model area configuration information, and the terminal performs model control of the AI/ML model according to the AI/ML model area information.
  • the terminal acquiring the AI/ML model area configuration information includes at least one of the following:
  • AI/ML model information reasons, indicators, terminal ID, and/or location information.
  • the base station acquires the AI/ML model area configuration information, and the base station performs model control of the AI/ML model according to the AI/ML model area information.
  • the base station acquiring the AI/ML model area configuration information includes at least one of the following:
  • AI/ML model information reasons, indicators, base station ID, and/or location information.
  • the AI/ML model region information request message and/or response message sent by the first wireless communication node to the second wireless communication node includes at least one of the following:
  • Wireless communication node capability information and/or
  • the wireless communication node sends the request message for the AI/ML model region information, including periodic transmission or event triggering.
  • the AI/ML model region information is updated in response to a periodic configuration, wherein the periodic configuration includes a duration, a start time, and/or an end time of the update.
  • the AI/ML model region information is updated in response to an event triggered configuration.
  • the periodic transmission and/or the event triggering corresponding periodic configuration and/or event triggering configuration are preconfigured, and/or configured, and/or fixed.
  • the AI/ML model region information request message and/or response message is carried by at least one of the following:
  • Non-access stratum NAS message radio resource control RRC message, medium access control element MAC-CE message, downlink control information DCI message, uplink channel, downlink channel, system message, N2 message and/or Xn message.
  • a wireless communication method for a terminal comprising:
  • the terminal acquires AI/ML model area information from the first base station, where the AI/ML model area information is used by the terminal to determine model control of the AI/ML model,
  • the terminal determines whether to execute model control of the AI/ML model based on the AI/ML model area information.
  • the terminal obtains an AI/ML model area configuration message from the first base station.
  • the terminal determines to perform model control of the AI/ML model.
  • the performing model control of the AI/ML model includes updating the AI/ML model region information between the terminal and the first base station.
  • the terminal sends a report message for the AI/ML model area configuration message and a result of executing the model control of the AI/ML model to the first base station, and the first base station sends the obtained result of executing the model control of the AI/ML model to the second base station.
  • a wireless communication method for a first base station comprising:
  • the first base station obtains a model control reporting message of the AI/ML model from the terminal, where the model control reporting message of the AI/ML model includes a result of the model control of the AI/ML model;
  • the first base station sends the cell switching request of the second base station and the result of the model control of the AI/ML model to the second base station.
  • a wireless communication method for a first base station comprising:
  • the first base station determines, according to a model control message of the AI/ML model fed back by the terminal, whether the second base station is in the AI/ML model area information;
  • the first base station executes the AI/ML model model. Type control.
  • the model control of the AI/ML model includes updating the AI/ML model region information between the first base station and the second base station.
  • the method further includes: the first base station sending the cell switching request of the second base station and the result of the model control to the second base station.
  • it also includes: if the second base station is in the AI/ML model area information, the first base station sends the AI/ML model information contained in the AI/ML model area information and a cell switching request to the second base station to the second base station.
  • a chip including: a processor configured to call and execute a computer program stored in a memory so that a device in which the chip is installed performs any one of the methods in the example embodiments.
  • FIG12 is a block diagram of an example system 700 for wireless communication according to an embodiment of the present disclosure.
  • the embodiments described herein may be implemented into a system using any appropriately configured hardware and/or software.
  • FIG12 illustrates a system 700, including a radio frequency (RF) circuit 710, a baseband circuit 720, a processing unit 730, a memory/storage 740, a display 750, a camera 760, a sensor 770, and an input/output (I/O) interface 780, coupled to each other as shown.
  • RF radio frequency
  • the processing unit 730 may include circuits, such as but not limited to one or more single-core or multi-core processors.
  • the processor may include any combination of a general-purpose processor and a special-purpose processor, such as a graphics processor and an application processor.
  • the processor may be coupled to a memory/storage and configured to execute instructions stored in the memory/storage to enable various applications and/or operating systems running on the system.
  • the RF circuit 710, the baseband circuit 720, the processing unit 730, the memory/storage 740, the display 750, the camera 760, the sensor 770, and the I/O interface 780 are well-known elements in the system 700, such as but not limited to laptop computing devices, tablet computing devices, netbooks, ultra-extreme notebooks, smart phones, etc.
  • instructions as software products may be stored in a readable storage medium in a computer.
  • the software product in the computer is stored in a storage medium, including multiple commands for a computing device (such as a personal computer, a server, or a network device) to run all or some of the steps disclosed in the embodiments of the present disclosure.
  • the storage medium includes a USB disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a floppy disk, or other types of media capable of storing program codes.
  • the embodiments of the present disclosure are a combination of techniques/processes that may be employed in 3GPP specifications to create a final product.

Landscapes

  • Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本公开提供一种无线通信系统中的人工智能/机器学习AI/ML模型的控制方法,包括:由无线通信系统中的节点获取AI/ML模型区域信息,节点包括终端或基站中的至少一者;由节点基于AI/ML模型区域信息来确定AI/ML模型的模型控制。本公开还提供一种用于终端的无线通信方法,包括:所述终端从第一基站获取AI/ML模型区域信息,所述AI/ML模型区域信息用于所述终端确定AI/ML模型的模型控制,其中,所述终端基于所述AI/ML模型区域信息确定是否执行所述AI/ML模型的模型控制。

Description

无线通信系统中的AI/ML模型的控制方法 技术领域
本公开涉及通信系统领域,更具体地,涉及一种用于无线通信系统中的人工智能/机器学习(AI/ML)模型的控制方法及其芯片、计算机可读存储介质和计算机程序产品、用于用户装备(终端)的无线通信方法及其芯片、计算机可读存储介质和计算机程序产品、以及用于第一基站的无线通信方法及其芯片、计算机可读存储介质和计算机程序产品。
背景技术
我们的社会正在经历数字化革命,在过去几年中,人工智能(AI)和机器学习(ML)方法被广泛应用于各个行业,以推动创新和提高流程效率。然而,在网络设计方面,随着数据量和网络复杂性的急剧增加,传统方法在许多情况下将无法提供快速解决方案。因此,AI/ML将成为推动未来无线通信网络性能改进的不可或缺的技术。
实际上,在通过空中接口使用AI/ML的无线通信系统中,由于用户装备(终端)移动性,信道可能变得更差,从而可能导致切换发生。当终端移动到新的服务基站(如gNB)时,系统可以在某些用例中使用传统方法或基于AI/ML的方法。如果采用基于AI/ML的方法,为了确保服务的连续性,在切换过程中可能需要执行AI/ML模型的模型控制(例如,模型选择、切换、更新等等)。但是,目前的无线通信网络并不具备支持此类模型控制的能力。
需要注意,在此章节中描述的内容不构成或不被视为对任何现有技术的承认。
发明内容
为了克服上述技术问题中的至少一个,本公开提供了一种用于无线通信系统中的人工智能/机器学习(AI/ML)模型的控制方法,其能够确保在切换过程中模型控制正常工作、减少信令开销,并且减少动态信息上报的延迟。
本公开的一个方面提供了一种无线通信系统中的人工智能/机器学习AI/ML模型的控制方法,包括:
所述无线通信节点获取AI/ML模型区域信息,无线通信节点基于所述AI/ML模型区域信 息确定AI/ML模型的模型控制,所述无线通信节点为基站或终端。
根据本公开的一个或多个实施例,所述AI/ML模型的模型控制包括以下各项中的至少一者:
AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调。
根据本公开的一个或多个实施例,所述AI/ML模型区域信息包括以下各项中的至少一者:
AI/ML模型区域标识ID,
位置信息,
场景信息,
无线通信节点能力信息,
和/或AI/ML模型信息。
根据本公开的一个或多个实施例,所述位置信息包括以下各项中的至少一者:
小区信息、AI/ML模型区域标识ID、终端位置信息、基站位置信息,场景信息、控制单元CU ID、数据单元DU ID。
根据本公开的一个或多个实施例,所述场景信息包括以下各项中的至少一者:
传输环境信息,速度信息,角度扩展信息,时延扩展信息,信噪比,信道接收功率,和/或干扰信息。
根据本公开的一个或多个实施例,所述无线通信节点能力信息包括以下各项中的至少一者:
计算能力,存储能力,AI/ML模型支持能力,负载情况,和/或传输能力,AI/ML模型支持AI/ML使用场景和/或AI/ML模型支持AI/ML特征。详细而言,在一些示例中例如一些特定计算能力,一些特定存储能力,AI/ML模型支持一些特定能力,一些特定负载情况,一些特定传输能力,AI/ML模型支持一些特定使用场景和/或支持支持一些特定特征。
无线网络能力节点能力信息;指示一个无线网络节点的能力,如节点支持的AI/ML模型的模型操作能力、节点的硬件能力信息(如存储能力,算力,电池能力等)、节点支持的AI/ML模型的功能性/特征,节点支持的AI/ML模型的数据,节点支持的AI/ML模型的场景等。
根据本公开的一个或多个实施例,所述AI/ML模型信息包括以下各项中的至少一者:
AI/ML模型ID信息,和/或AI/ML模型元信息。
根据本公开的一个或多个实施例,所述AI/ML模型区域信息通过以下各项中的至少一者传输:
非接入层NAS消息、无线电资源控制RRC消息、介质接入控制控制元素MAC-CE消息、 下行链路控制信息DCI消息、上行链路信道、下行链路信道、系统消息、N2消息和/或Xn消息
根据本公开的一个或多个实施例,由所述无线通信节点基于所述AI/ML模型区域信息确定AI/ML模型的模型控制包括以下各项中的至少一者:
如果所述无线通信节点的位置信息属于所述AI/ML模型区域信息中所指示的位置信息,则所述无线通信节点不执行AI/ML模型的模型控制;
如果所述无线通信节点的位置信息不属于所述AI/ML模型区域信息中所指示的位置信息,则所述无线通信节点执行AI/ML模型的模型控制;
如果所述无线通信节点的AI/ML模型信息属于所述AI/ML模型区域信息中所指示的AI/ML模型信息,则所述无线通信节点不执行AI/ML模型的模型控制;
如果所述无线通信节点的AI/ML模型信息不属于所述AI/ML模型区域信息中所指示的AI/ML模型信息,则所述无线通信节点执行AI/ML模型的模型控制.
根据本公开的一个或多个实施例,所述执行AI/ML模型的模型控制包括以下各项中的至少一者:
所述执行AI/ML模型的模型控制包括以下各项中的至少一者:
AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调、AI/ML模型的模型控制的结果上报,和/或AI/ML模型的模型控制的响应。
根据本公开的一个或多个实施例,当所述无线通信节点为终端时,所述终端获取所述AI/ML模型区域配置信息,所述终端根据所述AI/ML模型区域信息执行所述AI/ML模型的模型控制。
根据本公开的一个或多个实施例,所述终端获取所述AI/ML模型区域配置信息包括以下各项中的至少一者:
AI/ML模型信息、事由、指示符、终端ID、和/或位置信息。
根据本公开的一个或多个实施例,当所述无线通信节点为基站时,所述基站获取所述AI/ML模型区域配置信息,所述基站根据所述AI/ML模型区域信息执行所述AI/ML模型的模型控制。
根据本公开的一个或多个实施例,所述基站获取所述AI/ML模型区域配置信息包括以下各项中的至少一者:
AI/ML模型信息、事由、指示符、基站ID,和/或位置信息。
根据本公开的一个或多个实施例,第一无线通信节点向第二无线通信节点发送的所述AI/ML模型区域信息请求消息和/或响应消息包括以下各项中的至少一者:
AI/ML模型区域标识ID,
位置信息,
场景信息,
无线通信节点能力信息,和/或
AI/ML模型信息。
根据本公开的一个或多个实施例,所述无线通信节点发送所述AI/ML模型区域信息的请求消息,包括周期性发送或事件触发。
根据本公开的一个或多个实施例,在所述周期性发送的情况下,所述AI/ML模型区域信息响应于周期性配置而进行更新,其中,所述周期性配置包括所述更新的持续时间、起始时间、和/或结束时间。
根据本公开的一个或多个实施例,在所述方式为事件触发的情况下,所述AI/ML模型区域信息响应于事件触发配置而进行更新。
根据本公开的一个或多个实施例,所述周期性发送和/或所述事件触发相应的周期性配置和/或事件触发配置是预配置的、和/或配置的、和/或固定的。
根据本公开的一个或多个实施例,所述AI/ML模型区域信息请求消息和/或响应消息通过以下至少一项承载:
非接入层NAS消息、无线电资源控制RRC消息、介质接入控制控制元素MAC-CE消息、下行链路控制信息DCI消息、上行链路信道、下行链路信道、系统消息、N2消息和/或Xn消息。
本公开的一个方面提供了一种用于终端的无线通信方法,包括:
所述终端从第一基站获取AI/ML模型区域信息,所述AI/ML模型区域信息用于所述终端确定AI/ML模型的模型控制,
其中,所述终端基于所述AI/ML模型区域信息确定是否执行所述AI/ML模型的模型控制。
根据本公开的一个或多个实施例,还包括:
所述终端从所述第一基站获取AI/ML模型区域配置消息。
根据本公开的一个或多个实施例,如果第二基站不在所述AI/ML模型区域的情况下,所述终端确定执行所述AI/ML模型的模型控制。
根据本公开的一个或多个实施例,所述执行AI/ML模型的模型控制包括在所述终端与所 述第一基站之间更新所述AI/ML模型区域信息。
根据本公开的一个或多个实施例,还包括:
所述终端将针对所述AI/ML模型区域配置消息的上报消息以及所述执行所述AI/ML模型的模型控制的结果发送至所述第一基站,所述第一基站将获取的执行所述AI/ML模型的模型控制的结果发送至所述第二基站。
本公开的一个方面提供了一种用于第一基站的无线通信方法,包括:
所述第一基站从所述终端获取AI/ML模型的模型控制上报消息,所述AI/ML模型的模型控制上报消息包括所述AI/ML模型的模型控制的结果;
所述第一基站将所述第二基站的小区切换请求以及所述AI/ML模型的模型控制的结果发送至所述第二基站。
本公开的一个方面提供了一种用于第一基站的无线通信方法,包括:
所述第一基站根据终端反馈的AI/ML模型的模型控制消息,确定第二基站是否在AI/ML模型区域信息中;
如果所述第二基站不在所述AI/ML模型区域信息中,所述第一基站执行AI/ML模型的模型控制。
根据本公开的一个或多个实施例,所述AI/ML模型的模型控制包括在所述第一基站与所述第二基站之间更新所述AI/ML模型区域信息。
根据本公开的一个或多个实施例,还包括:
所述第一基站将所述第二基站的小区切换请求以及所述模型控制的结果发送至所述第二基站。
根据本公开的一个或多个实施例,还包括:
如果所述第二基站在所述AI/ML模型区域信息中,所述第一基站将所述AI/ML模型区域信息中所包含的AI/ML模型信息以及至所述第二基站的小区切换请求发送至所述第二基站。
本公开的一个方面提供了一种芯片,包括:
处理器,其被配置为调用和运行存储在存储器中的计算机程序,以使得其中安装有所述芯片的设备执行示例实施例中任一个的方法。
本公开的一个方面提供了一种芯片,包括:处理器,其被配置为调用和运行存储在存储器中的计算机程序,以使得其中安装有所述芯片的设备执行本公开任一方面的实施例的方法。
本公开的一个方面提供了一种计算机可读存储介质,其中存储有计算机程序,其中所述计算机程序使得计算机执行本公开任一方面的实施例的方法。
本公开的一个方面提供了一种计算机程序产品,包括计算机程序,其中所述计算机程序使得计算机执行本公开任一方面的实施例的方法。
附图说明
为了更清楚地说明本公开或相关技术的实施例,将在实施例中简要介绍以下附图。显然,附图仅仅是本公开的一些实施例,本领域的普通技术人员可以在不付出创造性劳动的前提下根据这些附图获得其他附图。
图1图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图2图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图3图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图4图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图5图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图6图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图7图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图8图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
图9图示了根据本公开的一些实施例的在切换过程期间的快速模型控制过程的信令图。
图10图示了根据本公开的一些实施例的在切换过程期间的快速模型控制过程的信令图。
图11图示了根据本公开的一些实施例的在切换过程期间的快速模型控制过程的信令图。
图12是根据本公开的实施例的用于无线通信的示例系统的框图。
具体实施方式
本公开的实施例参照所述附图详细描述了技术事项、结构特征、实现的目的和效果,如下文所描述。具体地,本公开的实施例中的术语仅用于描述特定实施例的目的,而不是限制本公开。
在本公开中,“A或B”可以意指“仅A”、“仅B”或“A和B二者”。
换句话说,在本公开中,“A或B”可以被解释为“A和/或B”。例如,在本公开中,“A、B或C”可以意指“仅A”、“仅B”、“仅C”或“A、B、C的任何组合”。
在本公开中使用的斜杠(/)或逗号可以意指“和/或”。例如,“A/B”可以意指“A和/或B”。因此,“A/B”可以意指“仅A”、“仅B”或“A和B二者”。例如,“A、B、C”可以意指“A、B或C”。
在本公开中,“A和B中的至少一个”可以意指“仅A”、“仅B”或“A和B二者”。另外,在本公开中,表述“A或B中的至少一个”或“A和/或B中的至少一个”可以被解释为“A和B中的至少一个”。
另外,在本公开中,“A、B和C中的至少一个”可以意指“仅A”、“仅B”、“仅C”或“A、B和C的任何组合”。另外,“A、B或C中的至少一个”或“A、B和/或C中的至少一个”可以意指“A、B和C中的至少一个”。
此外,术语“第一”、“第二”仅用于描述目的,而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此,限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者多个所述特征。在本申请的描述中,“多个”的含义是两个或两个以上,除非另有明确具体的限定。
在通过空中接口使用AI/ML的无线通信系统中,由于终端移动性,信道可能变得更差,从而可能导致切换发生。当终端移动到新的服务基站(如gNB)时,系统可以在某些用例中使用传统方法或基于AI/ML的方法。如果采用基于AI/ML的方法,为了确保服务的连续性,在切换过程中可能需要执行AI/ML模型的模型控制(例如,AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调等等,本公开对此不进行任何限制)。但是,目前的无线通信网络并不具备支持此类模型控制的能力。
根据协议的进展,判断终端侧的AI/ML模型是否需要进行模型控制(也可称为模型操作、模型管理等)主要基于模型监听的结果。也就是说,在AI/ML模型的生命周期管理过程中,诸如模型激活、去激活、选择、切换、回退、更新、训练、重训练等AI/ML模型的模型控制是在模型开启监听后根据监听结果做出的决策。在切换过程中,如果采用模型监听的方案来判断当前的AI/ML模型是否适配于第二基站,需要第二基站配置模型监听和监听上报。然而,这个配置过程通常需要在终端接入第二基站后,也就是切换完成后才能进行模型监听的判断。而且,要得到准确的模型监听结果,需要通过多次基于AI的测量和传统测量来进行模型监听。
此外,在不同的场景下,对于终端侧的模型,当模型监听位于终端侧,监听决策位于网络(例如但不限于基站(如gNB)等)侧时,终端需要上报监听结果以辅助网络侧进行新模型的选择,网络侧然后将新选择的模型下发给终端。当模型监听位于终端侧,监听决策也位于终端侧时,终端需要根据决策重新选择模型。
因此,对于切换来说,如果采用模型监听的方式,需要在终端接入新的第二基站后才能进行,并且整个模型监听到模型控制的过程会占用很长时间,给整个系统带来较大的延迟,无法 满足对服务连续性的要求。
另一种不采用模型监听的可选方案,即在发生小区切换时,为了使终端的AI/ML模型适配新的网络侧(如第二基站),可以直接进行AI模型的模型控制。例如新切换/选择/更新的AI模型可以由第二基站选择后直接传输给终端,或者由第二基站配置新模型的相关信息(如模型标识符(ID)、模型的元信息等)发送给终端,以辅助终端侧选择/切换新的AI模型。然而,这种方法会导致基站和终端之间需要直接交互AI模型或所需的AI模型信息,从而产生较大的信令开销。此外,考虑到新模型的传输和配置也是在终端接入新的第二基站后进行的,整个系统会面临较大的延迟,无法满足对服务连续性的要求。
根据RAN1和RAN2的进展,为了降低由于AI模型控制引起的切换时延问题,一些公司提出了在切换准备过程中交互模型信息的方法。通过这种方法,在切换可能发生时,可以使终端在切换后快速切换到正确的新的AI/ML上下文。虽然这种方法可以有效降低时延,但是执行模型控制的条件并不清楚。此外,在准备阶段存在大量的交互,如果当前的AI/ML模型仍然适用,那么会造成信令开销的浪费。
基于上述讨论,发明人洞察到设计一种更简单的方法来实现快速的模型控制是亟需且有实际意义的。
因而,本公开提出了一种用于无线通信系统中的人工智能/机器学习(AI/ML)模型的控制方法。该方法可以用于确定在切换过程中是否需要执行模型控制,并且可以用于在发生模型控制时辅助节点例如快速选择/切换/更新一个或多个合适的AI/ML模型。借助于本公开的方法,能够有效降低由于模型监听以及对AI/ML模型进行模型控制等所带来的延迟;并且能够降低执行模型控制时的信令交互开销。
在一些实施例中,终端可以为用户站、移动站、移动台、无线通信设备、用户代理等各种用户设备。此外,终端的一些示例可以为手机、平板电脑、笔记本电脑、掌上电脑、移动互联网设备、可穿戴设备、虚拟现实设备、增强现实设备等,并且本公开对此并不进行任何限定。
在一些实施例中,基站可以包括或可以被称为基站收发机、无线基站、接入点、无线收发机、节点B、e节点B(eNB)、下一代节点B或千兆节点B(其中任一个可以被称为gNB)、家庭节点B、家庭e节点B或其他合适的术语。无线通信系统可以包括不同类型的基站(例如,宏小区基站或小型小区基站等)。终端可以与各种基站和/或网络设备通信,包括宏eNB、小型小区eNB、gNB、中继基站等。
提供了一种AI/ML模型特定的区域信息,其可用于辅助节点(如终端、gNB等)执行AI/ML模型控制。其中,一个AI/ML模型可以用于特定AI/ML区域中的一个或多个位置。在一些实 施例中,AI/ML模型区域信息可以至少包括以下内容之一:AI/ML模型区域标识ID,位置信息,场景信息,无线通信节点能力信息,和/或AI/ML模型信息。在一些实施例中,AI/ML模型区域信息可以通过RRC消息、NAS消息、N2消息、Xn消息、DCI、MAC CE,系统消息,上行链路信道、下行链路信道等配置给节点,如终端和/或基站等。
部分本方面实施例出现的术语解释如下:
模型控制:也可称为模型操作、模型管理等,指的是生命周期管理(LCM)中的一些模型操作,例如模型选择、模型切换、模型更新、模型激活、模型去激活(模型停用)、模型回退、模型保留(微调)、模型训练、模型重训练等,本公开对此不进行任何限制;终端侧(端)模型:可以指位于终端侧(端)的AI/ML模型,包括终端侧模型之情形和双侧模型中的终端部分之情形;
AI/ML模型区域信息:是指AI特定区域(AI-specific area),在本公开中,也可以被可互换地指称为AI/ML模型特定区域,AI/ML模型区域信息与AI/ML模型相对应或相关联;
AI/ML模型区域指示符:用于标识AI特定区域;
AI/ML模型特定区域的特定区域代码(AI-specific area code of AI-specific area):可以用于指示AI特定区域的AI特定区域代码;
AI/ML模型信息:可以用于指示AI/ML模型的信息,例如模型ID、模型元信息等,此外,当没有特别指出时,AI/ML模型信息可以是一个AI/ML模型或多个AI/ML模型的信息;位置信息:小区信息、AI/ML模型区域标识ID、终端位置信息、基站位置信息,场景信息、控制单元CU ID、数据单元DU ID等;其中,
根据本公开的一个或多个实施例,所述无线通信节点能力信息包括以下各项中的至少一者:
计算能力,存储能力,AI/ML模型支持能力,负载情况,和/或传输能力,AI/ML模型支持AI/ML使用场景和/或AI/ML模型支持AI/ML特征。详细而言,在一些示例中例如一些特定计算能力,一些特定存储能力,AI/ML模型支持一些特定能力,一些特定负载情况,一些特定传输能力,AI/ML模型支持一些特定使用场景和/或支持支持一些特定特征。
无线网络能力节点能力信息;指示一个无线网络节点的能力,如节点支持的AI/ML模型的模型操作能力、节点的硬件能力信息(如存储能力,算力,电池能力等)、节点支持的 AI/ML模型的功能性/特征,节点支持的AI/ML模型的数据,节点支持的AI/ML模型的场景等。
基站信息:可以指示小区ID、传输接收点(TRP)ID、中央单元(CU)ID和/或分布单元(DU)ID等;
场景信息:可以传输环境信息,速度信息,角度扩展信息,时延扩展信息,信噪比,信道接收功率,和/或干扰信息等反映信道特征的一个或多个参数;
终端位置信息:可以包括例如终端ID、与终端相关联的小区ID、终端的物理位置区域、经度、纬度、海拔高度等。
模型ID:用于标识AI/ML模型,包括全局模型ID和/或本地模型ID,其中:
全局AI/ML模型ID用于标识AI/ML模型,其可以是全球唯一ID、PLMN特定唯一ID、运营商特定唯一ID或AI/ML管理平台特定唯一ID。
本地AI/ML模型ID用于标识网络中使用的AI/ML模型,并与全局ID存在一定的映射关系。
逻辑ID可以是全球唯一ID、PLMN特定唯一ID、运营商特定唯一ID、小区特定ID、链路特定ID、跟踪区域(TA)特定ID、CU特定ID、DU特定ID、用户面功能(UPF)特定ID、接入和移动性管理功能(AMF)特定ID、无线电资源控制(RRC)特定ID或网络切片特定ID。
术语“小区特定”表示每个AI/ML模型在特定小区内具有唯一ID。类似地,“链路特定”、“TA特定”、“CU特定”、“DU特定”、“UPF特定”、“AMF特定”和“网络切片特定”等指的是网络中特定上下文中AI/ML模型的唯一ID。
模型ID也可以通过隐式方式指示,例如RRC ID、DCI ID等。
模型元信息:用于描述AI/ML模型的信息,可以包括模型功能性/特征、模型版本信息、模型格式信息、供应商信息、计算复杂度、模型复杂度、模型性能、适用条件等。其中,
模型功能性/特征:指示一个模型的功能性,例如模型用于CSI压缩、波束预测、定位等。指示应该是子用例级别或甚至更细的级别。
模型版本信息:指示AI/ML模型的版本信息。
模型格式信息:指示AI/ML模型的格式信息,例如开放格式(如ONNX)、私有格式等。
供应商信息:指示AI/ML模型训练供应商的信息。
计算复杂度:用于指示模型训练的计算复杂度,例如FLOP(浮点运算数)、预处 理/后处理的级别。
模型复杂度:用于指示模型训练的复杂度,包括实值模型参数的数量、实值的数量等。
模型性能:用于指示模型的性能,其可以是模型准确度、模型偏差、模型方差等。
适用条件:用于指示AI/ML模型或功能性的适用条件,例如场景、配置或站点、终端内部能力等。
场景信息包括以下各项中的至少一者:传输环境信息,速度信息,角度扩展信息,时延扩展信息,信噪比,信道接收功率,和/或干扰信息,其中,
传输环境信息:用于描述当前无线通信节点所处的信道传输环境,比如处于室内/城市/城市郊区/农村/隧道等,或者信道分类信息,比如基于UMA/UM i/InH/RMA的分类信息等。
速度信息:用于描述当前无线通信节点所处的移动速度,该移动速度可以是终端测量得到的,也可以是基站测量得到的。
角度扩展信息:用于描述当前无线通信节点所出的信道的角度扩展情况,即信道的角度扩展实时信息或者统计信息,或者基于角度扩展相关联的其他信息。
时延扩展信息:用于描述当前无线通信节点所出的信道的时延扩展情况,即信道的时延扩展实时信息或者统计信息,或者基于时延扩展相关联的其他信息。
根据本公开的一个或多个实施例,所述无线通信节点能力信息包括以下各项中的至少一者:
计算能力,存储能力,AI/ML模型支持能力,负载情况,和/或传输能力,AI/ML模型支持AI/ML使用场景和/或AI/ML模型支持AI/ML特征。详细而言,在一些示例中例如一些特定计算能力,一些特定存储能力,AI/ML模型支持一些特定能力,一些特定负载情况,一些特定传输能力,AI/ML模型支持一些特定使用场景和/或支持支持一些特定特征。
无线网络能力节点能力信息;指示一个无线网络节点的能力,如节点支持的AI/ML模型的模型操作能力、节点的硬件能力信息(如存储能力,算力,电池能力等)、节点支持的AI/ML模型的功能性/特征,节点支持的AI/ML模型的数据,节点支持的AI/ML模型的场景等。
根据本公开的一个或多个实施例,所述AI/ML模型信息包括以下各项中的至少一者:
AI/ML模型ID信息,和/或AI/ML模型元信息。
如本公开所使用的,术语“模型控制”也可称为模型操作、模型管理等。
如本公开所使用的,当提到表述“不发起”时,可以指禁止执行、不执行、放弃执行等。
AI/ML模型区域信息配置
AI特定区域信息被引入用于辅助节点(如终端、记账等)正确执行AI/ML模型控制,和/或在模型控制发生时帮助节点例如选择/切换/更新一个或多个合适的AI/ML模型。
关于本公开所提出的AI/ML模型区域信息,以下问题和相关方法列举如下:
AI/ML模型区域信息可以由以下的方式构成。
AI/ML模型区域信息的组成包括以下各项中的至少一者:AI/ML模型区域标识ID,位置信息,场景信息,无线通信节点能力信息,和/或AI/ML模型信息.几种可能的选项如下
选项1:AI/ML模型区域信息可以由单个或多个位置信息构成;
选项2:AI/ML模型区域信息可以由单个或多个位置信息和一些AI/ML模型信息构成;
选项3:AI/ML模型区域信息可以由单个或多个位置信息,和一些AI/ML模型信息,和一些场景信息构成。
AI/ML模型区域标识ID,位置信息,场景信息,无线通信节点能力信息,和/或AI/ML模型信息任意两项组合及以上存在映射关系,所述映射关系可以预先配置或通过例如RRC消息、NAS消息、DCI消息等进行配置,也可以在特定情况下被固定。在一种实现方式种,所述AI/ML模型信息包括以下各项中的至少一者:AI/ML模型ID信息,和/或AI/ML模型元信息。
AI/ML模型区域信息与AI/ML模型信息之间的映射可以是一对一、一对多或多对一的关系。需要注意,AI/ML模型区域信息中的任意一个或多个元素与AI/ML模型信息中的任意一个或多个元素之间都可以存在映射关系,本公开对此不进行任何限制。
在一些示例中,可以使用AI/ML模型区域指示符表示位置信息和AI/ML模型信息之间的映射。表1展示了AI/ML模型区域信息与AI/ML模型信息之间的一种可能的映射关系,其中位置信息是小区ID,模型信息是模型ID;表2展示了AI/ML模型区域信息与AI/ML模型信息之间的另一种可能的映射关系,其中位置信息是小区ID,模型信息是模型特征和/或功能性、模型格式。
表1:AI/ML模型区域信息与AI/ML模型信息之间的映射关系

表2:AI/ML模型区域信息与AI/ML模型信息之间的映射关系
在一些示例中,AI/ML模型区域信息与AI/ML模型信息之间的一种可能映射关系如下表3所示,其中位置信息是小区ID,模型信息是模型ID:
表3:AI/ML模型区域信息与AI/ML模型信息之间的映射关系
在一些示例中,AI/ML模型区域信息与AI/ML模型信息之间的一种可能映射关系如下表4所示,其中位置信息是小区ID,模型信息是模型ID和AI特征:
表4:AI/ML模型区域信息与AI/ML模型信息之间的映射关系
在一些示例中,为了指示AI/ML模型区域信息与AI/ML模型信息之间的一种可能映射关系,可以引入一个索引,如下表5所示,其中位置信息是小区ID,模型信息是模型ID:
表5:AI/ML模型区域信息与AI/ML模型信息之间的映射关系

应当领会,上面描述的其中构成AI/ML模型区域信息的几种不同方式的示例仅出于说明性目的而被给出,本公开对此并不旨在进行任何限制。本领域技术人员在阅读上述示例后,容易想到其他合适的添加、删除、修改、变型等,而这些都落入本公开的范围。
关于AI/ML模型区域信息的配置,有若干种不同的选择:
在RRC激活、非激活或空闲状态下的终端(如UE)可以配置有AI/ML模型区域信息或AI/ML模型区域信息列表,可以通过以下选项进行配置:
选项1:核心网网元(CN,如AMF)通过例如NAS消息向终端发送AI/ML模型区域信息或AI/ML模型区域信息列表。本公开对所使用的消息类型不进行任何限制,并且不排除使用除NAS消息之外的任何其他合适的消息;
选项2:基站(如gNB)通过例如RRC消息向终端发送AI/ML模型区域信息或AI/ML模型区域信息列表,RRC消息例如包括RRC重配置消息、RRC释放消息或其他专用的RRC消息。本公开对所使用的消息类型不进行任何限制,并且不排除使用除RRC消息之外的任何其他合适的消息;
选项3:基站(如gNB)通过系统信息消息向终端发送AI/ML模型区域信息或AI/ML模型区域信息列表,包括SIB1或其他新定义的SIB。本公开对所使用的消息类型不进行任何限制,并且不排除使用除系统信息消息之外的任何其他合适的消息。
选项4:基站(如gNB)通过例如MAC CE消息向终端发送AI/ML模型区域信息或AI/ML模型区域信息列表;本公开对所使用的消息类型不进行任何限制,并且不排除使用除MAC CE消息之外的任何其他合适的消息;
选项5:基站(如gNB)通过例如DCI消息向终端发送AI/ML模型区域信息或AI/ML模型区域信息列表;本公开对所使用的消息类型不进行任何限制,并且不排除使用除DCI消息之外的任何其他合适的消息;
一个基站(如gNB)可以配置有AI/ML模型区域信息或AI/ML模型区域信息列表,可以通过以下选项进行配置:
选项1:核心网网元(如AMF)通过N2消息向基站(如gNB)发送AI/ML模型区域信息或AI/ML模型区域信息列表。本公开对所使用的消息类型不进行任何限制,并且不排除使 用除N2消息之外的任何其他合适的消息;
选项2:基站(如gNB)通过Xn消息向另一个基站(如gNB)发送AI/ML模型区域信息或AI/ML模型区域信息列表。本公开对所使用的消息类型不进行任何限制,并且不排除使用除Xn消息之外的任何其他合适的消息。
应当领会,上面描述的其中配置AI/ML模型区域信息的几种不同方式的示例仅出于说明性目的而被给出,本公开对此并不旨在进行任何限制。本领域技术人员在阅读上述示例后,容易想到其他合适的添加、删除、修改、变型等,而这些都落入本公开的范围。
关于AI/ML模型区域信息更新,可以考虑至少三种替代方案:
替代方案1:周期性请求更新,可以通过例如RRC消息、NAS消息、N2消息或系统消息,Xn消息,MAC CE,DCI等配置AI/ML模型区域信息更新周期,更新周期配置内容包括以下至少一个更新周期的持续时间、起始时间和结束时间,例如5分钟到720分钟。
替代方案2:事件触发请求更新,以下任一事件可能触发AI/ML模型区域信息更新:
事件1:AI/ML模型区域信息对应的模型不在配置的AI/ML模型区域信息中,将触发AI/ML模型区域信息更新;例如,如果模型发生变化(包括模型选择、模型切换、模型更新等),对应的AI/ML模型区域信息不属于配置的AI/ML模型区域信息。
事件2:候选基站和/或小区的位置信息不在配置的AI/ML模型区域信息中,将触发AI/ML模型区域信息更新。
事件3:当终端的小区重选过程选择了不属于配置的AI/ML模型区域信息的基站和/或小区,将触发AI/ML模型区域信息更新。
下面描述AI/ML模型区域信息更新的可能过程。
图1图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图1所示,图示了终端触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或周期触发。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括终端(如终端)向基站(如gNB)发送用于AI/ML模型区域信息更新的请求消息。该消息至少包括以下信息:
-模型信息:用于指示终端侧当前的AI模型,可以是处于激活态的AI模型,也可以处于非激活态的AI模型。
-事由:用于指示AI/ML模型区域信息更新的事由。在本例中,事由是终端侧当前AI 模型与AI/ML模型区域信息不属于配置的AI/ML模型区域信息列表。
-终端标识:用于标识特定的终端。
示例性的,该消息可以是专用的用于请求AI/ML模型区域信息更新的消息。一旦基站(如gNB)接收到该消息,基站(如gNB)就知道该消息是用于请求AI/ML模型区域信息更新的。或者,该消息可以是一个传统消息,其可包含一个指示符,用于指示该消息用于AI/ML模型区域信息更新请求。另外,该消息也可以是一个传统消息,其可包含一个事由,用于指示发起专用消息的原因。
示例性的,步骤2可以包括终端接收到更新的AI/ML模型区域信息。更具体地说,更新的AI/ML模型区域信息可以包括AI/ML模型区域指示符、AI/ML模型区域指示符列表、AI模型信息、位置信息或AI/ML模型区域信息与模型信息的映射规则/关系。通过接收这些信息,终端能够更新其AI/ML模型区域信息,并获得最新的AI模型信息和相关的位置信息和映射规则。这样,终端能够更好地适应当前的网络环境和需求,提供更高质量的服务。因此,本公开的实施例提供了一种有效的AI/ML模型区域信息更新方法,能够提升网络性能和用户体验。
需要注意,上述实施例仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对实施例进行各种修改和变化。
图2图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图2所示,图示了终端触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
在某些情况下,如果第一第一基站(例如,在基站为gNB的情况下为服务基站s-gNB)没有与请求的AI/ML模型区域信息对应的模型,则第一基站可以将请求转发给其他基站(例如,第二基站),如图2所示。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括终端向第一基站发送AI/ML模型区域信息更新的请求消息,其信息与图1所示示例中的步骤1相同。
示例性的,步骤2可以包括第一基站将在步骤1中收到的请求消息转发给(例如多个)其他基站。其他基站会发送更新后的AI/ML模型区域信息,更具体地说,更新后的AI/ML模型区域信息可以包括AI/ML模型区域标识ID,位置信息,场景信息,无线通信节点能力信息,和/或AI/ML模型信息。
示例性的,步骤3可以包括终端接收关于AI/ML模型区域信息更新请求的响应消息,其 中的信息与步骤2中其他基站发送给第一基站的信息相同。
通过上述示例过程,终端能够向基站,如第一基站以及进而其他基站请求AI/ML模型区域信息更新。通过接收更新后的AI/ML模型区域信息,终端可以确保具备最相关的AI模型,提高性能并适应不断变化的网络条件。因此,本公开的实施例提供了一种有效的终端触发的AI/ML模型区域信息更新过程,能够提升网络性能和用户体验。
需要注意,上述实施例仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对实施例进行各种修改和变化。
图3图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图3所示,图示了基站(如gNB)触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
在某些情况下,如果第一基站(s-gNB)没有与其他候选gNB对应的AI/ML模型区域信息(或AI/ML模型区域信息列表),则第一基站会触发AI/ML模型区域信息更新过程。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括第一基站触发AI/ML模型区域信息更新过程。具体来说,第一基站可以向其他基站发送AI/ML模型区域信息更新的请求消息,该信息可以包括模型信息和事由。
示例性的,步骤2可以包括其他基站向第一基站B发送关于AI/ML模型区域信息更新请求的响应消息。作为示例而非限制,响应消息可以指示AI/ML模型区域信息更新请求被准予,并且包括更新后的AI/ML模型区域信息。
通过上述示例过程,基站(如gNB)能够触发AI/ML模型区域信息的更新,以确保所配置的AI/ML模型区域信息(列表)的有效性,例如,包括与其他候选基站相对应的AI/ML模型区域信息(列表)。因此,本公开的实施例提供了一种由基站(如gNB)触发的AI/ML模型区域信息更新过程,能够提升网络性能和效率。
需要注意,上述实施例仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对实施例进行各种修改和变化。
图4图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图4所示,图示了终端触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步 骤/信令:
示例性的,步骤1可以包括终端向基站(如gNB)发送AI/ML模型区域信息更新的请求消息,该消息至少包括以下内容:
位置信息:用于指示基站(如gNB)的位置信息,包括候选基站。
事由:用于指示AI/ML模型区域信息更新的原因。在此情况下,事由可以是候选基站的小区不在所配置的AI/ML模型区域信息中,或者进行了模型(重新)选择,新的驻留的小区不属于所配置的AI/ML模型区域信息。
终端标识:用于指示终端的身份。
该消息可以是专用的用于请求AI/ML模型区域信息更新的消息。一旦基站(如gNB)接收到该消息,就知道它是用于要求AI/ML模型区域信息更新的。另外,该消息也可以是传统消息,其可包含一个指示符,用于指示该消息用于AI-TA更新请求。或者,该消息也可以是传统消息,其可包含一个用于指示为什么要发起专用消息的原因。
示例性的,步骤2可以包括终端接收更新后的AI/ML模型区域信息,更具体地说,更新后的AI/ML模型区域信息可以包括以下任意之一:AI/ML模型区域指示符、AI/ML模型区域指示符列表、AI模型信息、位置信息或AI/ML模型区域信息和模型信息的映射规则/关系。
通过上述示例过程,终端能够向基站(如gNB)发送AI/ML模型区域信息更新请求,并接收更新后的AI/ML模型区域信息,以提高网络性能和效率。
需要注意,上述实施例仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对实施例进行各种修改和变化。
图5图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图5所示,图示了终端触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括终端向基站(如gNB)发送AI/ML模型区域信息更新的请求消息,该消息至少包括以下内容之一:位置信息、事由、终端标识等。更具体地说,对于上述事件2而言,候选基站的信息需要包含在消息中,以指示所请求的AI/ML模型区域信息(列表)。
示例性的,步骤2可以包括第一基站将步骤1中的请求消息转发给(例如多个)其他基站,其他基站发送更新后的AI/ML模型区域信息,更具体地说,更新后的AI/ML模型区域信息至 少包含以下之一:AI/ML模型区域指示符、AI/ML模型区域指示符列表、AI模型信息、位置信息或AI/ML模型区域信息和模型信息的映射规则/关系。
示例性的,步骤3可以包括终端从第一基站接收更新后的AI/ML模型区域信息,更具体地说,更新后的AI/ML模型区域信息至少包含以下之一:AI/ML模型区域指示符、AI/ML模型区域指示符列表,AI模型信息、位置信息或AI/ML模型区域信息和模型信息的映射规则/关系。
通过上述示例过程,终端能够向基站,如第一基站以及进而其他基站发送AI/ML模型区域信息更新请求,并从第一基站接收更新后的AI/ML模型区域信息,以提高网络性能和效率。
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
图6图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图6所示,图示了基站(如gNB)触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括第一基站向其他基站发送AI/ML模型区域信息更新的请求消息,该消息至少包括以下内容之一:位置信息、事由、终端标识等。更具体地说,对于事件2而言,候选基站的小区信息需要包含在消息中,以指示所请求的AI/ML模型区域信息(列表)。
示例性的,步骤2可以包括其他基站对步骤1中的消息进行响应,其他基站发送更新后的AI/ML模型区域信息,更具体地说,更新后的AI/ML模型区域信息可以是AI/ML模型区域指示符、AI/ML模型区域指示符列表、AI模型信息、位置信息或AI/ML模型区域信息和模型信息的映射规则/关系。
通过上述示例过程,第一基站能够向其他基站发送AI/ML模型区域信息更新请求,并接收来自其他基站的更新后的AI/ML模型区域信息,以提高网络性能和效率。
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
图7图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图7所示,图示了终端触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
在某些情况下,终端触发了AI/ML模型区域信息更新过程,它将向基站(如gNB)发送 请求消息,基站(如gNB)将消息转发给CN节点,CN将通过更新AI/ML模型区域信息来响应终端。以下是一个关于小区重选的示例,AI/ML模型区域信息更新可能会在注册过程中发生,AI/ML模型区域信息由AMF发送给终端。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括终端通过注册请求消息向基站(如gNB)发送AI/ML模型区域信息更新的请求消息,该消息至少包括以下内容之一:位置信息、事由、终端标识、AI模型信息等。更具体地说,对于事件1而言,AI模型信息需要包含在消息中;对于事件2而言,候选基站的小区信息需要包含在消息中,以指示所请求的AI/ML模型区域信息(列表)。
示例性的,步骤2可以包括基站(如gNB)通过例如N2消息将AI/ML模型区域信息更新请求消息转发给AMF。
示例性的,步骤3可以包括如果终端的注册请求被接受,终端通过来自AMF的注册接受消息接收更新后的AI/ML模型区域信息。
通过上述示例过程,终端能够触发AI/ML模型区域信息更新过程,并从AMF接收更新后的AI/ML模型区域信息,以提高网络性能和效率。
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
图8图示了根据本公开的一些实施例的AI/ML模型区域信息更新过程的信令图。
如图8所示,图示了基站(如gNB)触发的AI/ML模型区域信息更新过程。该过程可以由事件触发或定期触发。
在某些情况下,基站(如gNB)触发了AI/ML模型区域信息更新过程,它将向CN节点(例如AMF、网络数据分析功能(NWDAF)等)发送请求消息。
根据一些实施例,提供了一种用于更新AI/ML模型区域信息的方法。以下是该方法的步骤/信令:
示例性的,步骤1可以包括基站(如gNB)通过例如N2消息向AMF发送AI/ML模型区域信息更新的请求消息。该请求消息至少包括以下内容之一:位置信息、事由、终端标识、AI模型信息等。更具体地说,对于事件1而言,AI模型信息需要包含在消息中;对于事件2而言,候选基站的小区信息需要包含在消息中,以指示所请求的AI/ML模型区域信息(列表)。
示例性的,步骤2可以包括基站通过例如N2消息从AMF接收更新后的AI/ML模型区域信息。
通过上述示例过程,基站(如gNB)能够通过N2消息向AMF发送AI/ML模型区域信息更新的请求消息,并且能够通过N2消息从AMF接收更新后的AI/ML模型区域信息,以实现更高效的网络通信和优化。
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
终端在切换过程中确定快速模型控制的过程
图9图示了根据本公开的一些实施例的在切换过程期间的快速模型控制过程的信令图。
如图9所示,展示了潜在切换过程中的快速模型控制。对于本例,模型控制(例如模型选择、模型切换、模型更新等)位于终端端。
示例性的,步骤0:终端接收有关AI/ML模型区域信息的配置消息,该消息用于在切换过程中帮助终端快速确定模型控制。该消息可包括AI/ML模型区域指示符、AI/ML模型区域指示符列表、AI模型信息、位置信息或AI/ML模型区域信息与模型信息的映射规则/关系,并且该消息可以在例如RRC消息、系统信息、MAC CE、DCI,NAS等中传递。
示例性的,步骤1:终端接收传统测量配置,测量配置通过专用信令(例如RRCReconfiguration或RRCResume等)提供。
示例性的,步骤2:终端检查配置的AI/ML模型区域信息,如果候选第二基站(如Target gNB,T-gNB)在配置的AI/ML模型区域信息列表范围内,则无需执行模型控制,然后终端(如终端)执行步骤4;如果候选第二基站(如Target gNB,T-gNB)不在配置的AI/ML模型区域信息列表范围内,则终端将执行步骤3模型控制,然后终端执行步骤3'和步骤4,步骤3'的详细信息在上文描述。
示例性的,步骤4:基于步骤2的检查结果,终端向第一基站发送测量上报消息,该消息可携带模型控制结果,消息中至少包括以下之一:模型信息、指示或原因。
指示:用于指示当前进行中的模型是否已更改。
模型信息:用于指示当前的AI/ML模型。
事由:用于指示AI/ML模型更改或未更改的原因。
示例性的,步骤5:第一基站发起切换(Handover,HO)决策。
示例性的,步骤6:第一基站向第二基站发送切换请求消息,该消息可携带模型信息,模型信息表示适用于第二基站的AI/ML模型,此步骤可以在终端和第二基站之间提前进行模型识别。
示例性的,步骤7:第二基站执行接纳控制。
示例性的,步骤8:第二基站向原基站发送关于切换请求的确认消息,其中携带第二基站的信息和终端的资源信息。
示例性的,步骤9:第一基站向终端发送带有切换命令的消息(如RRCReconfiguration消息等),其中携带用于访问第二基站的信息,例如小区ID。
需要注意,步骤3'可以在步骤4之前或之后发生。在某些示例中,如果步骤3'在步骤4之后发送,则步骤4和步骤6不携带AI模型信息。
还需要注意,如果步骤3发生并选择/切换/更新新的AI模型,并不意味着立即激活新模型,而是可以通过步骤9的RRC重配置指示进行激活。具体地,步骤9携带激活/去激活配置信息,包括至少以下之一指示模型的激活/去激活起始时间、偏置、持续时间和结束时间等。可选地,步骤9仅携带模型激活/去激活指示信息。当终端接收到该指示信息后,自行进行模型激活/去激活。
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
所描述的步骤(信令/框)的顺序不旨在被理解为限制,并且可以跳过或以任何顺序组合任意数量的所描述的步骤(信令/框)以实现方法或可替换的方法。
基站(如gNB)在切换过程中确定快速模型控制的过程
图10图示了根据本公开的一些实施例的在切换过程期间的快速模型控制过程的信令图。
如图10所示,展示了潜在切换过程中的快速模型控制。对于本例,模型控制(例如模型选择、模型切换、模型更新等)位于终端侧。
示例性的,步骤1:终端接收传统测量配置,该配置通过专用信令(例如RRCReconfiguration或RRCResume等)提供。
示例性的,步骤2:终端根据步骤1中的测量配置向第一基站(如Source gNB,S-gNB)发送测量上报消息。
示例性的,步骤3:基于步骤2中的上报消息,第一基站发起切换(如handover,HO)决策。
示例性的,步骤4:第一基站检查配置的AI/ML模型区域信息列表,如果候选第二基站(如Target gNB,T-gNB)在配置的AI/ML模型区域信息列表范围内,则无需执行模型控制。此时,第一基站将AI/ML模型信息发送给第二基站,并可在步骤8的切换请求消息中携带。如果候选第二基站不在配置的AI/ML模型区域信息列表范围内,则第一基站执行步骤5,步骤6,步骤7。
示例性的,步骤5:终端接收用于模型控制的消息,该消息用于指示终端当前的AI/ML模型需要进行模型控制。该消息至少包括以下内容之一:
位置信息:用于指示当前第二基站的位置信息;
AI/ML模型区域信息:用于指示第二基站对应的AI/ML模型区域信息,例如AI/ML模型区域指示符、AI/ML模型区域指示符列表、AI模型信息、位置信息或AI/ML模型区域信息与模型信息的映射规则/关系;还可以用于帮助终端选择/切换AI/ML模型;
指示符:用于指示需要执行模型控制,和/或当前AI/ML模型不再适用;
事由:用于指示AI/ML模型更改或未更改的原因。
示例性的,步骤6:终端根据步骤5执行模型控制。
示例性的,步骤7:终端向第一基站发送带有模型控制结果的消息,,例如AI/ML模型信息;或指示符,用于指示AI/ML模型已经更改;
示例性的,步骤8:第一基站向第二基站发送切换请求消息,该消息可能携带AI/ML模型信息,AI/ML模型信息表示适用于第二基站的AI模型。此步骤可在终端和第二基站之间提前进行模型识别。
示例性的,步骤9:第二基站执行接纳模型控制。
示例性的,步骤10:第二基站向第一基站发送关于切换请求的确认消息,其中携带第二基站的信息和终端的资源信息。
示例性的,步骤11:第一基站向终端发送带有切换命令的消息(如RRCReconfiguration消息等),其中携带用于访问第二基站的信息,例如小区ID。
需要注意,在某些示例中,步骤4-7可以与步骤3、步骤8、步骤9-11同时执行。
还需要注意,在某些示例中,步骤4之后,第一基站检查配置的AI/ML模型区域信息列表。如果候选第二基站在配置的AI/ML模型区域信息列表范围内,表明终端侧AI/ML适用于第二基站,则无需执行模型控制。此时,第一基站也可以向终端发送指示消息,该消息用于指示终端,终端侧AI/ML适用于第二基站,可以不必执行模型控制,或第一基站也可以向终端发送传统的消息,该消息包含指示符,该指示符用于指示终端,终端侧AI/ML适用于第二基站,可以不必执行模型控制
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
所描述的步骤(信令/框)的顺序不旨在被理解为限制,并且可以跳过或以任何顺序组合任意数量的所描述的步骤(信令/框)以实现方法或可替换的方法。
图11图示了根据本公开的一些实施例的在切换过程期间的快速模型控制过程的信令图。
如图11所示,展示了潜在切换过程中的快速模型控制。对于本例,模型控制(例如模型选择、模型切换、模型更新等)位于基站(如gNB)端。
示例性的,步骤1:终端接收传统测量配置,该传统测量配置通过专用信令(例如RRCReconfiguration、RRCResume等)提供。
示例性的,步骤2:终端根据步骤1中的测量配置向第一基站(如Source gNB,S-gNB)发送测量报告消息。
示例性的,步骤3:基于步骤2中的报告消息,第一基站发起切换HO决策。
示例性的,步骤4:第一基站检查配置的AI/ML模型区域信息列表,如果候选第二基站(如Target gNB,T-gNB)在配置的AI/ML模型区域信息列表范围内,则无需执行模型控制。第一基站执行步骤6,此时,第一基站可以将AI/ML模型信息发送给第二基站,并可以在步骤6的切换请求消息中携带。如果候选第二基站不在配置的AI/ML模型区域信息列表范围内,则第一基站执行步骤4',步骤5。
示例性的,步骤4':或者,第一基站发起AI/ML模型区域信息更新过程,具体细节在实施例2中显示。
示例性的,步骤5:终端根据步骤4'执行模型控制,例如选择/切换一个或多个适合的AIML模型。
示例性的,步骤6:第一基站向第二基站发送切换请求消息,该消息可能携带模型信息,模型信息表示适用于第二基站的AI/ML模型。此步骤可以在终端和第二基站之间提前进行模型识别。
示例性的,步骤7:第二基站执行接纳模型控制。
示例性的,步骤8:第二基站向第一基站发送关于切换请求的确认消息,其中携带第二基站的信息和终端的资源信息。
示例性的,步骤9:第一基站向终端发送带有切换命令的消息(如RRCReconfiguration消息等),其中携带用于访问第二基站的信息,例如小区ID。
需要注意,如果在步骤5中选择/切换/更新了新的AI模型,并不意味着立即激活新模型。可以通过步骤7的切换请求消息进行激活指示。具体地,步骤7携带激活/去激活配置信息,包括指示模型的激活/去激活起始时间、偏置、持续时间和结束时间等。可选地,仅携带模型激活/去激活指示信息。当收到该指示信息后,自行进行模型激活/去激活。
需要注意,模型信息可以通过步骤9的消息(如RRCReconfiguration消息等)传输,也可 以在步骤4之后通过专用的RRC消息、MAC CE或DCI发送。
需要注意,上述步骤仅为示例,并不限制本发明的范围。在不脱离本发明精神和范围的情况下,可以对步骤进行各种修改和变化。
所描述的步骤(信令/框)的顺序不旨在被理解为限制,并且可以跳过或以任何顺序组合任意数量的所描述的步骤(信令/框)以实现方法或可替换的方法。
本公开在上述实施方式中描述网络架构中终端与网元组件之间的通信的示例,其主要出于示例目的而不是限制性的。
所描述的步骤(信令/框)的顺序不旨在被解释为限制,并且能够跳过或以任何顺序组合任何数量的所描述的步骤(信令/框)以实现方法或替代方法。通常,能够使用软件、固件、硬件(例如,固定逻辑电路)、手动处理或其任意组合来实现本文中描述的组件、模块、方法和操作的任何一个。可以在存储在计算机处理系统本地和/或远程的计算机可读存储器上的可执行指令的一般场境中描述示例方法的一些操作,并且实施方式能够包括软件应用、程序、函数等。替代地或另外,本文中描述的任何功能能够至少部分地由一个或多个硬件逻辑组件执行,诸如但不限于现场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准产品(ASSP)、片上系统(SoC)、复杂可编程逻辑器件(CPLD)等。
此外,本公开实施例中描述的信令传递可以以本领域已知的任何方式来实现。例如,信令传递可以为显式和/或隐式的。此外,所图示的步骤(信令/框)仅出于示例目的,而不旨在限制本申请。
可以从以下枚举的示例实施例中理解本发明的各个方面:
根据示例实施例,提供了一种无线通信系统中的人工智能/机器学习AI/ML模型的控制方法,包括:
所述无线通信节点获取AI/ML模型区域信息,无线通信节点基于所述AI/ML模型区域信息确定AI/ML模型的模型控制,所述无线通信节点为基站或终端。
根据示例实施例,所述AI/ML模型的模型控制包括以下各项中的至少一者:
AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调。
根据示例实施例,所述AI/ML模型区域信息包括以下各项中的至少一者:
AI/ML模型区域标识ID,
位置信息,
场景信息,
和/或AI/ML模型信息。
根据示例实施例,所述位置信息包括以下各项中的至少一者:
小区信息、AI/ML模型区域标识ID、终端位置信息、基站位置信息,场景信息、控制单元CU ID、数据单元DU ID。
根据示例实施例,所述场景信息包括以下各项中的至少一者:
传输环境信息,速度信息,角度扩展信息,时延扩展信息,信噪比,信道接收功率,和/或干扰信息。
根据本公开的一个或多个实施例,所述无线通信节点能力信息包括以下各项中的至少一者:
计算能力,存储能力,AI/ML模型支持能力,负载情况,和/或传输能力,AI/ML模型支持AI/ML使用场景和/或AI/ML模型支持AI/ML特征。详细而言,在一些示例中例如一些特定计算能力,一些特定存储能力,AI/ML模型支持一些特定能力,一些特定负载情况,一些特定传输能力,AI/ML模型支持一些特定使用场景和/或支持支持一些特定特征。
无线网络能力节点能力信息;指示一个无线网络节点的能力,如节点支持的AI/ML模型的模型操作能力、节点的硬件能力信息(如存储能力,算力,电池能力等)、节点支持的AI/ML模型的功能性/特征,节点支持的AI/ML模型的数据,节点支持的AI/ML模型的场景等。
根据本公开的一个或多个实施例,所述AI/ML模型信息包括以下各项中的至少一者:
AI/ML模型ID信息,和/或AI/ML模型元信息。
根据示例实施例,所述AI/ML模型信息包括以下各项中的至少一者:
AI/ML模型ID信息,和/或AI/ML模型元信息。
根据示例实施例,所述AI/ML模型区域信息通过以下各项中的至少一者传输:
非接入层NAS消息、无线电资源控制RRC消息、介质接入控制控制元素MAC-CE消息、下行链路控制信息DCI消息、上行链路信道、下行链路信道、系统消息、N2消息和/或Xn消息。
根据示例实施例,由所述无线通信节点基于所述AI/ML模型区域信息确定AI/ML模型的模型控制包括以下各项中的至少一者:
如果所述无线通信节点的位置信息属于所述AI/ML模型区域信息中所指示的位置信息,则所述无线通信节点不执行AI/ML模型的模型控制;
如果所述无线通信节点的位置信息不属于所述AI/ML模型区域信息中所指示的位置信息,则所述无线通信节点执行AI/ML模型的模型控制;
如果所述无线通信节点的AI/ML模型信息属于所述AI/ML模型区域信息中所指示的 AI/ML模型信息,则所述无线通信节点不执行AI/ML模型的模型控制;
如果所述无线通信节点的AI/ML模型信息不属于所述AI/ML模型区域信息中所指示的AI/ML模型信息,则所述无线通信节点执行AI/ML模型的模型控制;
根据示例实施例,所述执行AI/ML模型的模型控制包括以下各项中的至少一者:
AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调、AI/ML模型的模型控制的结果上报,和/或AI/ML模型的模型控制的响应。
根据示例实施例,当所述无线通信节点为终端时,所述终端获取所述AI/ML模型区域配置信息,所述终端根据所述AI/ML模型区域信息执行所述AI/ML模型的模型控制。
根据示例实施例,所述终端获取所述AI/ML模型区域配置信息包括以下各项中的至少一者:
AI/ML模型信息、事由、指示符、终端ID、和/或位置信息。
根据示例实施例,当所述无线通信节点为基站时,所述基站获取所述AI/ML模型区域配置信息,所述基站根据所述AI/ML模型区域信息执行所述AI/ML模型的模型控制。
根据示例实施例,所述基站获取所述AI/ML模型区域配置信息包括以下各项中的至少一者:
AI/ML模型信息、事由、指示符、基站ID,和/或位置信息。
根据示例实施例,第一无线通信节点向第二无线通信节点发送的所述AI/ML模型区域信息请求消息和/或响应消息包括以下各项中的至少一者:
AI/ML模型区域标识ID,
位置信息,
场景信息,
无线通信节点能力信息,和/或
AI/ML模型信息。
根据示例实施例,所述无线通信节点发送所述AI/ML模型区域信息的请求消息,包括周期性发送或事件触发。
根据示例实施例,在所述周期性发送的情况下,所述AI/ML模型区域信息响应于周期性配置而进行更新,其中,所述周期性配置包括所述更新的持续时间、起始时间、和/或结束时间。
根据示例实施例,在所述方式为事件触发的情况下,所述AI/ML模型区域信息响应于事件触发配置而进行更新。
根据示例实施例,所述周期性发送和/或所述事件触发相应的周期性配置和/或事件触发配置是预配置的、和/或配置的、和/或固定的。
根据示例实施例,所述AI/ML模型区域信息请求消息和/或响应消息通过以下至少一项承载:
非接入层NAS消息、无线电资源控制RRC消息、介质接入控制控制元素MAC-CE消息、下行链路控制信息DCI消息、上行链路信道、下行链路信道、系统消息、N2消息和/或Xn消息。
根据示例实施例,提供了一种用于终端的无线通信方法,包括:
所述终端从第一基站获取AI/ML模型区域信息,所述AI/ML模型区域信息用于所述终端确定AI/ML模型的模型控制,
其中,所述终端基于所述AI/ML模型区域信息确定是否执行所述AI/ML模型的模型控制。
根据示例实施例,还包括:
所述终端从所述第一基站获取AI/ML模型区域配置消息。
根据示例实施例,如果第二基站不在所述AI/ML模型区域的情况下,所述终端确定执行所述AI/ML模型的模型控制。
根据示例实施例,所述执行AI/ML模型的模型控制包括在所述终端与所述第一基站之间更新所述AI/ML模型区域信息。
根据示例实施例,还包括:
所述终端将针对所述AI/ML模型区域配置消息的上报消息以及所述执行所述AI/ML模型的模型控制的结果发送至所述第一基站,所述第一基站将获取的执行所述AI/ML模型的模型控制的结果发送至所述第二基站。
根据示例实施例,提供了一种用于第一基站的无线通信方法,包括:
所述第一基站从所述终端获取AI/ML模型的模型控制上报消息,所述AI/ML模型的模型控制上报消息包括所述AI/ML模型的模型控制的结果;
所述第一基站将所述第二基站的小区切换请求以及所述AI/ML模型的模型控制的结果发送至所述第二基站。
根据示例实施例,提供了一种用于第一基站的无线通信方法,包括:
所述第一基站根据终端反馈的AI/ML模型的模型控制消息,确定第二基站是否在AI/ML模型区域信息中;
如果所述第二基站不在所述AI/ML模型区域信息中,所述第一基站执行AI/ML模型的模 型控制。
根据示例实施例,所述AI/ML模型的模型控制包括在所述第一基站与所述第二基站之间更新所述AI/ML模型区域信息。
根据示例实施例,还包括:所述第一基站将所述第二基站的小区切换请求以及所述模型控制的结果发送至所述第二基站。
根据示例实施例,还包括:如果所述第二基站在所述AI/ML模型区域信息中,所述第一基站将所述AI/ML模型区域信息中所包含的AI/ML模型信息以及至所述第二基站的小区切换请求发送至所述第二基站。
根据示例实施例,提供了一种芯片,包括:处理器,其被配置为调用和运行存储在存储器中的计算机程序,以使得其中安装有所述芯片的设备执行示例实施例中任一个的方法。
图12是根据本公开的实施例的用于无线通信的示例系统700的框图。可以使用任何适当配置的硬件和/或软件将本文描述的实施例实现到系统中。图12图示系统700,包括射频(RF)电路710、基带电路720、处理单元730、存储器/储存器740、显示器750、相机760、传感器770和输入/输出(I/O)接口780,如图所示彼此耦合。
处理单元730可以包括电路,例如但不限于一个或多个单核或多核处理器。处理器可以包括通用处理器和专用处理器的任何组合,例如图形处理器和应用处理器。处理器可以与存储器/储存器耦合,并且被配置为执行存储在存储器/储存器中的指令,以能够实现在系统上运行的各种应用和/或操作系统。RF电路710、基带电路720、处理单元730、存储器/储存器740、显示器750、相机760、传感器770和I/O接口780是系统700中的公知元件,例如但不限于膝上型计算设备、平板计算设备、上网本、超极致笔电、智能电话等。此外,作为软件产品的指令可以存储在计算机中的可读存储介质中。计算机中的软件产品存储在存储介质中,包括用于计算设备(诸如个人计算机、服务器或网络设备)的多个命令,以运行本公开的实施例所公开的所有或一些步骤。存储介质包括USB盘、移动硬盘、只读存储器(ROM)、随机存取存储器(RAM)、软盘或能够存储程序代码的其它类型的介质。
本公开的实施例是可以在3GPP规范中采用的技术/过程的组合以创建最终产品。
虽然已经结合被认为是最实用和优选的实施例描述了本公开,但是应当理解,本公开不限于所公开的实施例,而是旨在覆盖在不脱离所附权利要求的最广泛解释的范围的情况下所做出的各种布置。

Claims (32)

  1. 一种无线通信系统中的人工智能/机器学习AI/ML模型的控制方法,包括:
    所述无线通信节点获取AI/ML模型区域信息,无线通信节点基于所述AI/ML模型区域信息确定AI/ML模型的模型控制,所述无线通信节点为基站或终端。
  2. 根据权利要求1所述的方法,其中,所述AI/ML模型的模型控制包括以下各项中的至少一者:
    AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调。
  3. 根据权利要求1所述的方法,其中,所述AI/ML模型区域信息包括以下各项中的至少一者:
    AI/ML模型区域标识ID,
    位置信息,
    场景信息,
    无线通信节点能力信息,
    和/或AI/ML模型信息。
  4. 根据权利要求3所述的方法,其中,所述位置信息包括以下各项中的至少一者:
    小区信息、AI/ML模型区域标识ID、终端位置信息、基站位置信息,场景信息、控制单元CU ID、数据单元DU ID。
  5. 根据权利要求3所述的方法,其中,所述场景信息包括以下各项中的至少一者:
    传输环境信息,速度信息,角度扩展信息,时延扩展信息,信噪比,信道接收功率,和/或干扰信息。
  6. 根据权利要求3所述的方法,其中,所述无线通信节点能力信息包括以下各项中的至少一者:
    计算能力,存储能力,AI/ML模型支持能力,负载情况,传输能力,AI/ML模型支持AI/ML使用场景和/或AI/ML模型支持AI/ML特征。
  7. 根据权利要求3所述的方法,其中,所述AI/ML模型信息包括以下各项中的至少一者:
    AI/ML模型ID信息,和/或AI/ML模型元信息。
  8. 根据权利要求1所述的方法,其中,所述AI/ML模型区域信息通过以下各项中的至少一者传输:
    非接入层NAS消息、无线电资源控制RRC消息、介质接入控制控制元素MAC-CE消息、下行链路控制信息DCI消息、上行链路信道、下行链路信道、系统消息、N2消息和/或Xn消息。
  9. 根据权利要求3所述的方法,其中,由所述无线通信节点基于所述AI/ML模型区域信息确定AI/ML模型的模型控制包括以下各项中的至少一者:
    如果所述无线通信节点的位置信息属于所述AI/ML模型区域信息中所指示的位置信息,则所述无线通信节点不执行AI/ML模型的模型控制;
    如果所述无线通信节点的位置信息不属于所述AI/ML模型区域信息中所指示的位置信息,则所述无线通信节点执行AI/ML模型的模型控制;
    如果所述无线通信节点的AI/ML模型信息属于所述AI/ML模型区域信息中所指示的AI/ML模型信息,则所述无线通信节点不执行AI/ML模型的模型控制;
    如果所述无线通信节点的AI/ML模型信息不属于所述AI/ML模型区域信息中所指示的AI/ML模型信息,则所述无线通信节点执行AI/ML模型的模型控制。
  10. 根据权利要求9所述的方法,其中,所述执行AI/ML模型的模型控制包括以下各项中的至少一者:
    AI/ML模型激活、AI/ML模型去激活、AI/ML模型选择、AI/ML模型切换、AI/ML模型回退、AI/ML模型更新、AI/ML模型训练、AI/ML模型重训练、或AI/ML模型微调、AI/ML模型的模型控制的结果上报,和/或AI/ML模型的模型控制的响应。
  11. 根据权利要求10所述的方法,其中,当所述无线通信节点为终端时,所述终端获取所述AI/ML模型区域配置信息,所述终端根据所述AI/ML模型区域信息执行所述AI/ML模型的模型控制。
  12. 根据权利要求11所述的方法,其中,所述终端获取所述AI/ML模型区域配置信息包括以下各项中的至少一者:
    AI/ML模型信息、事由、指示符、终端ID、和/或位置信息。
  13. 根据权利要求10所述的方法,其中,当所述无线通信节点为基站时,所述基站获取所述AI/ML模型区域配置信息,所述基站根据所述AI/ML模型区域信息执行所述AI/ML模型的模型控制。
  14. 根据权利要求13所述的方法,其中,所述基站获取所述AI/ML模型区域配置信息包括以下各项中的至少一者:
    AI/ML模型信息、事由、指示符、基站ID,和/或位置信息。
  15. 根据权利要求1所述的方法,其中,第一无线通信节点向第二无线通信节点发送的所述AI/ML模型区域信息请求消息和/或响应消息包括以下各项中的至少一者:
    AI/ML模型区域标识ID,
    位置信息,
    场景信息,
    无线通信节点能力信息,和/或
    AI/ML模型信息。
  16. 根据权利要求15所述的方法,其中,所述无线通信节点发送所述AI/ML模型区域信息的请求消息,包括周期性发送或事件触发。
  17. 根据权利要求16所述的方法,其中,在所述周期性发送的情况下,所述AI/ML模型区域信息响应于周期性配置而进行更新,其中,所述周期性配置包括所述更新的持续时间、起始时间、和/或结束时间。
  18. 根据权利要求16所述的方法,其中,在所述方式为事件触发的情况下,所述AI/ML模型区域信息响应于事件触发配置而进行更新。
  19. 根据权利要求16所述的方法,其中,所述周期性发送和/或所述事件触发相应的周期性配置和/或事件触发配置是预配置的、和/或配置的、和/或固定的。
  20. 根据权利要求15所述的方法,所述AI/ML模型区域信息请求消息和/或响应消息通过以下至少一项承载:
    非接入层NAS消息、无线电资源控制RRC消息、介质接入控制控制元素MAC-CE消息、下行链路控制信息DCI消息、上行链路信道、下行链路信道、系统消息、N2消息和/或Xn消息。
  21. 一种用于终端的无线通信方法,包括:
    所述终端从第一基站获取AI/ML模型区域信息,所述AI/ML模型区域信息用于所述终端确定AI/ML模型的模型控制,
    其中,所述终端基于所述AI/ML模型区域信息确定是否执行所述AI/ML模型的模型控制。
  22. 根据权利要求21所述的方法,还包括:
    所述终端从所述第一基站获取AI/ML模型区域配置消息。
  23. 根据权利要求22所述的方法,其中,如果第二基站不在所述AI/ML模型区域的情况下,所述终端确定执行所述AI/ML模型的模型控制。
  24. 根据权利要求23所述的方法,其中,所述执行AI/ML模型的模型控制包括在所述终 端与所述第一基站之间更新所述AI/ML模型区域信息。
  25. 根据权利要求21所述的方法,还包括:
    所述终端将针对所述AI/ML模型区域配置消息的上报消息以及所述执行所述AI/ML模型的模型控制的结果发送至所述第一基站,所述第一基站将获取的执行所述AI/ML模型的模型控制的结果发送至所述第二基站。
  26. 一种终端,所述终端被配置成执行权利要求21至25中任一项所述的方法。
  27. 一种用于第一基站的无线通信方法,包括:
    所述第一基站从所述终端获取AI/ML模型的模型控制上报消息,所述AI/ML模型的模型控制上报消息包括所述AI/ML模型的模型控制的结果;
    所述第一基站将所述第二基站的小区切换请求以及所述AI/ML模型的模型控制的结果发送至所述第二基站。
  28. 一种用于第一基站的无线通信方法,包括:
    所述第一基站根据终端反馈的AI/ML模型的模型控制消息,确定第二基站是否在AI/ML模型区域信息中;
    如果所述第二基站不在所述AI/ML模型区域信息中,所述第一基站执行AI/ML模型的模型控制。
  29. 根据权利要求28所述的方法,其中,所述AI/ML模型的模型控制包括在所述第一基站与所述第二基站之间更新所述AI/ML模型区域信息。
  30. 根据权利要求29所述的方法,还包括:
    所述第一基站将所述第二基站的小区切换请求以及所述模型控制的结果发送至所述第二基站。
  31. 根据权利要求28所述的方法,还包括:
    如果所述第二基站在所述AI/ML模型区域信息中,所述第一基站将所述AI/ML模型区域信息中所包含的AI/ML模型信息以及至所述第二基站的小区切换请求发送至所述第二基站。
  32. 一种芯片,包括:
    处理器,其被配置为调用和运行存储在存储器中的计算机程序,以使得其中安装有所述芯片的设备执行权利要求1至31中任一项所述的方法。
PCT/CN2023/123047 2023-09-29 2023-09-29 无线通信系统中的ai/ml模型的控制方法 Pending WO2025065720A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/123047 WO2025065720A1 (zh) 2023-09-29 2023-09-29 无线通信系统中的ai/ml模型的控制方法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/123047 WO2025065720A1 (zh) 2023-09-29 2023-09-29 无线通信系统中的ai/ml模型的控制方法

Publications (1)

Publication Number Publication Date
WO2025065720A1 true WO2025065720A1 (zh) 2025-04-03

Family

ID=95204630

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/123047 Pending WO2025065720A1 (zh) 2023-09-29 2023-09-29 无线通信系统中的ai/ml模型的控制方法

Country Status (1)

Country Link
WO (1) WO2025065720A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109479309A (zh) * 2016-07-13 2019-03-15 三星电子株式会社 在移动通信中使用的接入控制方法和装置
US20230093963A1 (en) * 2021-09-28 2023-03-30 Qualcomm Incorporated Artificial intelligence based enhancements for idle and inactive state operations
WO2023133897A1 (en) * 2022-01-17 2023-07-20 Shenzhen Tcl New Technology Co., Ltd. User equipment, base station, and wireless communication methods based on machine learning/artificial intelligence

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109479309A (zh) * 2016-07-13 2019-03-15 三星电子株式会社 在移动通信中使用的接入控制方法和装置
US20230093963A1 (en) * 2021-09-28 2023-03-30 Qualcomm Incorporated Artificial intelligence based enhancements for idle and inactive state operations
WO2023133897A1 (en) * 2022-01-17 2023-07-20 Shenzhen Tcl New Technology Co., Ltd. User equipment, base station, and wireless communication methods based on machine learning/artificial intelligence

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
YUANYUAN ZHANG, MEDIATEK INC.: "Discussion on RAN2 aspects for LCM", 3GPP DRAFT; R2-2211610; TYPE DISCUSSION, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. RAN WG2, no. Toulouse, FR; 20221114 - 20221118, 4 November 2022 (2022-11-04), Mobile Competence Centre ; 650, route des Lucioles ; F-06921 Sophia-Antipolis Cedex ; France, XP052215715 *

Similar Documents

Publication Publication Date Title
CN114303347B (zh) 通信网络中与机器学习相关的方法、装置和机器可读介质
US20220279403A1 (en) Information transmission method and apparatus, storage medium and electronic apparatus
CN109951877A (zh) 一种切片信息更新方法及装置
WO2020103517A1 (zh) 终端的能力信息的获取方法、装置及系统
WO2017166988A1 (zh) 网络切片处理方法及装置、终端、基站、计算机存储介质
US12262261B2 (en) Clock synchronization method and device, and storage medium
US12219410B2 (en) Message sending method and apparatus, message receiving method and apparatus, and device and storage medium
WO2023066289A1 (zh) 小区重选方法、装置及相关设备
US20240340077A1 (en) Satellite switching method, apparatus, storage medium, and chip system
CN109121160B (zh) 一种小区信息上报的方法、指示方法、基站及终端
EP4322594A1 (en) Cell reselection method, communications apparatus, and computer-readable storage medium
CN112954711B (zh) 信息上报及接收方法、装置、相关设备及存储介质
CN110366861B (zh) 基站间切换方法、装置、系统及存储介质
US12615557B2 (en) Communication method, device, and system
CN116602016A (zh) 一种小区选择方法、寻呼方法及装置
US11470672B2 (en) Multi-connection recovery method in non-activated state and device therefor
US12550014B2 (en) Dual-connectivity handover method, handover device and storage medium
CN104507133B (zh) 语音切换的方法和装置
WO2020192387A1 (zh) 一种请求处理方法、相关装置及系统
CN115643616A (zh) 切换决策的确定方法、装置以及存储介质
CN118786707A (zh) 测量信息的上报方法、接收方法、装置、设备及存储介质
WO2025065720A1 (zh) 无线通信系统中的ai/ml模型的控制方法
JP7797706B2 (ja) サイドリンクポジショニング方法、装置、端末、サーバと無線アクセスネットワーク機器
WO2020063789A1 (zh) 删除小区的方法及装置
CN106376055B (zh) 一种gmr系统中的小区搜索方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23953861

Country of ref document: EP

Kind code of ref document: A1