WO2019206196A1 - 一种模型更新方法、装置及系统 - Google Patents

一种模型更新方法、装置及系统 Download PDF

Info

Publication number
WO2019206196A1
WO2019206196A1 PCT/CN2019/084148 CN2019084148W WO2019206196A1 WO 2019206196 A1 WO2019206196 A1 WO 2019206196A1 CN 2019084148 W CN2019084148 W CN 2019084148W WO 2019206196 A1 WO2019206196 A1 WO 2019206196A1
Authority
WO
WIPO (PCT)
Prior art keywords
entity
model
data
update
network element
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.)
Ceased
Application number
PCT/CN2019/084148
Other languages
English (en)
French (fr)
Inventor
吴中耀
池清华
徐以旭
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.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies 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 Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to JP2020560331A priority Critical patent/JP7159347B2/ja
Priority to EP19791815.4A priority patent/EP3780495A4/en
Priority to EP25172910.9A priority patent/EP4625925A3/en
Publication of WO2019206196A1 publication Critical patent/WO2019206196A1/zh
Priority to US17/081,570 priority patent/US11451452B2/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/23—Updating
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/2866—Architectures; Arrangements
    • H04L67/30—Profiles
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/22—Indexing; Data structures therefor; Storage structures
    • G06F16/2228—Indexing structures
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0894—Policy-based network configuration management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/50—Service provisioning or reconfiguring

Definitions

  • the present application relates to the field of information technology, and in particular, to a method, device, and system for updating a model.
  • AI artificial intelligence
  • the model generated by the machine learning technology in the artificial intelligence network architecture analyzes the data generated by the mobile network, and optimizes the mobile network based on the analysis result generated by the model, so as to better support the user service and become the trend of the mobile network evolution. .
  • the performance of the model may be degraded, and the network may not operate normally.
  • the model is updated by re-training the model through the newly generated data of the network element in a proactive manner through a fixed time interval. This way of actively updating the model does not accurately sense changes in the network environment.
  • the network environment may change before the model update time arrives, causing the model to no longer apply to the current situation, resulting in erroneous results, resulting in network performance degradation or inoperability.
  • the present application provides a model updating method, apparatus, and system for solving the problem of network performance degradation caused by a decrease in model performance in the prior art.
  • the embodiment of the present application provides a method for updating a model, including: a first function entity sends a model update policy to a second function entity, where the model update policy includes an update condition of a model of the first network element, The model of the first network element is used to guide parameter adjustment of the first network element; when the first function entity receives the update request, perform a process of updating a model of the first network element; The update request is triggered by the second functional entity when it is determined that the model satisfies the update condition.
  • the first functional entity may be an analysis and modeling function AMF entity
  • the second functional entity may be an intelligent collaborative function APF entity or a function execution MEF entity.
  • the model update mechanism of the present application replaces the model update mechanism from the active cycle update to the passive trigger update, the first functional entity assigns the model update policy to the second functional entity, and the second functional entity goes to the first when the monitored model meets the update condition
  • the function entity triggers the update request, so that the first function entity receives the update request and performs a process of updating the model, wherein the update condition indicates that the network environment changes, and when the network environment changes to meet the update condition, the current environment is updated.
  • the model avoids the problem of network performance degradation or resource waste that may result from proactively updating the model. Model update strategies can be configured according to the scenario, resulting in greater flexibility.
  • the model update policy further includes an index of the first data, and the first data is used to determine whether the model satisfies the update condition.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is an intelligent collaborative function APF entity
  • the first functional entity performs a model for updating the first network element
  • the method further includes:
  • the first functional entity performs a process of updating the model, including:
  • the AMF entity sends a data request to the data service function DSF entity, where the data request is used to request the second data corresponding to the first network element required for the model update;
  • the AMF entity retrains the model based on the second data and sends the retrained model to the MEF.
  • the model when the model is updated, the model is trained based on the model output result and the parameter adjustment action corresponding to the model result, and the accuracy of the model training is improved.
  • the system to which the method is applied includes a plurality of network elements including the first network element, where the plurality of network elements all correspond to a model of the first network element, and the The plurality of network elements are managed by at least two APF entities, the first functional entity is an AMF entity, and the second functional entity is any one of the at least two APF entities, and related data of the multiple network elements And being managed by the at least two DSF entities, where the corresponding models of the plurality of network elements are configured in the at least two MEFs;
  • the first function entity After the AMF entity sends the model update policy to each APF entity, the first function entity performs a process of updating the model, including:
  • the AMF entity sends a data request to the at least two DSF entities, where the data request is used to request third data corresponding to the plurality of network elements respectively required for performing model update;
  • the AMF entity retrains the model according to the third data corresponding to the plurality of network elements respectively, and sends the retrained model to the at least two MEFs respectively.
  • the first functional entity is an AMF entity
  • the second functional entity is an MEF entity.
  • the method may further include: the AMF entity sending the first network element to the MEF entity a process of the first functional entity performing a process of updating the model of the first network element, comprising: the AMF entity sending a data request to a DSF entity, where the data request is used to request a model of the first network element And updating, by the AMF entity, the fourth data that is sent by the first network element, where the AMF entity receives the fourth data that is sent by the DSF entity, where the fourth data includes the first The output result of the model of the network element, the data of the output result and the adjustment action of the first network element parameter corresponding to the output result; the AMF entity retrains the first network element according to the fourth data The model and send the retrained model to the MEF.
  • the above design through the MEF trigger model update.
  • the AMF configures the model update policy to the MEF so that the MEF triggers the model update when the update condition is met based on the model update policy.
  • the model is retrained according to the output result of the unupdated model and the parameter adjustment action corresponding to the output result, thereby improving the accuracy of the model training.
  • an embodiment of the present application provides a method for updating a model, including: a second function entity receives a model update policy sent by a first function entity, where the model update policy includes an update condition of the model, where the model is used for guiding Parameter adjustment of the network element; the second functional entity determines whether the model satisfies the update condition according to the model update policy; and when the second function entity determines that the model satisfies the update condition, The first functional entity sends an update request, and the update request is used to trigger the first functional entity to perform a process of updating the model.
  • the model update mechanism of the present application replaces the model update mechanism from the active cycle update to the passive trigger update, the first functional entity assigns the model update policy to the second functional entity, and the second functional entity monitors the model to meet the update condition. And triggering an update request to the first functional entity, so that the first functional entity receives the update request, and performs a process of updating the model, wherein the update condition indicates that the network environment changes, and the network environment changes to meet the update condition.
  • the model is updated in time, thereby avoiding the problem of network performance degradation or resource waste caused by actively updating the model mode.
  • Model update strategies can be configured according to the scenario, resulting in greater flexibility.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is an intelligent collaborative function APF entity
  • the model update policy further includes an index of the first data, where the a data is used to determine whether the model satisfies the update condition
  • the method further comprises: sending, by the APF entity, a data service function DSF entity a data request, the data request including an index of the first data; the APF entity receiving the first data sent by the DSF entity according to an index of the first data; the second functional entity determining the Whether the model satisfies the update condition includes: the APF entity determining, according to the first data, whether the model satisfies the update condition.
  • the APF subscribes to the data from the DSF to judge the state of the network environment. If the performance of the model does not decrease, the update may not be performed, thus saving the computational resources and the model transmission process of the training model; if the network performance is degraded, The APF sends an update request to trigger the AMF retraining model to achieve the purpose of updating the model.
  • the method further includes: the APF entity receiving an output result of the model sent by a model execution function MEF entity; the APF entity determining a parameter adjustment of the network element according to the output result action;
  • the APF entity sends a parameter adjustment action of the network element to the network element, and is used to instruct the network element to adjust parameters of the network element according to the parameter adjustment action.
  • the method further includes: the APF entity sending the output result and a parameter adjustment action of the network element corresponding to the output result to the DSF entity, to facilitate the DSF
  • the entity tags the data that produced the output for updating the model.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is a model execution function MEF entity
  • the model update policy further includes an index of the first data
  • the first data is used to determine whether the model satisfies the update condition
  • the method further includes: the MEF entity to a data service function DSF
  • the entity sends a first data request, where the first data request includes an index of the first data
  • the MEF entity receives the first data that is determined by the DSF entity and is determined according to an index of the first data
  • Determining whether the model satisfies the update condition the method includes: determining, by the MEF entity, whether the model satisfies the update condition according to the first data.
  • the MEF subscribes data from the DSF to judge the state of the network environment. If the performance of the model does not decrease, the update may not be performed, thereby saving the computational resources and the model transmission process of the training model; The MEF sends an update request to trigger the AMF retraining model to achieve the purpose of updating the model.
  • the method further includes: the MEF receiving the model sent by the AMF;
  • the MEF entity Sending, by the MEF entity, a second data request to the DSF entity, the second data request including an index of second data used to characterize a current network environment state in which the network element is located; the MEF entity receiving the DSF The second data determined by the entity according to the index of the second data; the MEF entity inputs the second data into the model to obtain an output result, where the result includes the network element in the current a parameter adjustment action in a network environment state; the MEF entity sends the output result to the APF entity.
  • the embodiment of the present application provides a method for updating a model, including: a data service function DSF entity receives a data request sent by a first functional entity, where the data request includes an index of the first data, and the first data is used by Determining whether the model run by the model execution function MEF satisfies an update condition, the model is used to guide parameter adjustment of the network element; the DSF entity sends the identifier to the first functional entity according to the index of the first data The first data; wherein the first functional entity is an intelligent cooperative function APF entity or is the MEF entity.
  • the first functional entity subscribes the first data from the DSP to monitor whether the model satisfies the update condition based on the subscribed first data, so as to trigger the update when it is determined that the update condition is met.
  • the method further includes: receiving, by the DSF entity, an output result of the model sent by the APF entity, and the output result Corresponding parameter adjustment action of the network element; the DSF marks data generating the output result for updating the model.
  • the method further includes: the DSF entity receiving a data request sent by an analysis and modeling function AMF entity, the data request being used to request second data required to update the model; Transmitting, by the DSF entity, second data required to update the model to the AMF entity, where the second data includes the output result, data generating the output result, and the network corresponding to the output result Meta parameter adjustment action.
  • the APF when the APF triggers the model update, the APF sends the output result of the model and the parameter adjustment action corresponding to the output result to the DSF, so that the DSF marks the data, and when the model needs to be updated, the AMF performs model update according to the data. More accurate.
  • the embodiment of the present application provides a model updating apparatus, where the apparatus may be applied to a first functional entity, where the apparatus may be a first functional entity, or may be a chip capable of implementing a function corresponding to the first functional entity.
  • the device has the functionality to implement the various embodiments of the first aspect described above. This function can be implemented in hardware or in hardware by executing the corresponding software.
  • the hardware or software includes one or more modules corresponding to the functions described above.
  • the first functional entity includes the following modules:
  • a sending module configured to send a model update policy to the second function entity, where the model update policy includes an update condition of a model of the first network element, where the model of the first network element is used to guide the first network element Parameter adjustment;
  • a receiving module configured to receive an update request, where the update request is triggered by the second functional entity when determining that the model satisfies the update condition
  • a processing module configured to: when the receiving module receives the update request, perform a process of updating a model of the first network element.
  • the model update policy further includes an index of the first data, and the first data is used to determine whether the model satisfies the update condition.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is an intelligent collaborative function APF entity
  • the sending module sends a model of the first network element to a model execution function MEF entity before the processing module performs a process of updating the model of the first network element; and receiving, by the receiving module, the update request
  • the sending module is further configured to send a data request to the data service function DSF entity, where the data request is used to request second data corresponding to the first network element required for performing the model update; the receiving module, And the second data corresponding to the first network element sent by the DSF entity, where the second data includes an output result of the model of the first network element determined by the MEF entity, and the And the APF entity re-trains the model according to the second data, and sends the retrained model to the MEF.
  • the system applied by the device includes a plurality of network elements including the first network element, where the plurality of network elements all correspond to a model of the first network element, and the multiple The network element is managed by at least two APF entities, the first functional entity is an AMF entity, and the second functional entity is any one of the at least two APF entities, and related data of the multiple network elements are The at least two DSF entities are managed, and the corresponding models of the plurality of network elements are configured in the at least two MEFs;
  • the processing module determines, when the process of updating the model is executed, that the receiving module receives the update in the preset duration.
  • the number of the requested APF entities reaches a preset threshold; the sending module is configured to send a data request to the at least two DSF entities, where the data request is used to request the multiple network elements required for performing the model update.
  • the sending module is configured to send a data request to the at least two DSF entities, where the data request is used to request the multiple network elements required for performing the model update.
  • the receiving module is configured to receive third data corresponding to the plurality of network elements respectively sent by the at least two DSF entities, where the processing module is configured to perform third according to the plurality of network elements respectively.
  • the data retrains the model and sends the retrained models to the at least two MEFs, respectively.
  • the first functional entity is an AMF entity
  • the second functional entity is an MEF entity
  • the sending module is further configured to send the model of the first network element to the MEF entity
  • the processing module is further configured to: when the receiving module receives the update request, trigger the sending module to send a data request to a DSF entity, where the data request is used to request to perform a model update of the first network element.
  • the fourth data corresponding to the first network element is required; the receiving module is further configured to receive the fourth data sent by the DSF entity, where the fourth data includes the The output result of the model of the first network element, the data of the output result and the adjustment action of the first network element parameter corresponding to the output result; the processing module is further configured to retrain according to the four data The model of the first network element and the retrained model is sent to the MEF.
  • an embodiment of the present application provides a device, where the device is applied to a first functional entity, including: a processor, a communication interface, and a memory; the memory is configured to store an instruction, when the device is running, the communication interface For transmitting and receiving data, the processor executes the instruction stored by the memory to cause the apparatus to perform the model updating method in any of the implementation methods of the first aspect or the first aspect.
  • the memory may be integrated in the processor or may be independent of the processor.
  • the communication interface is configured to send a model update policy to the second function entity, where the model update policy includes an update condition of the model of the first network element, where the model of the first network element is used to guide the first Parameter adjustment of the network element; and receiving an update request; wherein the update request is triggered by the second functional entity when determining that the model satisfies the update condition;
  • a processor configured to perform a process of updating a model of the first network element when receiving the update request by using the communication interface.
  • the model update policy further includes an index of the first data, and the first data is used to determine whether the model satisfies the update condition.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is an intelligent collaborative function APF entity
  • the communication interface is further configured to send the model of the first network element to a model execution function MEF entity before the processor performs a process of updating a model of the first network element;
  • the processor is specifically configured to: when performing a process of updating a model of the first network element:
  • Second data corresponding to the first network element that is sent by the DSF entity where the second data includes an output result of a model of the first network element determined by the MEF entity, and The parameter adjustment action of the first network element determined by the APF entity according to the output result;
  • the system for the device includes a plurality of network elements including the first network element, where the plurality of network elements respectively correspond to a model of the first network element, and the The plurality of network elements are managed by at least two APF entities, the first functional entity is an AMF entity, and the second functional entity is any one of the at least two APF entities, and related data of the multiple network elements And being managed by the at least two DSF entities, where the corresponding models of the plurality of network elements are configured in the at least two MEFs;
  • the processor after transmitting the model update policy to each APF entity through the communication interface, determines to receive the APF that sends the update request within a preset duration when performing the process of updating the model.
  • the number of the entities reaches a preset threshold; the data request is sent to the at least two DSF entities by the communication interface, where the data request is used to request the third of the plurality of network elements respectively required for the model update
  • the retrained models are sent to the at least two MEFs, respectively.
  • the first functional entity is an AMF entity
  • the second functional entity is an MEF entity
  • the communication interface is further configured to send the model of the first network element to the MEF entity;
  • the processor is specifically configured to: when performing a process of updating a model of the first network element:
  • the fourth data sent by the DSF entity, where the fourth data includes an output result of a model of the first network element determined by the MEF entity, and generating data of the output result And an adjustment action of the first network element parameter corresponding to the output result;
  • the embodiment of the present application provides a model updating apparatus, where the apparatus is applied to a second functional entity, and the apparatus may be a second functional entity, or may be a chip capable of implementing a function corresponding to the second functional entity.
  • the device has the functionality to implement the various embodiments of the second aspect described above. This function can be implemented in hardware or in hardware by executing the corresponding software.
  • the hardware or software includes one or more modules corresponding to the functions described above.
  • the second functional entity may be an APF entity or an MEF entity.
  • the second functional entity includes:
  • a receiving module configured to receive a model update policy sent by the first function entity, where the model update policy includes an update condition of the model, where the model is used to guide parameter adjustment of the network element;
  • a processing module configured to determine, according to the model update policy, whether the model meets the update condition
  • a sending module configured to send an update request to the first functional entity when the processing module determines that the model meets the update condition, where the update request is used to trigger the first functional entity to perform an update of the model Process.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is an intelligent collaborative function APF entity
  • the model update policy further includes an index of the first data, The first data is used to determine whether the model satisfies the update condition
  • the sending module in the APF entity is further configured to: before the processing module in the APF determines that the model meets the update condition, send a data request to a data service function DSF entity, where the data request includes the first Index of data;
  • the receiving module in the APF entity is further configured to receive the first data that is sent by the DSF entity according to the index of the first data; and the processing module in the APF entity determines whether the model meets the update.
  • the condition is specifically configured to determine, according to the first data, whether the model satisfies the update condition.
  • the receiving module in the APF entity is further configured to receive an output result of the model sent by the model execution function MEF entity;
  • the processing module in the APF entity is further configured to determine a parameter adjustment action of the network element according to the output result;
  • the sending module of the APF entity is further configured to send a parameter adjustment action for determining the network element to the network element, where the network element is instructed to adjust a parameter of the network element according to the parameter adjustment action.
  • the sending module in the APF entity is further configured to send the output result and the parameter adjustment action of the network element corresponding to the output result to the DSF entity, so as to facilitate The DSF entity marks the data that produced the output for updating the model.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is a model execution function MEF entity
  • the model update policy further includes an index of the first data
  • the first data is used to determine whether the model satisfies the update condition
  • the sending module in the MEF is further configured to send, to the data service function DSF entity, before the processing module determines whether the model satisfies the update condition a data request, the first data request including an index of the first data
  • the receiving module in the MEF entity is further configured to receive, by the DSF entity, the first data that is determined according to an index of the first data;
  • the receiving module in the MEF is further configured to receive the model sent by the AMF
  • the sending module in the MEF entity is further configured to send the second data to the DSF entity.
  • the request, the second data request includes an index of the second data used to represent the current network environment state in which the network element is located
  • the receiving module in the MEF entity is further configured to receive the base information sent by the DSF entity
  • the second data determined by the index of the second data
  • the processing module in the MEF entity is further configured to input the second data into the model to obtain an output result, where the result includes the network element
  • the sending module in the MEF entity is further configured to send the output result to the APF entity.
  • the embodiment of the present application provides a device, where the device is applied to a second functional entity, including: a processor, a communication interface, and a memory; the memory is configured to store an instruction, when the device is running, the communication interface For transmitting and receiving data, the processor executes the instruction stored by the memory to cause the apparatus to perform the model updating method in any of the implementation methods of the second aspect or the second aspect.
  • the memory may be integrated in the processor or may be independent of the processor.
  • the communication interface is configured to receive a model update policy sent by the first functional entity, where the model update policy includes an update condition of the model, where the model is used to guide parameter adjustment of the network element, and the processor is configured to Determining whether the model satisfies the update condition; and determining that the model satisfies the update condition, sending an update request to the first functional entity through the communication interface, the update request being used to trigger The first functional entity performs a process of updating the model.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is an intelligent collaborative function APF entity
  • the model update policy further includes an index of the first data, The first data is used to determine whether the model satisfies the update condition
  • the processor in the APF entity is further configured to: before determining that the model meets the update condition, send a data request to the data service function DSF entity by using the communication interface, where the data request includes the first data Receiving, by the communication interface in the APF entity, the first data that is sent by the DSF entity according to the index of the first data; when determining whether the model meets the update condition, specifically used according to The first data determines whether the model satisfies the update condition.
  • the communication interface in the APF entity is further configured to receive an output result of the model sent by the model execution function MEF entity; the processor in the APF entity is further configured to use the output result according to the output Determining a parameter adjustment action of the network element; the communication interface in the APF entity is further configured to send a parameter adjustment action for determining the network element to the network element, where the network element is instructed to adjust according to the parameter The action adjusts parameters of the network element.
  • the communication interface in the APF entity is further configured to send the output result and the parameter adjustment action of the network element corresponding to the output result to the DSF entity, so as to facilitate the The DSF entity marks the data that produced the output for updating the model.
  • the first functional entity is an analysis and modeling function AMF entity
  • the second functional entity is a model execution function MEF entity
  • the model update policy further includes an index of the first data
  • the first data is used to determine whether the model satisfies the update condition
  • the processor in the MEF is further configured to: pass, by using a communication interface in the MEF, before determining whether the model meets the update condition
  • the data service function DSF entity sends a first data request, the first data request includes an index of the first data, and receives, by using a communication interface in the MEF, an index according to the first data sent by the DSF entity.
  • the determining, by the processor in the MEF whether the model satisfies the update condition according to the first data, when determining whether the model meets the update condition.
  • the communication interface in the MEF is further configured to: receive the model sent by the AMF; send a second data request to the DSF entity, where the second data request includes An index of the second data that is indicative of the current network environment state in which the network element is located; receiving the second data determined by the DSF entity according to the index of the second data; the processor in the MEF, further And inputting the second data into the model to obtain an output result, where the result includes a parameter adjustment action of the network element in the current network environment state; a communication interface in the MEF is further used to The output is sent to the APF entity.
  • the embodiment of the present application provides a model updating apparatus, where the apparatus is applied to a data service function DSF entity, and the apparatus may be a DSF or a chip capable of implementing a function corresponding to the DSF.
  • the device has the functionality to implement the various embodiments of the third aspect described above. This function can be implemented in hardware or in hardware to implement the corresponding software.
  • the hardware or software includes one or more modules corresponding to the functions described above.
  • a receiving module configured to receive a data request sent by the first functional entity, where the data request includes an index of the first data, where the first data is used to determine whether a model run by the model execution function MEF satisfies an update condition, and the model Used to guide parameter adjustment of network elements;
  • a processing module configured to determine first data according to an index of the first data
  • a sending module configured to send the first data determined by the processing module to the first functional entity, where the first functional entity is an intelligent cooperative function APF entity or is the MEF entity.
  • the receiving module is further configured to receive an output result of the model sent by the APF entity and corresponding to the output result.
  • the parameter adjustment action of the network element; the processing module is further configured to mark data that generates the output result, and is used to update the model.
  • the receiving module is further configured to receive a data request sent by an analysis and modeling function AMF entity, where the data request is used to request second data required to update the model; a module, configured to send, to the AMF entity, second data required to update the model, where the second data includes the output result, data that generates the output result, and a corresponding location of the output result The parameter adjustment action of the network element.
  • the embodiment of the present application provides an apparatus, where the apparatus is applied to a DSF entity, including: a processor, a communication interface, and a memory; the memory is configured to store an instruction, and when the apparatus is running, the communication interface is used. Transmitting and transmitting data, the processor executing the instruction stored in the memory, to cause the apparatus to perform the model updating method in any of the implementation methods of the third aspect or the third aspect.
  • the memory may be integrated in the processor or may be independent of the processor.
  • the DSF entity includes: a communication interface, configured to receive a data request sent by the first functional entity, where the data request includes an index of the first data, where the first data is used to determine whether a model run by the model execution function MEF is The update condition is met, the model is used to guide parameter adjustment of the network element, and the processor is configured to obtain an index of the first data in the data request, and determine the first data according to the index of the first data.
  • the communication interface is further configured to send the first data to the first functional entity, where the first functional entity is an intelligent cooperative function APF entity or is the MEF entity.
  • the communication interface is further configured to receive an output result of the model sent by the APF entity and corresponding to the output result. Parameter adjustment action of the network element;
  • the processor is further configured to mark data that generates the output result for updating the model.
  • the communication interface is further configured to receive a data request sent by an analysis and modeling function AMF entity, where the data request is used to request second data required to update the model;
  • the processor is further configured to determine the second data according to the data request, send the second data to the AMF entity by using the communication interface, where the second data includes the output result, Generating the data of the output result and the parameter adjustment action of the network element corresponding to the output result.
  • the embodiment of the present application provides a system, including: an analysis and modeling function AMF entity, an intelligent collaboration function APF entity, and the AMF entity, configured to send a model update policy to the APF entity, where the model
  • the update policy includes an update condition of a model of the network element, the model is used to guide parameter adjustment of the network element, and the APF entity is configured to receive the model update policy sent by the AMF entity, where And when the model satisfies the update condition included in the model update policy, sending an update request to the AMF entity, where the update request is used to trigger the AMF entity to perform a process of updating the model; the AMF entity is configured to receive When the update request is made, a process of updating the model is performed.
  • the AMF entity is used to implement the method flow executed by the first functional entity described in any of the first aspects.
  • the APF entity is used to implement the method flow of the APF entity execution described in any of the second aspects.
  • the system can also include a DSP entity for performing the method flow of the DSF entity execution of any of the three aspects of the design.
  • the system may further include other devices that interact with the two functional entities in the solution provided by the embodiment of the present application, for example, a network element and an MEF entity.
  • the embodiment of the present application provides another system, including: an analysis and modeling function AMF entity and a model execution function MEF entity; the AMF entity, configured to send a model update policy to the MEF entity, where
  • the model update policy includes an update condition of a model of the network element, the model is used to guide parameter adjustment of the network element, and the MEF entity is configured to receive the model update policy sent by the AMF entity, and determine And when the model satisfies the update condition included in the model update policy, sending an update request to the AMF entity, where the update request is used to trigger the AMF entity to perform a process of updating the model; the AMF entity is used to Upon receiving the update request sent by the MEF entity, a process of updating the model is performed.
  • the AMF entity is used to implement the method flow executed by the first functional entity described in any of the first aspects.
  • the MEF entity is used to implement the method flow of the APF entity execution described in any of the second aspects.
  • the system can also include a DSP entity for performing the method flow of the DSF entity execution of any of the three aspects of the design.
  • the system may further include other devices that interact with the two functional entities in the solution provided by the embodiment of the present application, for example, a network element and an APF entity.
  • the embodiment of the present application further provides a readable storage medium, where the readable storage medium stores a program or an instruction, and when it is run on a computer, the method for selecting any network element of the foregoing aspects is selected. Executed.
  • the embodiment of the present application further provides a computer program product comprising instructions, when executed on a computer, causing the computer to perform a selection method of any of the foregoing network elements.
  • FIG. 1 is a schematic structural diagram of a core network according to an embodiment of the present application.
  • FIG. 2 is a schematic structural diagram of an access network according to an embodiment of the present application.
  • FIG. 3 is a schematic diagram of a network architecture according to an embodiment of the present application.
  • FIG. 4 is a flowchart of a method for updating a model according to an embodiment of the present application.
  • FIG. 5 is a schematic flowchart of a method for updating a model corresponding to an APF entity triggered according to an embodiment of the present disclosure
  • FIG. 6 is a schematic flowchart of a method for updating a global triggering model according to an embodiment of the present disclosure
  • FIG. 7 is a schematic flowchart of a method for updating a model triggered by an MEF entity according to an embodiment of the present disclosure
  • FIG. 8 is a schematic structural diagram of an apparatus 800 according to an embodiment of the present application.
  • FIG. 9 is a schematic structural diagram of an apparatus 900 according to an embodiment of the present application.
  • FIG. 10 is a schematic structural diagram of an apparatus 1000 according to an embodiment of the present application.
  • FIG. 11 is a schematic structural diagram of a device 1100 according to an embodiment of the present application.
  • FIG. 12 is a schematic structural diagram of an apparatus 1200 according to an embodiment of the present application.
  • FIG. 13 is a schematic structural diagram of an apparatus 1300 according to an embodiment of the present application.
  • FIG. 14 is a schematic structural diagram of a device 1400 according to an embodiment of the present application.
  • FIG. 15 is a schematic structural diagram of an apparatus 1500 according to an embodiment of the present application.
  • the model updating method provided by the embodiment of the present application is applicable to different wireless access technology communication systems, such as a third generation (3rd generation, 3G) communication system, a long term evolution (LTE) system, and a fifth generation ( 5th generation, 5G) communication system, or more possible communication systems in the future.
  • 3rd generation, 3G third generation
  • LTE long term evolution
  • 5th generation, 5G fifth generation
  • the access network is responsible for the wireless side access of the terminal, and the possible deployment modes of the access network (AN) device include: a centralized unit (CU) and a distributed unit (DU) separation scenario; Single site scenario.
  • a single site includes a gNB/NR-NB, a transmission reception point (TRP), an evolved Node B (eNB), a radio network controller (RNC), and a Node B (Node B, NB), base station controller (BSC), base transceiver station (BTS), home base station (for example, home evolved NodeB, or home Node B, HNB), baseband unit (BBU) ), or wireless fidelity (Wifi) access point (AP), etc.
  • the single site is gNB/NR-NB.
  • the CU supports protocols such as radio resource control (RRC), packet data convergence protocol (PDCP), and service data adaptation protocol (SDAP).
  • RRC radio resource control
  • PDCP packet data convergence protocol
  • SDAP service data adaptation protocol
  • the CU is generally deployed at the central node and has rich computing resources.
  • the DU mainly supports radio link control (RLC), media access control (MAC), and physical layer (PHY) protocols.
  • RLC radio link control
  • MAC media access control
  • PHY physical layer
  • the DU generally adopts a distributed deployment mode. In a normal case, one CU needs to connect more than one DU.
  • the gNB has the functions of CU and DU and is usually deployed as a single site. DU and gNB are limited by factors such as the size and power consumption of the device, and usually the computing resources are limited.
  • the operation support system (OSS) of the access network is mainly used to configure parameters of the terminal device, collect alarms, performance statistics, running status, and logs of the terminal device.
  • a terminal device also called a user equipment (UE), a mobile station (MS), a mobile terminal (MT), etc., is a device that provides voice and/or data connectivity to users.
  • the terminal device includes a handheld device having a wireless connection function, an in-vehicle device, and the like.
  • the terminal devices can be: mobile phones, tablets, laptops, PDAs, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality ( Augmented reality, AR) wireless terminal in equipment, industrial control, wireless terminal in self driving, wireless terminal in remote medical surgery, smart grid A wireless terminal, a wireless terminal in a transportation safety, a wireless terminal in a smart city, or a wireless terminal in a smart home.
  • MIDs mobile internet devices
  • VR virtual reality
  • AR Augmented reality
  • wireless terminal in equipment industrial control
  • wireless terminal in self driving wireless terminal in remote medical surgery
  • smart grid A wireless terminal, a wireless terminal in a transportation safety, a wireless terminal in a smart city, or a wireless terminal in a smart home.
  • the mobile core network includes access and mobility management functions, session management functions (SMF), user plane functions (UPF), policy control functions (PCF), application functions, and the like. It is usually deployed on a cloud computing system in a centralized manner and connected to the access network through a transport network.
  • SMF session management functions
  • UPF user plane functions
  • PCF policy control functions
  • the model update manner provided by the embodiment of the present application can be implemented by a data analysis (DA) network architecture configured in a wireless network.
  • the DA network architecture can be configured in the core network or configured in the radio access network.
  • NWDA network data analysis
  • RANDA radio access network data analysis
  • NWDA is introduced in the core network.
  • NWDA is used to collect network data, training data analysis and evaluation models (referred to as models) from the functional entities of each core network, implement data analysis and prediction, and provide data analysis for PCF.
  • the PCF can make decisions based on the results of data analysis and the user's current business, generate new policy and charging control rules (PCC) rules, and improve the quality of service (QoS of users). ) to improve the user experience of the business.
  • PCC policy and charging control rules
  • QoS of users quality of service
  • the DA can include four functional entities: data service function (DSF) entity, analysis and modeling function (AMF) entity, and model execution function (MEF) entity. And adaptive policy function (APF) entities. See Figure 2 for a schematic diagram of the architecture of the DA deployed in the access network.
  • DSF data service function
  • AMF analysis and modeling function
  • MEF model execution function
  • APF adaptive policy function
  • the model is updated by signaling interaction between the four functional entities, and the network element parameter adjustment action is determined according to the updated model, and is sent to the network element (NE element) for performing the action. And according to the change of the network environment state after the network element performs the network element parameter adjustment action, to evaluate whether the model needs to be updated, and use the updated evaluation model to determine the next network element parameter adjustment action.
  • the DSF is used to collect data and pre-process the collected data, provide the AMF with the data needed to train or update the evaluation model, and provide the MEF with the network data needed to perform the evaluation model.
  • the network data may also be simply referred to as data.
  • the AMF is used to subscribe to the data required to train or update the evaluation model from the DSF, train or update the evaluation model based on the subscribed data, and send the evaluation model to the MEF. And, after receiving the network element parameter adjustment action of the APF feedback, iteratively updates the evaluation model according to the network element parameter adjustment action, and sends the updated evaluation model to the MEF or the APF.
  • the MEF is used to obtain an evaluation model from the AMF, obtain network data from the DSF, use the evaluation model to perform online prediction on the network data, obtain a network element parameter adjustment action, and send the network element parameter adjustment action to the APF.
  • APF is used to trigger a policy (such as a conflict handling policy) based on the result of analysis or prediction to change the network status, such as parameter adjustment, traffic engineering, resource scheduling, and the like.
  • a policy such as a conflict handling policy
  • the network element parameter adjustment action is obtained from the MEF, and the network element parameter adjustment action is sent to the network element that actually performs the network element parameter adjustment, thereby improving the capacity or performance of the network element. And, for feeding back the network element parameter adjustment action to the AMF.
  • the above four functional entities may be deployed on the access network or on the network elements of the core network. For example, they may be deployed on the CU, DU, gNB, and OSS of the access network, or deployed on the UPF or deployed as a whole. In the core network and connected to the PCF.
  • the above four functional entities may also be deployed in the access network for managing the entire access network, and then the above four functional entities are respectively deployed in each local network element.
  • the DA is deployed in the RAN.
  • CUDA, DUDA, and gNBDA are deployed separately in CU, DU, and gNB.
  • the above four functional entities may be deployed in the same network element or deployed in different network elements.
  • the four functional entities in the same network element perform signaling interaction to complete the model update method in the embodiment of the present application; in other application scenarios, the functional entities deployed in different network elements pass The interface between the network elements performs signaling interaction to complete the model update method in the embodiment of the present application.
  • the computing resources of the DU are limited, and the evaluation model can be trained or updated by the AMF entity in the CU, and the training model or the updated evaluation model is executed by the MEF entity in the DU.
  • the evaluation model can be trained or updated by the DA in the RAN as a global deployment (specifically, the AMF in the RANDA), by using the DU, CU as a local deployment.
  • the DA (specifically MEF) in the gNB performs the training or the updated evaluation model.
  • the network element parameter involved in the embodiment of the present application may refer to various parameters in radio resource management (RRM), or various parameters in a radio transmission technology (RTT), or in an operation and maintenance system.
  • RRM radio resource management
  • RTT radio transmission technology
  • Various parameters may be: pilot power, reference signal (RS) power, antenna downtilt, long term evolution (LTE) reusable level difference threshold, measurement report (MR) Interference decision thresholds, etc.
  • the method of actively updating the model is generally adopted. For example, fixed time period update, set the model update time (such as 1 week, 1 day, 1 hour, etc.) when configuring the model. After the model is deployed, start recording the running time. After the update time period is reached, use this period of time. The data trains the new model and deploys the trained model. This way of actively updating the model does not accurately sense changes in the network environment. During the model deployment run, the network may change before the update time point, causing the model to no longer apply to the current situation, resulting in erroneous results, resulting in network performance degradation or inoperability.
  • the model deployment run the network may change before the update time point, causing the model to no longer apply to the current situation, resulting in erroneous results, resulting in network performance degradation or inoperability.
  • the model can be deployed close to the application network element. Each distributed network element needs to be updated every cycle. A large number of model transmissions waste the transmission resources of the network. In short, the way to actively update the model cannot flexibly configure and sense changes in the network environment, which will occupy redundant computing resources and bandwidth.
  • the present application provides a model updating method and apparatus for solving the problem of low flexibility of the existing update model.
  • the method and the device are based on the same inventive concept. Since the principles of the method and the device for solving the problem are similar, the implementation of the device and the method can be referred to each other, and the repeated description is not repeated.
  • FIG. 4 it is a schematic flowchart of a method for updating a model provided by an embodiment of the present application.
  • the first functional entity sends a model update policy to the second functional entity.
  • the model update policy includes an update condition of a model of the first network element, where the model is used to guide parameter adjustment of the first network element.
  • the model guidance in the different network environment states adjusts the parameters of the first network element differently.
  • the second functional entity receives a model update policy sent by the first functional entity.
  • the second functional entity determines, according to the model update policy, whether the model satisfies the update condition.
  • the second function entity sends an update request to the first function entity when determining that the model meets the update condition, where the update request is used to trigger the first function entity to perform update of the model. Process.
  • the first functional entity may be configured in the CU CUDA
  • the second functional entity may be the DUDA configured in the DU
  • the first functional entity may be the RANDA configured in the RAN
  • the second The functional entity may be a DA configured in the CU/DU/gNB
  • the first functional entity may be an OSSDA configured in the OSS
  • the second functional entity may be a DA configured in the CU/DU/gNB.
  • the first functional entity may be an AMF entity
  • the second functional entity may be an APF entity, that is, the update is triggered by the APF entity.
  • the AMF entity and the APF entity may be located in the same network element or in different network elements.
  • both AMF entities and APF entities are on CUDA.
  • the AMF entity is located on the CUDA
  • the APF entity is located on the DUDA.
  • the AMF entity is located on RANDA
  • the APF entity is located on gNBDA.
  • the AMF entity may configure a model update policy to the APF entity, and the APF entity obtains the monitoring data from the ADF entity based on the model update policy (or is referred to as the first data, etc., which is not limited in this application), and the monitoring data is used to determine Whether the model meets the update criteria in the model update strategy.
  • the APF entity determines whether the model needs to be updated according to the model update strategy and the monitoring data, and instructs the AMF entity to update the model when it is determined that the model needs to be updated.
  • the first functional entity may be an AMF entity in the DA
  • the second functional entity may be an MEF entity, that is, the update is triggered by the MEF entity.
  • the AMF entity and the MEF entity may be located in the same network element or in different network elements.
  • both AMF entities and MEF entities are on CUDA.
  • the AMF entity is located on the CUDA
  • the MEF entity is located on the DUDA.
  • the AMF entity is located on RANDA
  • the APFMEF entity is located on gNBDA.
  • the AMF entity configures a model update policy to the MEF entity, and the MEF entity acquires monitoring data from the DSF entity based on the model update policy.
  • the MEF entity determines whether the model needs to be updated according to the model update strategy and the monitoring data, and instructs the AMF entity to update the model when it is determined that the model needs to be updated.
  • model update method is described in detail by taking the APF entity trigger model update as an example.
  • the AMF entity, the APF entity, the MEF entity, and the DSF entity shown in FIG. 5 may be located in the same network element or in different network elements.
  • AMF entities, APF entities, MEF entities, and DSF entities are all on CUDA.
  • the AMF entity is located on the CUDA, and the APF entity, the MEF entity, and the DSF entity are located on the DUDA.
  • the AMF entity is located on the RANDA, and the APF entity, the MEF entity, and the DSF entity are located on the gNBDA.
  • the AMF entity sends model installation information to the MEF entity, where the model installation information includes a model of the first network element.
  • the model installation information may also include an index of data that the MEF needs to subscribe to the DSF entity when the model is run, and the data required to run the model is referred to as running data for convenience of subsequent description.
  • the AMF entity sends a model update policy to the APF entity.
  • the model update policy includes the update conditions of the model.
  • the model update strategy may also include an index of the monitoring data.
  • the model needs to adjust the target key performance indicator (KPI) in the NE, and the KPI may be a circuit switching (CS) call drop rate; or, traffic or The load is calculated in such a way that the traffic or load is a weighted sum of multiple cells.
  • KPI target key performance indicator
  • CS circuit switching
  • the update condition may be configured according to an actual application scenario to meet requirements in different application scenarios.
  • the threshold of the KPI triggers the update of the model when the KPI is below the threshold.
  • the traffic prediction scenario may be a threshold of prediction error, and the prediction error may be a difference between the actual traffic and the predicted traffic.
  • an adaptive modulation and coding (AMC) scenario may be a throughput threshold.
  • the traffic prediction scenario the actual traffic needs to be monitored. Therefore, the monitoring data can be traffic, or data used to calculate the process.
  • the monitoring data In the AMC scenario, the throughput needs to be monitored, the monitoring data can be throughput, or the data used to calculate the throughput. .
  • S501 and S502 may be sent in one message or may be sent in different messages.
  • the order of execution of S501 and S502 is not specifically limited.
  • the MEF entity sends a data request 1 to the DSF entity, the data request 1 being used to request data required to run the model.
  • the running data is acquired from the DSF entity according to the index of the running data in step S501. That is, the data request 1 includes an index of the running data.
  • the DSF entity sends the operation data to the MEF according to the data request 1. Specifically, the DSF entity periodically sends the running data to the MEF according to the index of the running data.
  • the operational data sent per cycle is data reflecting the state of the network environment during the cycle.
  • the APF entity sends a data request 2 to the DSF entity, the data request 2 being used to request monitoring data.
  • the monitoring data is acquired from the DSF entity according to the index of the monitoring data in step S502. That is, the data request 2 includes an index of the monitoring data.
  • the DSF entity sends the monitoring data to the APF entity according to the data request 2. Specifically, the DSF entity periodically sends the monitoring data to the MEF entity according to the index of the monitoring data.
  • sequential execution order between S503a and S504a is not limited.
  • S505 The MEF entity inputs the running data input model obtained by S503b to generate an output result, and sends the output result to the APF entity. Specifically, after acquiring the running data every cycle, the MEF entity inputs the obtained running data input model to generate an output result, and sends the output result to the APF entity.
  • the APF entity After receiving the output result, the APF entity determines a parameter adjustment action of the first network element according to the output result, and sends a parameter adjustment action of the first network element to the first network element.
  • the APF entity stores a correspondence between the output result and the parameter adjustment action, so as to determine a parameter adjustment action corresponding to the received output result according to the correspondence relationship.
  • the foregoing correspondence may also be stored in the MEF entity, so that after obtaining the output result of the model, the MEF entity determines the parameter adjustment action of the first network element according to the output result, and sends the parameter adjustment action of the first network element.
  • the APF entity is forwarded, so that the APF entity forwards the parameter adjustment action of the first network element to the first network element.
  • the model can also directly output the parameter adjustment action, that is, the output result is a parameter adjustment action.
  • the APF entity sends the output result of the model and the parameter adjustment action corresponding to the output result to the DSF entity.
  • the specific sending content may include: an identifier of the model, an output result of the model, an index of the data that generates the output result, and a parameter adjustment action corresponding to the output result.
  • the sending content may also not include an index of the data that generates the output result. Since the DSF entity periodically sends the running data to the MEF entity, the MEF entity also periodically runs the data input model to generate an output result, and periodically outputs the output result. The APF entity is also periodically sent to the DSF entity, so that the DSF entity can determine, based on which period of data sent, the content of the received APF entity is generated.
  • the DSF entity marks the data that generates the output result, and is used for updating the model.
  • the APF entity receives the monitoring data sent by the DSF entity, and determines whether the model performance is degraded according to the model update policy received in S502, that is, whether the model meets the update condition included in the model update policy according to the monitoring data. If so, an update request is sent to the AMF entity for triggering the model update.
  • the traffic of each time period (hourly) of the cell needs to be predicted, and the output of the model is the traffic of the next hour, that is, the predicted traffic, and the model update strategy may be The error threshold of the predicted traffic and the real traffic is calculated.
  • the prediction error between the predicted traffic and the real traffic is calculated, and the prediction error is compared with the threshold of the prediction error. If the threshold is greater than the prediction error, the trigger is triggered.
  • the model update process sends a model update request and vice versa.
  • the AMF entity after receiving the update request sent by the APF entity, the AMF entity sends a data request 3 to the DSF entity, where the data request 3 is used to request data of the first network element required for model update, and subsequently, for convenience of description, The data is called update data.
  • An index of the update data may be included in the data request 3.
  • the DSF entity sends the update data to the AMF entity according to the data request 3.
  • the update data includes an output result of the model determined by the MEF entity received at S508, and a parameter adjustment action corresponding to the output result determined by the APF entity according to the output result.
  • the AMF entity retrains the model according to the updated data, and sends the trained model to the MEF entity.
  • the AMF entity may update the model update policy and send the updated model update policy to the APF entity.
  • the model update mechanism is replaced by the active cycle update to the passive trigger update, so that the network can sense the effect of the model operation and the network state, and the model update policy can be configured according to the scenario, so that the flexibility is high.
  • the APF subscribes to the data from the DSF to judge the state of the network environment. If the performance of the model does not decrease, the update may not be performed, which saves the computing resources and the model transmission process of the training model; if the network performance decreases, the APF sends an update request. Trigger the AMF to retrain the model to achieve the purpose of updating the model.
  • the AMF entity triggered by the APF entity is updated only for the model with degraded performance.
  • a plurality of network elements including the first network element are included in the system, the models corresponding to the multiple network elements are the same, and the multiple network elements are composed of at least two APFs.
  • the entity management is performed by the at least two DSF entities, and the corresponding models of the plurality of network elements are configured in the at least two MEFs.
  • the AMF entity may be located in the RANDA, the APF entity, the MEF entity, and the DSF entity are located in the gNBDA, or the AMF entity is located in the CUDA, and the APF entity, the MEF entity, and the DSF entity are located in the DUDA.
  • the AMF may trigger an update for the models of the multiple network elements of the system.
  • the AMF entity determines that the number of APF entities that receive the update request reaches a preset threshold within a preset duration. Transmitting, by the AMF entity, a data request to the at least two DSF entities, where the data request is used to request data corresponding to the plurality of network elements required for performing model update; and the AMF entity receives the at least two Data corresponding to the plurality of network elements respectively sent by the DSF entity; the AMF entity retrains the model according to the data corresponding to the plurality of network elements respectively, and sends the retrained model to the at least two MEFs respectively .
  • the AMF entity triggers an update for the models of multiple network elements in the system.
  • the user configures a global update policy in the AMF entity.
  • a global update is triggered when the conditions included in the global update policy are met.
  • the condition may be that the number of update requests corresponding to the local model reaches a preset threshold.
  • the four network elements are network element 1 to network element 4.
  • the model of the four network elements is configured in four MEF entities, four network elements are managed by four APF entities, and related data of the N network elements are managed by four DSF entities.
  • the AMF entity performs the method steps described in S501 and S502 for the MEF entity corresponding to each network element, and details are not described herein again.
  • the N MEF entities, the N APF entities, and the N DSF entities are all performed according to the method steps described in S503a to S509, and details are not described herein again.
  • step S509 the AMF entity retrains the model by:
  • the AMF entity statistics receives the number of APF entities that send the update request within a preset duration.
  • the AMF entity When it is determined that the number of statistics reaches a preset threshold, the AMF entity sends a data request 4 to the four DSF entities respectively.
  • the data request 4 is used to request data corresponding to the four network elements required for model update.
  • the AMF entity receives update data corresponding to the four network elements sent by the four DSF entities.
  • Each network element has updated data corresponding to the current network environment state.
  • the update data corresponding to each of the plurality of network elements includes an output result of the model of the network element, and a parameter adjustment action corresponding to the output result.
  • the AMF entity retrains the model according to update data corresponding to the four network elements respectively.
  • the AMF entity sends the retrained models to four MEF entities respectively.
  • model update method is described in detail by taking the model update by the MEF entity as an example.
  • the AMF entity, the APF entity, the MEF entity, and the DSF entity shown in FIG. 7 may be located in the same network element or in different network elements.
  • AMF entities, APF entities, MEF entities, and DSF entities are all on CUDA.
  • the AMF entity is located on the CUDA, and the APF entity, the MEF entity, and the DSF entity are located on the DUDA.
  • the AMF entity is located on the RANDA, and the APF entity, the MEF entity, and the DSF entity are located on the gNBDA.
  • the AMF entity sends model installation information to the MEF entity, where the model installation information includes a model of the first network element.
  • the model installation information includes a model of the first network element.
  • the AMF entity sends a model update policy to the MEF entity.
  • model update policy refers to any of the above embodiments, and details are not described herein again.
  • the AMF entity may send the model installation information and the model update policy to the MEF entity through a message, and may also send the model installation information and the model update policy to the MEF entity by using two messages, which is not specifically limited in this application.
  • the MEF entity sends a data request 1 to the DSF entity, where the data request 1 is used to request acquisition of running data and monitoring data.
  • the data request 1 is used to request acquisition of running data and monitoring data.
  • the MEF entity may separately request to obtain the running data and the monitoring data by using two messages, which is not specifically limited in this application.
  • the DSF entity sends the running data and the monitoring data to the MEF entity. Specifically, the DSF entity periodically sends the running data and the monitoring data to the MEF entity.
  • the MEF entity receives the operational data input model to generate an output result, and sends the output result to the APF entity. Specifically, after acquiring the running data every cycle, the MEF entity inputs the obtained running data input model to generate an output result, and sends the output result to the APF entity.
  • the output result here is the parameter adjustment action.
  • the MEF entity sends the output result to the DSF entity.
  • the specific sending content may include: an identifier of the model, an output result of the model, and an index of the data that generates the output result.
  • the sending content may also not include an index of the data that generates the output result. Since the DSF entity periodically sends the running data to the MEF entity, the MEF entity also periodically runs the data input model to generate an output result, and periodically outputs the output result.
  • the APF entity is also periodically sent to the DSF entity, so that the DSF entity can determine, based on which period of data sent, the content of the received APF entity is generated.
  • the DSF entity After receiving the output result of the model, the DSF entity marks the data that generates the output result, so as to be used for updating the model.
  • the MEF entity receives the monitoring data sent by the DSF entity, determines whether the performance of the model is degraded according to the model update policy, that is, determines whether the model satisfies the update condition included in the model update policy according to the monitoring data, and if yes, sends the data to the AMF entity.
  • An update request is triggered to perform the model update.
  • the AMF entity After receiving the update request sent by the MEF entity, the AMF entity sends a data request 2 to the DSF entity, where the data request 2 is used to request update data.
  • An index of the update data may be included in the data request 2.
  • the DSF entity sends the update data to the AMF entity according to the data request 2.
  • the output of the model is included in the update data.
  • the AMF entity retrains the model according to the updated data, and sends the trained model to the MEF entity.
  • the AMF entity may update the model update policy and send the updated model update policy to the MEF entity.
  • the model update mechanism is replaced by the active cycle update to the passive trigger update, so that the network can sense the effect of the model operation and the network state, and the model update policy can be configured according to the scenario, so that the flexibility is high.
  • the MEF subscribes to the data from the DSF to judge the state of the network environment. If the performance of the model does not decrease, the update may not be performed, thus saving the computing resources and the model transmission process of the training model; if the network performance is degraded, the MEF sends an update request. Trigger the AMF to retrain the model to achieve the purpose of updating the model.
  • a plurality of network elements including the first network element are included in the system, the models corresponding to the multiple network elements are the same, and the multiple network elements are composed of at least two APFs.
  • the entity management is performed by the at least two DSF entities, and the corresponding models of the plurality of network elements are configured in the at least two MEFs.
  • the AMF entity may be located in the RANDA, the APF entity, the MEF entity, and the DSF entity are located in the gNBDA, or the AMF entity is located in the CUDA, and the APF entity, the MEF entity, and the DSF entity are located in the DUDA.
  • the AMF may trigger an update for the models of the multiple network elements of the system. Specifically, after the AMF entity sends the model update policy to each MEF entity, the AMF entity determines that the number of MEF entities that receive the update request reaches a preset threshold within a preset duration.
  • the AMF entity Transmitting, by the AMF entity, a data request to the at least two DSF entities, where the data request is used to request data corresponding to the plurality of network elements required for performing model update; and the AMF entity receives the at least two Data corresponding to the plurality of network elements respectively sent by the DSF entity; the AMF entity retrains the model according to the data corresponding to the plurality of network elements respectively, and sends the retrained model to the at least two MEFs respectively entity.
  • the embodiment of the present application further provides an apparatus.
  • the apparatus 800 is applied to an AMF entity.
  • the device 800 may specifically be a processor in an AMF entity, or a chip or chip system, or a functional module or the like.
  • the device may include a sending module 801, a receiving module 802, and a processing module 803.
  • the processing module 803 is configured to control and manage the actions of the device 800.
  • the sending module 801 is configured to perform the sending action performed by the first functional entity or the AMF entity in any embodiment of the present application, and the processing module 803 is used by the first functional entity and the processing performed by the AMF entity in any embodiment of the present application.
  • the receiving module 802 is configured to perform the receiving action performed by the first functional entity and the AMF entity in any embodiment of the present application, and the details are not repeated herein.
  • the processing module 803 can also be used to indicate a process related to the first functional entity or the AMF entity in any of the above embodiments and/or other processes of the technical solutions described herein.
  • the AMF entity 9100 may include a communication interface 910 and a processor 920.
  • the AMF entity 910 may further include a memory 930.
  • the memory 930 may be disposed inside the AMF entity, and may also be disposed outside the AMF entity.
  • the processing modules 803 shown in FIG. 8 above may each be implemented by the processor 920.
  • the sending module 801 and the receiving module 802 can be implemented by the communication interface 910.
  • the processor 920 receives information or messages through the communication interface 910 and is used to implement the methods performed by the AMF entity or the first functional entity described in Figures 4-7. In the implementation process, each step of the processing flow may complete the method executed by the AMF entity described in FIG. 4 to FIG. 7 through the integrated logic circuit of the hardware in the processor 920 or the instruction in the form of software.
  • the communication interface 910 in the embodiment of the present application may be a circuit, a bus, a transceiver, or any other device that can be used for information interaction.
  • the other device may be a device connected to the device 900, for example, the other device may be a DSF entity or an MEF entity or an APF entity or the like.
  • the processor 920 in the embodiment of the present application may be a general-purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or a transistor logic device, and a discrete hardware component, which may be implemented or executed.
  • a general purpose processor can be a microprocessor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application may be directly implemented as a hardware processor, or may be performed by a combination of hardware and software units in the processor.
  • the program code executed by the processor 920 for implementing the above method may be stored in the memory 930. Memory 930 is coupled to processor 920.
  • the coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, and may be in an electrical, mechanical or other form for information interaction between devices, units or modules.
  • Processor 920 may operate in conjunction with memory 930.
  • the memory 930 may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or may be a volatile memory such as a random access memory (random access memory). -access memory, RAM).
  • Memory 930 is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.
  • connection medium between the communication interface 910, the processor 920, and the memory 930 is not limited in the embodiment of the present application.
  • the memory 930, the processor 920, and the communication interface 910 are connected by a bus in FIG. 9.
  • the bus is indicated by a thick line in FIG. 9, and the connection manner between other components is only schematically illustrated. Not limited to limits.
  • the bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in Figure 9, but it does not mean that there is only one bus or one type of bus.
  • the embodiment of the present application further provides an apparatus.
  • the apparatus 1000 is applied to an MEF entity.
  • the device 1000 may specifically be a processor in an MEF entity, or a chip or chip system, or a functional module or the like.
  • the apparatus may include a receiving module 1001, a processing module 1002, and a transmitting module 1003.
  • the processing module 1002 is configured to control and manage the actions of the device 1000.
  • the sending module 1003 is configured to perform a sending action performed by the second functional entity or the MEF entity in any embodiment of the present application, where the processing module 1002 is used by the second functional entity and the MEF entity in any one of the embodiments of the present application.
  • the receiving module 1001 is configured to perform the receiving operation performed by the second functional entity and the MEF entity in any embodiment of the present application, and the repeated operations are not repeated herein.
  • the processing module 1002 may also be used to indicate a process involving a second functional entity or MEF entity in any of the above embodiments and/or other processes of the technical solutions described herein.
  • the MEF entity 1100 may include a communication interface 1110 and a processor 1120.
  • the memory 1130 may also be included in the MEF entity 1100.
  • the memory 1130 may be disposed inside the MEF entity, and may also be disposed outside the MEF entity.
  • the processing modules 1002 shown in FIG. 10 above may each be implemented by the processor 1120.
  • the sending module 1003 and the receiving module 1001 can be connected to the communication interface 1110.
  • the processor 1120 receives information or messages through the communication interface 1110 and is used to implement the method performed by the MEF entity or the second functional entity described in FIGS. 4-7. In the implementation process, each step of the processing flow may complete the method executed by the MEF entity described in FIG. 4 to FIG. 7 through the integrated logic circuit of the hardware in the processor 1120 or the instruction in the form of software.
  • the communication interface 1110 in the embodiment of the present application may be a circuit, a bus, a transceiver, or any other device that can be used for information interaction.
  • the other device may be a device connected to the device 1100, for example, the other device may be a DSF entity or an AMF entity or an APF entity or the like.
  • the processor 1120 in the embodiment of the present application may be a general-purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or a transistor logic device, and a discrete hardware component, which may be implemented or executed.
  • a general purpose processor can be a microprocessor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application may be directly implemented as a hardware processor, or may be performed by a combination of hardware and software units in the processor.
  • Program code for processor 1120 to implement the above methods may be stored in memory 1130. Memory 1130 is coupled to processor 1120.
  • the coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, and may be in an electrical, mechanical or other form for information interaction between devices, units or modules.
  • Processor 1120 may operate in conjunction with memory 1130.
  • the memory 1130 may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or a volatile memory such as a random access memory (random). -access memory, RAM).
  • Memory 1130 is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.
  • connection medium between the communication interface 1110, the processor 1120, and the memory 1130 is not limited in the embodiment of the present application.
  • the embodiment of the present application is connected by a bus between the memory 1130, the processor 1120, and the communication interface 1110 in FIG. 11.
  • the bus is indicated by a thick line in FIG. 11, and the connection manner between other components is only schematically illustrated. Not limited to limits.
  • the bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in Figure 11, but it does not mean that there is only one bus or one type of bus.
  • the embodiment of the present application further provides an apparatus.
  • the apparatus 1200 is applied to an APF entity.
  • the device 1200 may specifically be a processor in an APF entity, or a chip or chip system, or a functional module or the like.
  • the apparatus may include a receiving module 1201, a processing module 1202, and a transmitting module 1203.
  • the processing module 1202 is configured to control and manage the actions of the device 1200.
  • the sending module 1203 is configured to perform a sending action performed by the second function entity or the APF entity in any embodiment of the present application, where the processing module 1202 is used by the second function entity and the APF entity in any one of the embodiments of the present application.
  • the receiving module 1201 is configured to perform the receiving action performed by the second functional entity and the APF entity in any embodiment of the present application, and the details are not repeated herein.
  • the processing module 1202 may also be used to indicate a process involving a second functional entity or an APF entity in any of the above embodiments and/or other processes of the technical solutions described herein.
  • the APF entity 1300 may include a communication interface 1310 and a processor 1320.
  • the APF entity 1300 may further include a memory 1330.
  • the memory 1330 may be disposed inside the APF entity, and may also be disposed outside the APF entity.
  • the processing modules 1202 shown in FIG. 12 above may each be implemented by the processor 1320.
  • the transmitting module 1203 and the receiving module 1201 may be configured by the communication interface 1310.
  • the processor 1320 receives information or messages through the communication interface 1310 and is used to implement the method performed by the APF entity or the second functional entity described in FIGS. 4-7. In the implementation process, each step of the processing flow may complete the method performed by the APF entity described in FIG. 4 to FIG. 7 through the integrated logic circuit of the hardware in the processor 1320 or the instruction in the form of software.
  • the communication interface 1310 in the embodiment of the present application may be a circuit, a bus, a transceiver, or any other device that can be used for information interaction.
  • the other device may be a device connected to the device 1300, for example, the other device may be a DSF entity or an AMF entity or an MEF entity or the like.
  • the processor 1320 in the embodiment of the present application may be a general-purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or a transistor logic device, and a discrete hardware component, which may be implemented or executed.
  • a general purpose processor can be a microprocessor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application may be directly implemented as a hardware processor, or may be performed by a combination of hardware and software units in the processor.
  • Program code for processor 1320 to implement the above methods may be stored in memory 1330. Memory 1330 is coupled to processor 1320.
  • the coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, and may be in an electrical, mechanical or other form for information interaction between devices, units or modules.
  • Processor 1320 may operate in conjunction with memory 1330.
  • the memory 1330 may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or a volatile memory such as a random access memory (random). -access memory, RAM).
  • Memory 1330 is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.
  • connection medium between the communication interface 1310, the processor 1320, and the memory 1330 is not limited in the embodiment of the present application.
  • the embodiment of the present application is connected by a bus between the memory 1330, the processor 1320, and the communication interface 1310 in FIG. 13, and the bus is indicated by a thick line in FIG. 13, and the connection manner between other components is only schematically illustrated. Not limited to limits.
  • the bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in FIG. 13, but it does not mean that there is only one bus or one type of bus.
  • the embodiment of the present application further provides a device.
  • the device 1400 is applied to a DSF entity.
  • the device 1400 may specifically be a processor in a DSF entity, or a chip or chip system, or a functional module or the like.
  • the apparatus may include a receiving module 1401, a processing module 1402, and a transmitting module 1403.
  • the processing module 1402 is configured to control and manage the actions of the device 1400.
  • the sending module 1403 is configured to perform a sending action performed by a DSF entity in any embodiment of the present application, and the processing module 1402 is used in a processing action performed by a DSF entity in any embodiment of the present application, such as outputting a data input model.
  • the receiving module 1401 is configured to perform the receiving action performed by the DSF entity in any embodiment of the present application, and the details are not repeated herein.
  • the processing module 1402 may also be used to indicate a process involving a DSF entity in any of the above embodiments and/or other processes of the technical solutions described herein.
  • the DSF entity 1500 may include a communication interface 1510 and a processor 1520.
  • the memory 1530 may also be included in the DSF entity 1500.
  • the memory 1530 may be disposed inside the DSF entity, and may also be disposed outside the DSF entity.
  • the processing modules 1402 shown in FIG. 14 above may each be implemented by the processor 1520.
  • the sending module 1403 and the receiving module 1401 may be configured by the communication interface 1510.
  • the processor 1520 receives the information or message through the communication interface 1510 and is used to implement the method performed by the DSF entity or the second functional entity described in FIGS. 4-7. In the implementation process, each step of the processing flow may complete the method executed by the DSF entity described in FIG. 4 to FIG. 7 through the integrated logic circuit of the hardware in the processor 1520 or the instruction in the form of software.
  • the communication interface 1510 in the embodiment of the present application may be a circuit, a bus, a transceiver, or any other device that can be used for information interaction.
  • the other device may be a device connected to the device 1500, for example, the other device may be a DSF entity or an AMF entity or an MEF entity or the like.
  • the processor 1520 in the embodiment of the present application may be a general-purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or a transistor logic device, and a discrete hardware component, which may be implemented or executed.
  • a general purpose processor can be a microprocessor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application may be directly implemented as a hardware processor, or may be performed by a combination of hardware and software units in the processor.
  • Program code for processor 1520 to implement the above methods may be stored in memory 1530. Memory 1530 is coupled to processor 1520.
  • the coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, and may be in an electrical, mechanical or other form for information interaction between devices, units or modules.
  • Processor 1520 may operate in conjunction with memory 1530.
  • the memory 1530 may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or a volatile memory such as a random access memory (random). -access memory, RAM).
  • Memory 1530 is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.
  • connection medium between the communication interface 1510, the processor 1520, and the memory 1530 is not limited in the embodiment of the present application.
  • the memory 1530, the processor 1520, and the communication interface 1510 are connected by a bus in FIG. 15, and the bus is indicated by a thick line in FIG. 15, and the connection manner between other components is only schematically illustrated. Not limited to limits.
  • the bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in Figure 15, but it does not mean that there is only one bus or one type of bus.
  • the embodiment of the present application further provides a computer storage medium, where the software program stores a software program, and the software program can implement any one or more of the foregoing when being read and executed by one or more processors.
  • the computer storage medium may include various media that can store program codes, such as a USB flash drive, a removable hard disk, a read only memory, a random access memory, a magnetic disk, or an optical disk.
  • the embodiment of the present application further provides a chip, where the chip includes a processor, for implementing functions related to any one or more of the foregoing embodiments, for example, acquiring or processing information involved in the foregoing method or data.
  • the chip further includes a memory for the processor to execute necessary program instructions and data.
  • the chip can be composed of a chip, and can also include a chip and other discrete devices.
  • embodiments of the present application can be provided as a method, system, or computer program product.
  • the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment in combination of software and hardware.
  • the application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.
  • the computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device.
  • the apparatus implements the functions specified in one or more blocks of a flow or a flow and/or block diagram of the flowchart.
  • These computer program instructions can also be loaded onto a computer or other programmable data processing device such that a series of operational steps are performed on a computer or other programmable device to produce computer-implemented processing for execution on a computer or other programmable device.
  • the instructions provide steps for implementing the functions specified in one or more of the flow or in a block or blocks of a flow diagram.

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Stored Programmes (AREA)

Abstract

本申请公开了一种模型更新方法、装置及系统，用以解决现有技术存在的由于模型性能下降导致的网络性能下降的问题。方法，包括：第一功能实体向第二功能实体发送模型更新策略，所述模型更新策略包括第一网元的模型的更新条件，所述第一网元的模型用于指导对所述第一网元的参数调整；所述第一功能实体在接收到更新请求时，执行更新所述第一网元的模型的流程；其中，所述更新请求是所述第二功能实体在确定所述模型满足所述更新条件时触发的。其中，所述第一功能实体可以为分析和建模功能AMF实体，第二功能实体可以为智能协同功能APF实体或者为模型执行功能MEF实体。

Description

一种模型更新方法、装置及系统
本申请要求在2018年04月27日提交中国专利局、申请号为201810393730.0、发明名称为“一种模型更新方法、装置及系统”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及信息技术领域，尤其涉及一种模型更新方法、装置及系统。
背景技术
随着人工智能(artifical intelligence，AI)技术的发展，AI在无线网络中的应用更加广泛。人工智能网络架构中基于机器学习技术产生的模型来对移动网络产生的数据进行分析，并基于通过模型产生的分析结果来优化移动网络，以便于更好地支持用户业务，成为移动网络演进的趋势。
在模型运行的过程中，由于无线网络环境的变化，可能会导致模型的性能下降，使得网络不能正常运行，这就要一个模型更新的机制来及时触发模型更新，保证模型的有效性。现有的智能网络架构中，模型的更新是通过固定的时间间隔，以主动的方式通过网元新产生的数据重新训练模型。这种主动更新模型的方式不能准确感知网络环境的变化。在模型下发运行过程中，网络环境可能在模型更新时间到达之前就发生了改变，致使模型不再适用当时的情况，产生错误的结果，导致网络性能下降或者不能运行。
发明内容
本申请提供一种模型更新方法、装置及系统，用以解决现有技术存在的由于模型性能下降导致的网络性能下降的问题。
第一方面，本申请实施例提供了一种模型更新方法，包括：第一功能实体向第二功能实体发送模型更新策略，所述模型更新策略包括第一网元的模型的更新条件，所述第一网元的模型用于指导对所述第一网元的参数调整；所述第一功能实体在接收到更新请求时，执行更新所述第一网元的模型的流程；其中，所述更新请求是所述第二功能实体在确定所述模型满足所述更新条件时触发的。
其中，所述第一功能实体可以为分析和建模功能AMF实体，第二功能实体可以为智能协同功能APF实体或者为模型执行功能MEF实体。
本申请的模型更新机制，将模型更新机制从主动周期更新替换为被动触发更新，第一功能实体将模型更新策略配给第二功能实体，第二功能实体在监控到模型满足更新条件时向第一功能实体触发更新请求，从而第一功能实体在接收到更新请求，执行更新所述模型的流程，其中，满足更新条件则表明网络环境发生变化，从而网络环境发生变化满足更新条件时，及时更新所述模型，从而避免了主动更新模型方式可能导致的网络性能下降，或者资源浪费的问题。模型更新策略可以根据场景配置，从而灵活性较高。
在一种可能的设计中，所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；所述第一功能实体执行更新所述第一网元的模型的流程之前，所述方法还包括：
所述AMF实体向模型执行功能MEF实体发送所述第一网元的模型；
所述第一功能实体执行更新所述模型的流程，包括：
所述AMF实体向数据服务功能DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述第一网元对应的第二数据；
所述AMF实体接收所述DSF实体发送的所述第一网元对应的第二数据，所述第二数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，以及所述APF实体根据所述输出结果确定的所述第一网元的参数调整动作；
所述AMF实体根据所述第二数据重新训练模型，并将重新训练的模型发送给所述MEF。
通过上述设计，在更新模型时，基于模型输出结果以及模型结果对应的参数调整动作来训练模型，提高了模型训练的准确度。
在一种可能的设计中，所述方法应用的系统中包括所述第一网元在内的多个网元，所述多个网元均对应所述第一网元的模型，且所述多个网元由至少两个APF实体管理，所述第一功能实体为AMF实体，所述第二功能实体为所述至少两个APF实体中的任一个，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中；
所述AMF实体在分别向每个APF实体发送所述模型更新策略后，所述第一功能实体执行更新所述模型的流程时，包括：
所述AMF实体确定在预设时长内接收到发送所述更新请求的APF实体的数量达到预设阈值；
所述AMF实体分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的第三数据；
所述AMF实体接收所述至少两个DSF实体发来的所述多个网元分别对应的第三数据；
所述AMF实体根据所述多个网元分别对应的第三数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF。
通过上述设计，在多个模型性能发生恶化而满足更新条件，即有多个APF请求数据进行模型更新，从而AMF根据多个网元的数据对模型进行重新训练，从而提高了模型训练的准确度，并由于不需要针对每个网元分别训练模型，从而节省了资源。
在一种可能的设计中，所述第一功能实体为AMF实体，第二功能实体为MEF实体；所述方法还可以包括：所述AMF实体向所述MEF实体发送所述第一网元的模型；所述第一功能实体执行更新所述第一网元的模型的流程，包括：所述AMF实体向DSF实体发送数据请求，所述数据请求用于请求进行所述第一网元的模型更新所需的所述第一网元对应的第四数据；所述AMF实体接收所述DSF实体发送的所述第四数据，所述第四数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，产生所述输出结果的数据以及所述输出结果对应的所述第一网元参数的调整动作；所述AMF实体根据所述四数据重新训练所述第一网元的模型，并将重新训练的模型发送给所述MEF。
上述设计，通过MEF触发模型更新。AMF将模型更新策略配置给MEF，从而MEF 根据模型更新策略确定满足更新条件时触发模型更新。在训练模型时，根据未更新的模型的输出结果以及输出结果对应的参数调整动作来重新训练模型，从而提高了模型训练的准确度。
第二方面，本申请实施例提供了一种模型更新方法，包括：第二功能实体接收第一功能实体发送的模型更新策略，所述模型更新策略包括模型的更新条件，所述模型用于指导对网元的参数调整；所述第二功能实体根据所述模型更新策略确定所述模型是否满足所述更新条件；所述第二功能实体在确定所述模型满足所述更新条件时，向所述第一功能实体发送更新请求，所述更新请求用于触发所述第一功能实体执行更新所述模型的流程。
通过上述方案，本申请的模型更新机制，将模型更新机制从主动周期更新替换为被动触发更新，第一功能实体将模型更新策略配给第二功能实体，第二功能实体在监控到模型满足更新条件时向第一功能实体触发更新请求，从而第一功能实体在接收到更新请求，执行更新所述模型的流程，其中，满足更新条件则表明网络环境发生变化，从而网络环境发生变化满足更新条件时，及时更新所述模型，从而避免了主动更新模型方式可能导致的网络性能下降，或者资源浪费的问题。模型更新策略可以根据场景配置，从而灵活性较高。
在一种可能设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；所述第二功能实体在确定所述模型满足所述更新条件之前，所述方法还包括：所述APF实体向数据服务功能DSF实体发送数据请求，所述数据请求包括所述第一数据的索引；所述APF实体接收所述DSF实体根据所述第一数据的索引发送的所述第一数据；所述第二功能实体确定所述模型是否满足所述更新条件，包括：所述APF实体根据所述第一数据确定所述模型是否满足所述更新条件。
上述设计中，由APF从DSF中订阅数据来对网络环境状态进行判断，若模型性能没有下降，则可以不用进行更新，这样节省了训练模型的计算资源和模型传输过程；若网络性能下降，由APF发送更新请求触发AMF重新训练模型，达到更新模型的目的。
在一种可能的设计中，所述方法还包括：所述APF实体接收模型执行功能MEF实体发送的所述模型的输出结果；所述APF实体根据所述输出结果确定所述网元的参数调整动作；
所述APF实体将确定所述网元的参数调整动作发送给所述网元，用于指示所述网元根据所述参数调整动作调整所述网元的参数。
在一种可能的设计中，所述方法还包括：所述APF实体将所述输出结果以及所述输出结果对应的所述网元的参数调整动作发送给所述DSF实体，以便于所述DSF实体对产生所述输出结果的数据进行标记，用于更新所述模型。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为模型执行功能MEF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；所述第二功能实体在确定所述模型是否满足所述更新条件之前，所述方法还包括：所述MEF实体向数据服务功能DSF实体发送第一数据请求，所述第一数据请求包括所述第一数据的索引；所述MEF实体接收所述DSF实体发送的根据所述第一数据的索引确定的所述第一数据；所述第二功能实体确定所述模型是否满足所述更新条件，包括：所述MEF实体根据所述第一数据确定所述模型是否满足所述更新条件。
上述设计中，由MEF从DSF中订阅数据来对网络环境状态进行判断，若模型性能没有下降，则可以不用进行更新，这样节省了训练模型的计算资源和模型传输过程；若网络性能下降，由MEF发送更新请求触发AMF重新训练模型，达到更新模型的目的。
在一种可能的设计中，所述方法还包括：所述MEF接收所述AMF发送的所述模型；
所述MEF实体向所述DSF实体发送第二数据请求，所述第二数据请求包括用于表征所述网元所在的当前网络环境状态的第二数据的索引；所述MEF实体接收所述DSF实体发送的根据所述第二数据的索引确定的所述第二数据；所述MEF实体将所述第二数据输入所述模型得到输出结果，所述结果中包括所述网元在所述当前网络环境状态下的参数调整动作；所述MEF实体将所述输出结果发送给所述APF实体。
第三方面，本申请实施例提供了一种模型更新方法，包括：数据服务功能DSF实体接收第一功能实体发送的数据请求，所述数据请求包括第一数据的索引，所述第一数据用于确定模型执行功能MEF所运行的模型是否满足更新条件，所述模型用于指导对网元的参数调整；所述DSF实体根据所述第一数据的索引向所述第一功能实体发送所述第一数据；其中，所述第一功能实体为智能协同功能APF实体或者为所述MEF实体。
通过上述设计，第一功能实体从DSP订阅第一数据，从而基于订阅的第一数据来监控模型是否满足更新条件，以便于在确定满足更新条件时，触发更新。
在一种可能的设计中，所述第一功能实体为智能协同功能APF实体时，所述方法还包括：所述DSF实体接收所述APF实体发送的所述模型的输出结果以及所述输出结果对应的所述网元的参数调整动作；所述DSF对产生所述输出结果的数据进行标记，用于更新所述模型。
在一种可能的设计中，所述方法还包括：所述DSF实体接收分析和建模功能AMF实体发送的数据请求，所述数据请求用于请求更新所述模型所需的第二数据；所述DSF实体将更新所述模型所需的第二数据发送给所述AMF实体，所述第二数据中包括所述输出结果、产生所述输出结果的数据以及所述输出结果对应的所述网元的参数调整动作。
通过上述设计，由APF触发模型更新时，APF将模型的输出结果以及输出结果对应的参数调整动作发送给DSF，从而DSF对数据进行标记，在模型需要更新时，AMF根据这个数据进行模型更新相对较准确。
第四方面，本申请实施例提供了一种模型更新装置，所述装置可以应用于第一功能实体，该装置可以是第一功能实体，也可以是能够实现第一功能实体对应的功能的芯片。该装置具有实现上述第一方面的各实施例的功能。该功能可以通过硬件实现，也可以通过硬件执行相应的软件实现。该硬件或软件包括一个或多个与上述功能相对应的模块。
具体的，第一功能实体包括如下模块：
发送模块，用于向第二功能实体发送模型更新策略，所述模型更新策略包括第一网元的模型的更新条件，所述第一网元的模型用于指导对所述第一网元的参数调整；
接收模块，用于接收更新请求；其中，所述更新请求是所述第二功能实体在确定所述模型满足所述更新条件时触发的；
处理模块，用于在所述接收模块接收到所述更新请求时，执行更新所述第一网元的模型的流程。
在一种可能的设计中，所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；
所述发送模块，在处理模块在执行更新所述第一网元的模型的流程之前，向模型执行功能MEF实体发送所述第一网元的模型；在所述接收模块接收到所述更新请求时，所述发送模块，还用于向数据服务功能DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述第一网元对应的第二数据；所述接收模块，还用于接收所述DSF实体发送的所述第一网元对应的第二数据，所述第二数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，以及所述APF实体根据所述输出结果确定的所述第一网元的参数调整动作；所述处理模块，具体用于根据所述第二数据重新训练模型，并将重新训练的模型发送给所述MEF。
在一种可能设计中，所述装置所应用系统中包括所述第一网元在内的多个网元，所述多个网元均对应所述第一网元的模型，且所述多个网元由至少两个APF实体管理，所述第一功能实体为AMF实体，所述第二功能实体为所述至少两个APF实体中的任一个，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中；
所述发送模块在分别向每个APF实体发送所述模型更新策略后，所述处理模块，在执行更新所述模型的流程时，确定在预设时长内所述接收模块接收到发送所述更新请求的APF实体的数量达到预设阈值；所述发送模块，用于分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的第三数据；
所述接收模块，用于接收所述至少两个DSF实体发来的所述多个网元分别对应的第三数据；所述处理模块，用于根据所述多个网元分别对应的第三数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF。
在一种可能的设计中，所述第一功能实体为AMF实体，第二功能实体为MEF实体；所述发送模块，还用于向所述MEF实体发送所述第一网元的模型；所述处理模块，还用于在所述接收模块接收到所述更新请求时，触发所述发送模块向DSF实体发送数据请求，所述数据请求用于请求进行所述第一网元的模型更新所需的所述第一网元对应的第四数据；所述接收模块，还用于接收所述DSF实体发送的所述第四数据，所述第四数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，产生所述输出结果的数据以及所述输出结果对应的所述第一网元参数的调整动作；所述处理模块，还用于根据所述四数据重新训练所述第一网元的模型，并将重新训练的模型发送给所述MEF。
第五方面，本申请实施例提供一种装置，该装置应用于第一功能实体，包括：处理器、通信接口，还可以包括存储器；该存储器用于存储指令，当该装置运行时，通信接口用于收发数据，该处理器执行该存储器存储的该指令，以使该装置执行上述第一方面或第一方面的任一实现方法中的模型更新方法。需要说明的是，该存储器可以集成于处理器中，也可以是独立于处理器之外。
具体的，通信接口，用于向第二功能实体发送模型更新策略，所述模型更新策略包括第一网元的模型的更新条件，所述第一网元的模型用于指导对所述第一网元的参数调整；以及接收更新请求；其中，所述更新请求是所述第二功能实体在确定所述模型满足所述更新条件时触发的；
处理器，用于在通过所述通信接口接收到所述更新请求时，执行更新所述第一网元的 模型的流程。
在一种可能的设计中，所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，所述第二功能实体为智能协同功能APF实体；
所述通信接口，还用于在所述处理器执行更新所述第一网元的模型的流程之前，向模型执行功能MEF实体发送所述第一网元的模型；
所述处理器，在执行更新所述第一网元的模型的流程时，具体用于：
通过所述通信接口向数据服务功能DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述第一网元对应的第二数据；
通过所述通信接口接收所述DSF实体发送的所述第一网元对应的第二数据，所述第二数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，以及所述APF实体根据所述输出结果确定的所述第一网元的参数调整动作；
根据所述第二数据重新训练模型，并将重新训练的模型发送给所述MEF。
在一种可能的设计中，所述装置应用的系统中包括所述第一网元在内的多个网元，所述多个网元均对应所述第一网元的模型，且所述多个网元由至少两个APF实体管理，所述第一功能实体为AMF实体，所述第二功能实体为所述至少两个APF实体中的任一个，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中；
所述处理器，在通过所述通信接口分别向每个APF实体发送所述模型更新策略后，在执行更新所述模型的流程时，确定在预设时长内接收到发送所述更新请求的APF实体的数量达到预设阈值；通过所述通信接口分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的第三数据；通过所述通信接口接收所述至少两个DSF实体发来的所述多个网元分别对应的第三数据；根据所述多个网元分别对应的第三数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF。
在一种可能的设计中，所述第一功能实体为AMF实体，第二功能实体为MEF实体；
所述通信接口，还用于向所述MEF实体发送所述第一网元的模型；
所述处理器，在执行更新所述第一网元的模型的流程时，具体用于：
通过所述通信接口向DSF实体发送数据请求，所述数据请求用于请求进行所述第一网元的模型更新所需的所述第一网元对应的第四数据；
通过所述通信接口接收所述DSF实体发送的所述第四数据，所述第四数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，产生所述输出结果的数据以及所述输出结果对应的所述第一网元参数的调整动作；
根据所述四数据重新训练所述第一网元的模型，并将重新训练的模型发送给所述MEF。
第六方面，本申请实施例提供了一种模型更新装置，所述装置应用于第二功能实体，该装置可以是第二功能实体，也可以是能够实现第二功能实体对应的功能的芯片。该装置具有实现上述第二方面的各实施例的功能。该功能可以通过硬件实现，也可以通过硬件执行相应的软件实现。该硬件或软件包括一个或多个与上述功能相对应的模块。第二功能实体可以是APF实体也可以是MEF实体。
具体的，第二功能实体包括：
接收模块，用于接收第一功能实体发送的模型更新策略，所述模型更新策略包括模型的更新条件，所述模型用于指导对网元的参数调整；
处理模块，用于根据所述模型更新策略确定所述模型是否满足所述更新条件；
发送模块，用于在所述处理模块确定所述模型满足所述更新条件时，向所述第一功能实体发送更新请求，所述更新请求用于触发所述第一功能实体执行更新所述模型的流程。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；
所述APF实体中的发送模块，还用于在所述APF中的处理模块确定所述模型满足所述更新条件之前，向数据服务功能DSF实体发送数据请求，所述数据请求包括所述第一数据的索引；
所述APF实体中的接收模块，还用于接收所述DSF实体根据所述第一数据的索引发送的所述第一数据；所述APF实体中处理模块在确定所述模型是否满足所述更新条件，具体用于根据所述第一数据确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述APF实体中的接收模块，还用于接收模型执行功能MEF实体发送的所述模型的输出结果；
所述APF实体中的处理模块，还用于根据所述输出结果确定所述网元的参数调整动作；
所述APF实体中的发送模块，还用于将确定所述网元的参数调整动作发送给所述网元，用于指示所述网元根据所述参数调整动作调整所述网元的参数。
在一种可能的设计中，所述APF实体中的发送模块，还用于将所述输出结果以及所述输出结果对应的所述网元的参数调整动作发送给所述DSF实体，以便于所述DSF实体对产生所述输出结果的数据进行标记，用于更新所述模型。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为模型执行功能MEF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；所述MEF中的发送模块，还用于在处理模块确定所述模型是否满足所述更新条件之前，向数据服务功能DSF实体发送第一数据请求，所述第一数据请求包括所述第一数据的索引；
所述MEF实体中的接收模块，还用于接收所述DSF实体发送的根据所述第一数据的索引确定的所述第一数据；
所述MEF实体中的处理模块，在确定所述模型是否满足所述更新条件时，具体用于根据所述第一数据确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述MEF中的接收模块，还用于接收所述AMF发送的所述模型；所述MEF实体中的发送模块，还用于向所述DSF实体发送第二数据请求，所述第二数据请求包括用于表征所述网元所在的当前网络环境状态的第二数据的索引；所述MEF实体中的接收模块，还用于接收所述DSF实体发送的根据所述第二数据的索引确定的所述第二数据；所述MEF实体中的处理模块，还用于将所述第二数据输入所述模型得到输出结果，所述结果中包括所述网元在所述当前网络环境状态下的参数调整动作；所述MEF实体中的发送模块，还用于将所述输出结果发送给所述APF实体。
第七方面，本申请实施例提供一种装置，该装置应用于第二功能实体，包括：处理器、通信接口，还可以包括存储器；该存储器用于存储指令，当该装置运行时，通信接口用于 收发数据，该处理器执行该存储器存储的该指令，以使该装置执行上述第二方面或第二方面的任一实现方法中的模型更新方法。需要说明的是，该存储器可以集成于处理器中，也可以是独立于处理器之外。
具体的，通信接口，用于接收第一功能实体发送的模型更新策略，所述模型更新策略包括模型的更新条件，所述模型用于指导对网元的参数调整；处理器，用于根据所述模型更新策略确定所述模型是否满足所述更新条件；在确定所述模型满足所述更新条件时，通过所述通信接口向所述第一功能实体发送更新请求，所述更新请求用于触发所述第一功能实体执行更新所述模型的流程。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；
所述APF实体中的处理器，还用于：在确定所述模型满足所述更新条件之前，通过所述通信接口向数据服务功能DSF实体发送数据请求，所述数据请求包括所述第一数据的索引；通过所述APF实体中的通信接口接收所述DSF实体根据所述第一数据的索引发送的所述第一数据；在确定所述模型是否满足所述更新条件时，具体用于根据所述第一数据确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述APF实体中通信接口，还用于接收模型执行功能MEF实体发送的所述模型的输出结果；所述APF实体中处理器，还用于根据所述输出结果确定所述网元的参数调整动作；所述APF实体中通信接口，还用于将确定所述网元的参数调整动作发送给所述网元，用于指示所述网元根据所述参数调整动作调整所述网元的参数。
在一种可能的设计中，所述APF实体中通信接口，还用于将所述输出结果以及所述输出结果对应的所述网元的参数调整动作发送给所述DSF实体，以便于所述DSF实体对产生所述输出结果的数据进行标记，用于更新所述模型。
在一种可能的设计中，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为模型执行功能MEF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；所述MEF中的处理器，还用于：在确定所述模型是否满足所述更新条件之前，通过所述MEF中的通信接口向数据服务功能DSF实体发送第一数据请求，所述第一数据请求包括所述第一数据的索引；通过所述MEF中的通信接口接收所述DSF实体发送的根据所述第一数据的索引确定的所述第一数据；所述MEF中的处理器，在确定所述模型是否满足所述更新条件时，具体用于根据所述第一数据确定所述模型是否满足所述更新条件。
在一种可能的设计中，所述MEF中的通信接口，还用于：接收所述AMF发送的所述模型；向所述DSF实体发送第二数据请求，所述第二数据请求包括用于表征所述网元所在的当前网络环境状态的第二数据的索引；接收所述DSF实体发送的根据所述第二数据的索引确定的所述第二数据；所述MEF中的处理器，还用于将所述第二数据输入所述模型得到输出结果，所述结果中包括所述网元在所述当前网络环境状态下的参数调整动作；所述MEF中的通信接口，还用于将所述输出结果发送给所述APF实体。
第八方面，本申请实施例提供了一种模型更新装置，所述装置应用于数据服务功能DSF实体，该装置可以是DSF，也可以是能够实现DSF对应的功能的芯片。该装置具有实现上述第三方面的各实施例的功能。该功能可以通过硬件实现，也可以通过硬件执行相应的软 件实现。该硬件或软件包括一个或多个与上述功能相对应的模块。
接收模块，用于接收第一功能实体发送的数据请求，所述数据请求包括第一数据的索引，所述第一数据用于确定模型执行功能MEF所运行的模型是否满足更新条件，所述模型用于指导对网元的参数调整；
处理模块，用于根据所述第一数据的索引确定第一数据；
发送模块，用于将所述处理模块确定的第一数据发送给所述第一功能实体；其中，所述第一功能实体为智能协同功能APF实体或者为所述MEF实体。
在一种可能的设计中，所述第一功能实体为智能协同功能APF实体时，所述接收模块，还用于接收所述APF实体发送的所述模型的输出结果以及所述输出结果对应的所述网元的参数调整动作；所述处理模块，还用于对产生所述输出结果的数据进行标记，用于更新所述模型。
在一种可能的设计中，所述接收模块，还用于接收分析和建模功能AMF实体发送的数据请求，所述数据请求用于请求更新所述模型所需的第二数据；所述发送模块，还用于将更新所述模型所需的第二数据发送给所述AMF实体，所述第二数据中包括所述输出结果、产生所述输出结果的数据以及所述输出结果对应的所述网元的参数调整动作。
第九方面，本申请实施例提供一种装置，该装置应用于DSF实体，包括：处理器、通信接口，还可以包括存储器；该存储器用于存储指令，当该装置运行时，通信接口用于收发数据，该处理器执行该存储器存储的该指令，以使该装置执行上述第三方面或第三方面的任一实现方法中的模型更新方法。需要说明的是，该存储器可以集成于处理器中，也可以是独立于处理器之外。
具体的，DSF实体包括：通信接口，用于接收第一功能实体发送的数据请求，所述数据请求包括第一数据的索引，所述第一数据用于确定模型执行功能MEF所运行的模型是否满足更新条件，所述模型用于指导对网元的参数调整；处理器，用于获取所述数据请求中的第一数据的索引，并根据所述第一数据的索引确定所述第一数据；所述通信接口，还用于向所述第一功能实体发送所述第一数据；其中，所述第一功能实体为智能协同功能APF实体或者为所述MEF实体。
在一种可能的设计中，所述第一功能实体为智能协同功能APF实体时，所述通信接口，还用于接收所述APF实体发送的所述模型的输出结果以及所述输出结果对应的所述网元的参数调整动作；
所述处理器，还用于对产生所述输出结果的数据进行标记，用于更新所述模型。
在一种可能的设计中，所述通信接口，还用于接收分析和建模功能AMF实体发送的数据请求，所述数据请求用于请求更新所述模型所需的第二数据；
所述处理器，还用于根据所述数据请求确定所述第二数据，通过所述通信接口将所述第二数据发送给所述AMF实体，所述第二数据中包括所述输出结果、产生所述输出结果的数据以及所述输出结果对应的所述网元的参数调整动作。
第十方面，本申请实施例提供了一种系统，包括：分析和建模功能AMF实体，智能协同功能APF实体；所述AMF实体，用于向所述APF实体发送模型更新策略，所述模型更新策略包括网元的模型的更新条件，所述模型用于指导对所述网元的参数调整；所述APF实体，用于接收所述AMF实体发送的所述模型更新策略，在确定所述模型满足所述模型更新策略包括的更新条件时，向所述AMF实体发送更新请求，所述更新请求用于触发所 述AMF实体执行更新所述模型的流程；所述AMF实体，用于在接收到所述更新请求时，执行更新所述模型的流程。
具体的，AMF实体用于实现第一方面中任一设计所述的第一功能实体所执行的方法流程。APF实体，用于实现第二方面任一设计所述的APF实体执行的方法流程。系统还可以包括DSP实体，用于执行第三方面任一设计所述的DSF实体执行的方法流程。在一个可能的设计中，该系统还可以包括本申请实施例提供的方案中与该两种功能实体进行交互的其他设备，例如，网元、MEF实体。
第十一方面，本申请实施例提供了另一种系统，包括：分析和建模功能AMF实体以及模型执行功能MEF实体；所述AMF实体，用于向所述MEF实体发送模型更新策略，所述模型更新策略包括网元的模型的更新条件，所述模型用于指导对所述网元的参数调整；所述MEF实体，用于接收所述AMF实体发送的所述模型更新策略，在确定所述模型满足所述模型更新策略包括的更新条件时，向所述AMF实体发送更新请求，所述更新请求用于触发所述AMF实体执行更新所述模型的流程；所述AMF实体，用于在接收到所述MEF实体发送的更新请求时，执行更新所述模型的流程。
具体的，AMF实体用于实现第一方面中任一设计所述的第一功能实体所执行的方法流程。MEF实体，用于实现第二方面任一设计所述的APF实体执行的方法流程。系统还可以包括DSP实体，用于执行第三方面任一设计所述的DSF实体执行的方法流程。在一个可能的设计中，该系统还可以包括本申请实施例提供的方案中与该两种功能实体进行交互的其他设备，例如，网元、APF实体。
第十二方面，本申请实施例还提供一种可读存储介质，所述可读存储介质中存储有程序或指令，当其在计算机上运行时，使得上述各方面的任意网元的选择方法被执行。
第十三方面，本申请实施例还提供一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行上述各方面中的任意网元的选择方法。
另外，第四方面至第十三方面中任一种设计方式所带来的技术效果可参见第一方面至第三方面中不同实现方式所带来的技术效果，此处不再赘述。
附图说明
图1为本申请实施例提供的一种核心网架构示意图；
图2为本申请实施例提供的一种接入网架构示意图；
图3为本申请实施例提供的一种网络架构示意图；
图4为本申请实施例提供的一种模型更新方法流程图；
图5为本申请实施例提供的APF实体触发对应的模型更新方法流程示意图；
图6为本申请实施例提供的全局触发的模型更新方法流程示意图；
图7为本申请实施例提供的MEF实体触发的模型更新方法流程示意图；
图8为本申请实施例提供的装置800结构示意图；
图9为本申请实施例提供的装置900结构示意图；
图10为本申请实施例提供的装置1000结构示意图；
图11为本申请实施例提供的装置1100结构示意图；
图12为本申请实施例提供的装置1200结构示意图；
图13为本申请实施例提供的装置1300结构示意图；
图14为本申请实施例提供的装置1400结构示意图；
图15为本申请实施例提供的装置1500结构示意图。
具体实施方式
下面将结合附图，对本申请实施例进行详细描述。
本申请实施例提供的模型更新方法适用于不同的无线接入技术的通信系统中，例如第三代(3rd Generation，3G)通信系统、长期演进(long term evolution，LTE)系统、第五代(5th generation，5G)通信系统，或者未来更多可能的通信系统中。
接入网负责终端的无线侧接入，接入网(access network，AN)设备可能的部署形态包括：集中式单元(centralized unit，CU)和分布式单元(distributed unit，DU)分离场景；以及单站点的场景。单站点包括gNB/NR-NB、传输接收点(transmission reception point，TRP)、演进型节点B(evolved Node B，eNB)、无线网络控制器(radio network controller，RNC)、节点B(Node B，NB)、基站控制器(base station controller，BSC)、基站收发台(base transceiver station,BTS)、家庭基站(例如，home evolved NodeB，或home Node B，HNB)、基带单元(base band unit，BBU)，或无线保真(wireless fidelity，Wifi)接入点(access point，AP)等。在5G通信系统中，单站点为gNB/NR-NB。其中，CU支持无线资源控制(radio resource control，RRC)、分组数据汇聚协议(packet data convergence protocol，PDCP)、业务数据适配协议(service data adaptation protocol，SDAP)等协议。CU一般会部署在中心节点，具有较为丰富的计算资源。DU主要支持无线链路控制层(radio link control，RLC)、媒体接入控制层(media access control，MAC)和物理层(PHY)协议。DU一般采用分布式部署方式，在通常情况下一个CU要连接一个以上的DU。gNB具有CU和DU的功能，并且通常作为单站点的形态部署。DU和gNB受限于设备的体积、功耗等因素，通常计算资源较为有限。
接入网的运营支持系统(operation support system，OSS)主要用于配置终端设备的参数、收集终端设备的告警、性能统计、运行状态和日志等信息数据。终端设备，又称之为用户设备(user equipment，UE)、移动台(mobile station，MS)、移动终端(mobile terminal，MT)等，是一种向用户提供语音和/或数据连通性的设备。例如，终端设备包括具有无线连接功能的手持式设备、车载设备等。目前，终端设备可以是：手机(mobile phone)、平板电脑、笔记本电脑、掌上电脑、移动互联网设备(mobile internet device，MID)、可穿戴设备，虚拟现实(virtual reality，VR)设备、增强现实(augmented reality，AR)设备、工业控制(industrial control)中的无线终端、无人驾驶(self driving)中的无线终端、远程手术(remote medical surgery)中的无线终端、智能电网(smart grid)中的无线终端、运输安全(transportation safety)中的无线终端、智慧城市(smart city)中的无线终端，或智慧家庭(smart home)中的无线终端等。
移动核心网络包含有接入和移动性管理功能、会话管理(session management function，SMF)、用户面功能(user plane function，UPF)和策略控制功能(policy control function，PCF)、应用功能等实体，通常会以集中化部署在云计算系统上，并通过传输网连接接入网。
本申请实施例提供的模型更新方式可以通过配置在无线网络中的数据分析(data analysis，DA)网络架构来实现。DA网络架构可以配置于核心网中，或者配置于无线接入网中。作为一种示例，当DA网络架构配置于核心网中，可以称为NWDA(network data  analysis)，当DA网络架构配置于无线接入网中，可以称为RANDA(radio access network data analysis)，当DA部署于CU中时，可以称为CUDA，当DA部署于DU中时，可以称为DUDA。当DA部署于gNB中时，可以称为gNBDA。
参见图1所示，在核心网中引入了NWDA，NWDA用来从各个核心网的功能实体中收集网络数据、训练数据分析评估模型(简称模型)、实现数据分析预测以及为PCF提供数据分析后的结果，让PCF可以基于数据分析结果和用户当前的业务进行决策，生成新的策略与计费控制规则(policy and charging control rule，PCC)规则、改善用户业务的服务质量(quality of service，QoS)，从而改善用户业务的使用体验。
DA中可以包括四个功能实体，分别为：数据服务功能(data service function，DSF)实体、分析和建模功能(analyzing and modeling function，AMF)实体、模型执行功能(model execution function，MEF)实体、和自适应策略功能(adaptive policy function，APF)实体。参见图2所示，为DA部署于接入网中的架构示意图。
本申请实施例中，通过上述四个功能实体之间进行信令交互，实现模型更新，根据更新后的模型确定网元参数调整动作，并作发送给执行动作的网元(net element，NE)，并根据网元执行网元参数调整动作后网络环境状态的改变，来评估模型是否需要更新，利用更新后的评估模型确定下一次的网元参数调整动作。
以下介绍在实现模型更新时上述四个功能实体所执行的功能。
DSF，用于采集数据，并对采集的数据进行预处理，向AMF提供训练或更新评估模型所需的数据，并向MEF提供执行评估模型时所需的网络数据。本申请实施例的描述中，网络数据也可以简称为数据。
AMF，用于从DSF订阅训练或更新评估模型所需的数据，根据订阅的数据训练或更新评估模型，将评估模型发送给MEF。以及，用于在接收到APF反馈的网元参数调整动作后，根据网元参数调整动作进行迭代更新评估模型，并将更新的评估模型发送给MEF或APF。MEF，用于从AMF获取评估模型，并从DSF获取网络数据，采用评估模型对网络数据进行在线预测，获得网元参数调整动作，将网元参数调整动作发送给APF。
APF，用于根据分析或预测的结果来触发策略(如冲突处理策略)，以改变网络状态，如调参、流量工程、资源调度等。具体用于从MEF获取网元参数调整动作，将网元参数调整动作发送给实际执行网元参数调整的网元，从而改善网元的容量或者性能。以及，用于将网元参数调整动作反馈给AMF。
上述四个功能实体可以部署在接入网或者部署在核心网的网元上，例如，可以部署在接入网的CU、DU、gNB和OSS上，或者部署在UPF上或者作为一个整体部署与核心网中且与PCF连接。上述四个功能实体还可以部署于接入网中用于管理整个接入网，再在每个局部网元中分别部署上述四个功能实体，参见图3所示，在RAN中部署DA，称为RANDA，在CU、DU以及gNB再分别部署CUDA、DUDA以及gNBDA。具体的，上述四个功能实体在部署时，可以部署在同一个网元中，也可以分散的部署在不同的网元中。或者说，在一些应用场景下，同一个网元中的四个功能实体进行信令交互来完成本申请实施例的模型更新方法；在另一些应用场景下，部署在不同网元的功能实体通过网元之间的接口进行信令交互来完成本申请实施例的模型更新方法。例如，对于一些实时性要求高的参数，DU的计算资源受限，可以通过CU中的AMF实体来训练或更新评估模型，通过DU中的MEF实体来执行训练或更新后的评估模型。再比如，对于整个接入网中相同模型需要更新时， 可以通过作为全局部署的RAN中的DA(具体可以是RANDA中的AMF)来训练或者更新评估模型，通过作为局部部署的DU、CU、gNB中的DA(具体可以是MEF)来执行训练或者更新后的评估模型。
需要说明的是，上述四个功能实体的名称本申请中不作限定，本领域技术人员将上述功能实体的名称更换为其它名称而执行相同的功能，均属于本申请保护的范围。
本申请实施例涉及的网元参数可以是指无线资源管理(radio resource management，RRM)中的各种参数，或无线传输技术(radio transmission technology，RTT)中的各种参数，或运维系统中的各种参数。例如网元参数可以是：导频功率、参考信号(reference signal，RS)功率、天线下倾角、长期演进(long term evolution，LTE)可复用电平差门限、测量报告(measurement report，MR)干扰判决门限等。
在现有智能网络架构中，模型下发至运行网元后，一般采用主动更新模型的方式。比如固定时间周期更新，在配置模型时设置模型更新时间(如1周、1天、1小时等)，在模型部署完成后，开始记录运行时间，达到更新时间周期后，使用这段时间内产生的数据训练新的模型，部署训练后的模型。这种主动更新模型的方式不能准确感知网络环境的变化。在模型部署运行过程中，网络可能在更新时间点到来之前就发生了改变，致使模型不再适用当时的情况，产生错误的结果，导致网络性能下降或者不能运行。
另一方面，若模型部署运行后，在更新时间周期内，网络环境没有变化，模型运行效果良好，主动训练新的模型来更新会占用DA、NodeB、CU、DU、UPF等额外的计算能力，在智能网络架构中承载多种模型更新任务时，带来额外的消耗。在智能网络架构中(如RANDA)，模型可以部署在靠近应用网元的位置，每一个周期都需要对所有的分布式网元进行更新，大量的模型传输会浪费网络的传输资源。总之，主动更新模型的方式不能灵活的配置和感知网络环境的变化，会占用多余的计算资源和带宽。
基于此，本申请提供一种模型更新方法及装置，用以解决现有更新模型的灵活性较低的问题。其中，方法和装置是基于同一发明构思的，由于方法及装置解决问题的原理相似，因此装置与方法的实施可以相互参见，重复之处不再赘述。
参见图4所示，为本申请实施例提供的模型更新方法流程示意图。
S401，第一功能实体向第二功能实体发送模型更新策略。
其中，所述模型更新策略包括第一网元的模型的更新条件，所述模型用于指导对第一网元的参数调整。
其中，不同的网络环境状态下模型指导对第一网元的参数调整不同。
S402，第二功能实体接收第一功能实体发送的模型更新策略。
S403，所述第二功能实体根据所述模型更新策略确定所述模型是否满足所述更新条件。
S404，所述第二功能实体在确定所述模型满足所述更新条件时，向所述第一功能实体发送更新请求，所述更新请求用于触发所述第一功能实体执行更新所述模型的流程。
S405，所述第一功能实体接收所述第二功能实体发送的更新请求时，执行更新所述第一网元的模型的流程。
在一种可能的实施方式中，第一功能实体可以是配置于CU中CUDA，第二功能实体可以是配置于DU中的DUDA，或者第一功能实体可以是配置于RAN中的RANDA，第二功能实体可以是配置于CU/DU/gNB中的DA，或者第一功能实体可以是配置于OSS中的OSSDA，第二功能实体可以是配置于CU/DU/gNB中的DA。
在一种可能的实施方式中，第一功能实体可以是AMF实体，第二功能实体可以是APF实体，即由APF实体触发更新。其中，AMF实体和APF实体可以位于同一网元中，也可以位于不同网元中。比如AMF实体和APF实体均在CUDA上。再比如AMF实体位于CUDA上，APF实体位于DUDA上。再比如AMF实体位于RANDA上，APF实体位于gNBDA上。
具体的，AMF实体可以向APF实体配置模型更新策略，由APF实体基于模型更新策略向DSF实体获取监测数据(或者称为第一数据等等，本申请对此不作限定)，监测数据用来确定模型是否满足模型更新策略中的更新条件。从而APF实体根据模型更新策略以及监测数据来判断是否需要更新模型，在确定需要更新模型时指示AMF实体来更新模型。
在另一种可能的实施方式中，第一功能实体可以是DA中的AMF实体，第二功能实体可以是MEF实体，即由MEF实体触发更新。其中，AMF实体和MEF实体可以位于同一网元中，也可以位于不同网元中。比如AMF实体和MEF实体均在CUDA上。再比如AMF实体位于CUDA上，MEF实体位于DUDA上。再比如AMF实体位于RANDA上，APFMEF实体位于gNBDA上。
具体的，AMF实体向MEF实体配置模型更新策略，由MEF实体基于模型更新策略向DSF实体获取监测数据。从而MEF实体根据模型更新策略以及监测数据来判断是否需要更新模型，在确定需要更新模型时指示AMF实体来更新模型。
参见图5所示，以由APF实体触发模型更新为例，对模型更新方法进行详细说明。
图5所示的AMF实体、APF实体、MEF实体以及DSF实体可以位于同一网元中，也可以位于不同网元中。比如AMF实体、APF实体、MEF实体以及DSF实体均在CUDA上。再比如AMF实体位于CUDA上，APF实体、MEF实体以及DSF实体位于DUDA上。再比如AMF实体位于RANDA上，APF实体、MEF实体以及DSF实体位于gNBDA上。
S501，AMF实体向MEF实体发送模型安装信息，其中该模型安装信息中包括第一网元的模型。模型安装信息还可以包括MEF在运行该模型时需要向DSF实体订阅的数据的索引，为了后续描述方便将该运行模型需要的数据称为运行数据。
S502，AMF实体向APF实体发送模型更新策略。
模型更新策略中包括了模型的更新条件。模型更新策略中还可以包括监测数据的索引，比如模型需要调整NE中的目标关键性能指标(key performance indicator，KPI)，KPI可以是电路交换(circuit switching，CS)掉话率；或者，流量或负载的计算方式，如流量或负载为多个小区的加权和。
更新条件可以是根据实际的应用场景来配置，以满足不同的应用场景中的需求。比如KPI的阈值，在KPI低于阈值时触发模型的更新。比如流量预测场景可以为预测误差的阈值，预测误差可以为实际流量与预测流量之间的差值，再比如自适应调制编码(adaptive modulation and coding，AMC)场景可以为吞吐率阈值。在流量预测场景，需要监测实际流量，因此，监测数据可以为流量，或者用于计算流程的数据，在AMC场景，需要监测吞吐率，监测数据可以为吞吐量，或者用于计算吞吐量的数据。
本申请实施例中，S501与S502可以在一个消息中发送，也可以在不同消息中发送。在不同消息中发送时，对S501与S502的执行先后顺序，不作具体限定。
S503a，MEF实体向DSF实体发送数据请求1，该数据请求1用于请求运行模型所需要的数据。具体的根据步骤S501中的运行数据的索引向DSF实体获取运行数据。即数据 请求1中包括运行数据的索引。
S503b，DSF实体根据数据请求1向MEF发送运行数据。具体的，DSF实体根据运行数据的索引，周期性的向MEF发送运行数据。每周期发送的运行数据为反映在该周期内的网络环境状态的数据。
S504a，APF实体向DSF实体发送数据请求2，该数据请求2用于请求监测数据。具体的根据步骤S502中的监测数据的索引向DSF实体获取监测数据。即数据请求2中包括监测数据的索引。
S504b，DSF实体根据数据请求2向APF实体发送监测数据。具体的，DSF实体根据监测数据的索引，周期性的向MEF实体发送监测数据。
本申请实施例中，不限定S503a与S504a之间的先后执行顺序。
S505，MEF实体将S503b获取到的运行数据输入模型产生输出结果，并将输出结果发送给APF实体。具体的，MEF实体每周期获取到运行数据后，将获取到的运行数据输入模型产生输出结果，并将输出结果发送给APF实体。
S506，APF实体在接收到输出结果后，根据输出结果确定第一网元的参数调整动作，并将第一网元的参数调整动作发送给第一网元。
示例性的，APF实体中存储有输出结果与参数调整动作之间的对应关系，从而根据对应关系确定接收到的输出结果对应的参数调整动作。
可选地，上述对应关系也可以存储于MEF实体中，从而MEF实体获取到模型的输出结果后，根据输出结果确定第一网元的参数调整动作，并将第一网元的参数调整动作发送给APF实体，从而APF实体将第一网元的参数调整动作转发给第一网元。另外，模型也可以直接输出参数调整动作，即输出结果为参数调整动作。
S507，APF实体将模型的输出结果、输出结果对应的参数调整动作发送给DSF实体。
示例性的，具体发送内容可以包括：模型的标识、模型的输出结果，产生该输出结果的数据的索引，该输出结果对应的参数调整动作。发送内容中也可以不包括产生该输出结果的数据的索引，由于DSF实体周期性向MEF实体发送运行数据，进而MEF实体也周期性将运行数据输入模型产生输出结果，并将输出结果周期性的发送给APF实体，APF实体也周期性的发送给DSF实体，从而DSF实体能够确定每次接收到的APF实体的发送内容是基于哪个周期发送的数据产生的。
S508，DSF实体接收到模型的输出结果、参数调整动作后，对产生该输出结果的数据进行标记，从而用于模型的更新。
S509，在S504b后，APF实体接收到DSF实体发送的监测数据，根据在S502接收到的模型更新策略，确定模型性能是否下降，即根据监测数据确定该模型是否满足模型更新策略包括的更新条件，若是满足，则向AMF实体发送用于触发进行所述模型更新的更新请求。比如在小区流量预测的应用场景中，需要对小区未来的各个时间段(每小时)的流量进行预测流量进行，则模型的输出是下一小时的流量，即预测流量，而模型更新策略可以是预测流量与真实流量的误差阈值，在下一小时获取真实流量之后，计算预测流量与真实流量之间的预测误差，并对预测误差与预测误差的阈值进行比较，若大于预测误差的阈值，则触发模型更新流程，发送模型更新请求，反之继续运行。
S510，AMF实体在接收到APF实体发送的更新请求后，向DSF实体发送数据请求3，该数据请求3用于请求进行模型更新所需的第一网元的数据，后续为了描述方便，将该数 据称为更新数据。该数据请求3中可以包括更新数据的索引。
S511，DSF实体根据数据请求3将更新数据发送给AMF实体。更新数据中包括了在S508时接收到的所述MEF实体确定的模型的输出结果，以及所述APF实体根据所述输出结果确定的该输出结果对应的参数调整动作。
S512，AMF实体根据更新数据重新训练模型，并将训练后的模型发送给MEF实体。
可选地，AMF实体可以更新模型更新策略，并将更新后的模型更新策略发送给APF实体。
通过本申请的模型更新机制，将模型更新机制从主动周期更新替换为被动触发更新，使得网络能够感知模型运行的效果和网络状态，模型更新策略可以根据场景配置，从而灵活性较高。由APF从DSF中订阅数据来对网络环境状态进行判断，若模型性能没有下降，则可以不用进行更新，这样节省了训练模型的计算资源和模型传输过程；若网络性能下降，由APF发送更新请求触发AMF重新训练模型，达到更新模型的目的。
上述图5对应的实施例中，由APF实体触发AMF实体仅针对性能下降的模型进行更新。
在一种可能的实施方式中，在系统中包括有上述第一网元在内的多个网元，所述多个网元对应的模型相同，且所述多个网元由至少两个APF实体管理，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中。在该实施方式中，AMF实体可以位于RANDA中，APF实体、MEF实体以及DSF实体位于gNBDA中，或者AMF实体位于CUDA中，APF实体、MEF实体以及DSF实体位于DUDA中。在多个APF实体均向AMF实体触发模型更新时，则AMF可以针对系统的上述多个网元的模型均触发更新。
具体的，在AMF实体在分别向每个APF实体发送所述模型更新策略后，所述AMF实体确定在预设时长内接收到发送所述更新请求的APF实体的数量达到预设阈值时，所述AMF实体分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的数据；所述AMF实体接收所述至少两个DSF实体发来的所述多个网元分别对应的数据；所述AMF实体根据所述多个网元分别对应的数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF。
参见图6所示，对AMF实体针对系统中的多个网元的模型均触发更新进行详细描述。用户在AMF实体中配置全局更新策略。在满足全局更新策略包括的条件时，触发全局更新。条件可以是局部模型对应的更新请求数量达到预设阈值。以系统中包括的4个网元的模型相同为例，4个网元分别为网元1～网元4。该4个网元的模型配置于4个MEF实体中，4个网元由4个APF实体管理，N个网元的相关数据由4个DSF实体管理。
AMF实体针对每个网元对应的MEF实体均执行S501以及S502所述的方法步骤，此处不再赘述。
N个MEF实体、N个APF实体以及N个DSF实体均按照S503a至S509所述的方法步骤执行，此处不再赘述。
在步骤S509后，AMF实体通过如下方式重新训练模型：
S601，AMF实体统计在预设时长内接收到发送所述更新请求的APF实体的数量。
S602，在确定统计的数量达到预设阈值时，所述AMF实体分别向所述4个DSF实体发送数据请求4。所述数据请求4用于请求进行模型更新所需的所述4个网元分别对应的 数据。
S603，所述AMF实体接收4个DSF实体发来的所述4个网元分别对应的更新数据。
每个网元均有在当前网络环境状态下对应的更新数据。
所述多个网元中每个网元对应的更新数据包括该网元的模型的输出结果，以及所述输出结果对应的参数调整动作。
S604，所述AMF实体根据所述4个网元分别对应的更新数据重新训练所述模型。
S605，所述AMF实体将重新训练的所述模型分别发送给4个MEF实体。
参见图7所示，以由MEF实体触发模型更新为例，对模型更新方法进行详细说明。
图7所示的AMF实体、APF实体、MEF实体以及DSF实体可以位于同一网元中，也可以位于不同网元中。比如AMF实体、APF实体、MEF实体以及DSF实体均在CUDA上。再比如AMF实体位于CUDA上，APF实体、MEF实体以及DSF实体位于DUDA上。再比如AMF实体位于RANDA上，APF实体、MEF实体以及DSF实体位于gNBDA上。
S701，AMF实体向MEF实体发送模型安装信息，其中该模型安装信息中包括第一网元的模型。具体参见S501的相关描述，此处不再赘述。
S702，AMF实体向MEF实体发送模型更新策略。针对模型更新策略的描述具体可以参见上述任一实施例，此处不再赘述。
其中，AMF实体可以将模型安装信息以及模型更新策略通过一条消息发送给MEF实体，还可以通过两条消息分别将模型安装信息以及模型更新策略发送给MEF实体，本申请对此不作具体限定。
S703，MEF实体向DSF实体发送数据请求1，该数据请求1用于请求获取运行数据以及监测数据。针对运行数据以及监测数据相关说明，可以参见任一实施例所述，此处不再赘述。
可选地，MEF实体也可以通过两条消息分别请求获取运行数据和监测数据，本申请对此不作具体限定。
S704，DSF实体将运行数据以及监测数据发送给MEF实体。具体的，DSF实体周期性的将运行数据以及监测数据发送给MEF实体。
S705，MEF实体将接收到运行数据输入模型产生输出结果，并将输出结果发送给APF实体。具体的，MEF实体每周期获取到运行数据后，将获取到的运行数据输入模型产生输出结果，并将输出结果发送给APF实体。此处输出结果即为参数调整动作。
S706，APF实体在接收到输出结果后，将参数调整动作发送给第一网元。
S707，MEF实体将输出结果发送给DSF实体。示例性的，具体发送内容可以包括：模型的标识、模型的输出结果，产生该输出结果的数据的索引。发送内容中也可以不包括产生该输出结果的数据的索引，由于DSF实体周期性向MEF实体发送运行数据，进而MEF实体也周期性将运行数据输入模型产生输出结果，并将输出结果周期性的发送给APF实体，APF实体也周期性的发送给DSF实体，从而DSF实体能够确定每次接收到的APF实体的发送内容是基于哪个周期发送的数据产生的。
S708，DSF实体接收到模型的输出结果后，对产生该输出结果的数据进行标记，从而用于模型的更新。
S709，MEF实体接收到DSF实体发送的监测数据，根据模型更新策略，确定模型性 能是否下降，即根据监测数据确定该模型是否满足模型更新策略包括的更新条件，若是满足，则向AMF实体发送用于触发进行所述模型更新的更新请求。
S710，AMF实体在接收到MEF实体发送的更新请求后，向DSF实体发送数据请求2，该数据请求2用于请求更新数据。该数据请求2中可以包括更新数据的索引。
S711，DSF实体根据数据请求2将更新数据发送给AMF实体。更新数据中包括了模型的输出结果。
S712，AMF实体根据更新数据重新训练模型，并将训练后的模型发送给MEF实体。
可选地，AMF实体可以更新模型更新策略，并将更新后的模型更新策略发送给MEF实体。
通过本申请的模型更新机制，将模型更新机制从主动周期更新替换为被动触发更新，使得网络能够感知模型运行的效果和网络状态，模型更新策略可以根据场景配置，从而灵活性较高。由MEF从DSF中订阅数据来对网络环境状态进行判断，若模型性能没有下降，则可以不用进行更新，这样节省了训练模型的计算资源和模型传输过程；若网络性能下降，由MEF发送更新请求触发AMF重新训练模型，达到更新模型的目的。
在一种可能的实施方式中，在系统中包括有上述第一网元在内的多个网元，所述多个网元对应的模型相同，且所述多个网元由至少两个APF实体管理，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中。在该实施方式中，AMF实体可以位于RANDA中，APF实体、MEF实体以及DSF实体位于gNBDA中，或者AMF实体位于CUDA中，APF实体、MEF实体以及DSF实体位于DUDA中。在多个APF实体均向AMF实体触发模型更新时，则AMF可以针对系统的上述多个网元的模型均触发更新。具体的，在AMF实体在分别向每个MEF实体发送所述模型更新策略后，所述AMF实体确定在预设时长内接收到发送所述更新请求的MEF实体的数量达到预设阈值时，所述AMF实体分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的数据；所述AMF实体接收所述至少两个DSF实体发来的所述多个网元分别对应的数据；所述AMF实体根据所述多个网元分别对应的数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF实体。
基于与方法实施例同样的发明构思，本申请实施例还提供了一种装置，参见图8所示，该装置800应用于AMF实体。该装置800具体可以是AMF实体中的处理器，或者芯片或者芯片系统，或者是一个功能模块等。该装置可以包括发送模块801、接收模块802、处理模块803。处理模块803用于对装置800的动作进行控制管理。其中发送模块801用于执行本申请任一实施例中第一功能实体或者AMF实体所执行的发送动作，处理模块803用于本申请任一实施例中第一功能实体以及AMF实体所执行的处理动作，比如训练模型等，接收模块802用于执行本申请任一实施例中第一功能实体以及AMF实体所执行的接收动作，重复之处，此处不再赘述。处理模块803还可以用于指示上述任意实施例中涉及第一功能实体或者AMF实体的处理过程和/或本申请所描述的技术方案的其他过程。
本申请实施例还提供另外一种AMF的结构，如图9所示，AMF实体9100中可以包括通信接口910、处理器920。可选的，AMF实体910中还可以包括存储器930。其中，存储器930可以设置于AMF实体内部，还可以设置于AMF实体外部。上述图8中所示的处理模块803均可以由处理器920实现。发送模块801、接收模块802可以由通信接口910 实现。处理器920通过通信接口910接收信息或者消息，并用于实现图4～图7中所述的AMF实体或者第一功能实体所执行的方法。在实现过程中，处理流程的各步骤可以通过处理器920中的硬件的集成逻辑电路或者软件形式的指令完成图4～图7中所述的AMF实体所执行的方法。
本申请实施例中通信接口910可以是电路、总线、收发器或者其它任意可以用于进行信息交互的装置。其中，示例性地，该其它装置可以是与该装置900相连的设备，比如，该其它装置可以是DSF实体或者MEF实体或者APF实体等。
本申请实施例中处理器920可以是通用处理器、数字信号处理器、专用集成电路、现场可编程门阵列或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件，可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件单元组合执行完成。处理器920用于实现上述方法所执行的程序代码可以存储在存储器930中。存储器930和处理器920耦合。
本申请实施例中的耦合是装置、单元或模块之间的间接耦合或通信连接，可以是电性，机械或其它的形式，用于装置、单元或模块之间的信息交互。
处理器920可能和存储器930协同操作。存储器930可以是非易失性存储器，比如硬盘(hard disk drive，HDD)或固态硬盘(solid-state drive，SSD)等，还可以是易失性存储器(volatile memory)，例如随机存取存储器(random-access memory，RAM)。存储器930是能够用于携带或存储具有指令或数据结构形式的期望的程序代码并能够由计算机存取的任何其他介质，但不限于此。
本申请实施例中不限定上述通信接口910、处理器920以及存储器930之间的具体连接介质。本申请实施例在图9中以存储器930、处理器920以及通信接口910之间通过总线连接，总线在图9中以粗线表示，其它部件之间的连接方式，仅是进行示意性说明，并不引以为限。所述总线可以分为地址总线、数据总线、控制总线等。为便于表示，图9中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
基于与方法实施例同样的发明构思，本申请实施例还提供了一种装置，参见图10所示，该装置1000应用于MEF实体。该装置1000具体可以是MEF实体中的处理器，或者芯片或者芯片系统，或者是一个功能模块等。该装置可以包括接收模块1001、处理模块1002、发送模块1003。处理模块1002用于对装置1000的动作进行控制管理。其中发送模块1003用于执行本申请任一实施例中第二功能实体或者MEF实体所执行的发送动作，处理模块1002用于本申请任一实施例中第二功能实体以及MEF实体所执行的处理动作，比如将数据输入模型得到输出结果等，接收模块1001用于执行本申请任一实施例中第二功能实体以及MEF实体所执行的接收动作，重复之处，此处不再赘述。处理模块1002还可以用于指示上述任意实施例中涉及第二功能实体或者MEF实体的处理过程和/或本申请所描述的技术方案的其他过程。
本申请实施例还提供另外一种MEF实体的结构，如图11所示，MEF实体1100中可以包括通信接口1110、处理器1120。可选的，MEF实体1100中还可以包括存储器1130。其中，存储器1130可以设置于MEF实体内部，还可以设置于MEF实体外部。上述图10中所示的处理模块1002均可以由处理器1120实现。发送模块1003、接收模块1001可以 由通信接口1110。处理器1120通过通信接口1110接收信息或者消息，并用于实现图4～图7中所述的MEF实体或者第二功能实体所执行的方法。在实现过程中，处理流程的各步骤可以通过处理器1120中的硬件的集成逻辑电路或者软件形式的指令完成图4～图7中所述的MEF实体所执行的方法。
本申请实施例中通信接口1110可以是电路、总线、收发器或者其它任意可以用于进行信息交互的装置。其中，示例性地，该其它装置可以是与该装置1100相连的设备，比如，该其它装置可以是DSF实体或者AMF实体或者APF实体等。
本申请实施例中处理器1120可以是通用处理器、数字信号处理器、专用集成电路、现场可编程门阵列或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件，可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件单元组合执行完成。处理器1120用于实现上述方法所执行的程序代码可以存储在存储器1130中。存储器1130和处理器1120耦合。
本申请实施例中的耦合是装置、单元或模块之间的间接耦合或通信连接，可以是电性，机械或其它的形式，用于装置、单元或模块之间的信息交互。
处理器1120可能和存储器1130协同操作。存储器1130可以是非易失性存储器，比如硬盘(hard disk drive，HDD)或固态硬盘(solid-state drive，SSD)等，还可以是易失性存储器(volatile memory)，例如随机存取存储器(random-access memory，RAM)。存储器1130是能够用于携带或存储具有指令或数据结构形式的期望的程序代码并能够由计算机存取的任何其他介质，但不限于此。
本申请实施例中不限定上述通信接口1110、处理器1120以及存储器1130之间的具体连接介质。本申请实施例在图11中以存储器1130、处理器1120以及通信接口1110之间通过总线连接，总线在图11中以粗线表示，其它部件之间的连接方式，仅是进行示意性说明，并不引以为限。所述总线可以分为地址总线、数据总线、控制总线等。为便于表示，图11中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
基于与方法实施例同样的发明构思，本申请实施例还提供了一种装置，参见图12所示，该装置1200应用于APF实体。该装置1200具体可以是APF实体中的处理器，或者芯片或者芯片系统，或者是一个功能模块等。该装置可以包括接收模块1201、处理模块1202、发送模块1203。处理模块1202用于对装置1200的动作进行控制管理。其中发送模块1203用于执行本申请任一实施例中第二功能实体或者APF实体所执行的发送动作，处理模块1202用于本申请任一实施例中第二功能实体以及APF实体所执行的处理动作，比如将数据输入模型得到输出结果等，接收模块1201用于执行本申请任一实施例中第二功能实体以及APF实体所执行的接收动作，重复之处，此处不再赘述。处理模块1202还可以用于指示上述任意实施例中涉及第二功能实体或者APF实体的处理过程和/或本申请所描述的技术方案的其他过程。
本申请实施例还提供另外一种APF实体的结构，如图13所示，APF实体1300中可以包括通信接口1310、处理器1320。可选的，APF实体1300中还可以包括存储器1330。其中，存储器1330可以设置于APF实体内部，还可以设置于APF实体外部。上述图12中所示的处理模块1202均可以由处理器1320实现。发送模块1203、接收模块1201可以由 通信接口1310。处理器1320通过通信接口1310接收信息或者消息，并用于实现图4～图7中所述的APF实体或者第二功能实体所执行的方法。在实现过程中，处理流程的各步骤可以通过处理器1320中的硬件的集成逻辑电路或者软件形式的指令完成图4～图7中所述的APF实体所执行的方法。
本申请实施例中通信接口1310可以是电路、总线、收发器或者其它任意可以用于进行信息交互的装置。其中，示例性地，该其它装置可以是与该装置1300相连的设备，比如，该其它装置可以是DSF实体或者AMF实体或者MEF实体等。
本申请实施例中处理器1320可以是通用处理器、数字信号处理器、专用集成电路、现场可编程门阵列或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件，可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件单元组合执行完成。处理器1320用于实现上述方法所执行的程序代码可以存储在存储器1330中。存储器1330和处理器1320耦合。
本申请实施例中的耦合是装置、单元或模块之间的间接耦合或通信连接，可以是电性，机械或其它的形式，用于装置、单元或模块之间的信息交互。
处理器1320可能和存储器1330协同操作。存储器1330可以是非易失性存储器，比如硬盘(hard disk drive，HDD)或固态硬盘(solid-state drive，SSD)等，还可以是易失性存储器(volatile memory)，例如随机存取存储器(random-access memory，RAM)。存储器1330是能够用于携带或存储具有指令或数据结构形式的期望的程序代码并能够由计算机存取的任何其他介质，但不限于此。
本申请实施例中不限定上述通信接口1310、处理器1320以及存储器1330之间的具体连接介质。本申请实施例在图13中以存储器1330、处理器1320以及通信接口1310之间通过总线连接，总线在图13中以粗线表示，其它部件之间的连接方式，仅是进行示意性说明，并不引以为限。所述总线可以分为地址总线、数据总线、控制总线等。为便于表示，图13中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
基于与方法实施例同样的发明构思，本申请实施例还提供了一种装置，参见图14所示，该装置1400应用于DSF实体。该装置1400具体可以是DSF实体中的处理器，或者芯片或者芯片系统，或者是一个功能模块等。该装置可以包括接收模块1401、处理模块1402、发送模块1403。处理模块1402用于对装置1400的动作进行控制管理。其中发送模块1403用于执行本申请任一实施例中DSF实体所执行的发送动作，处理模块1402用于本申请任一实施例中DSF实体所执行的处理动作，比如将数据输入模型得到输出结果等，接收模块1401用于执行本申请任一实施例中DSF实体所执行的接收动作，重复之处，此处不再赘述。处理模块1402还可以用于指示上述任意实施例中涉及DSF实体的处理过程和/或本申请所描述的技术方案的其他过程。
本申请实施例还提供另外一种DSF实体的结构，如图15所示，DSF实体1500中可以包括通信接口1510、处理器1520。可选的，DSF实体1500中还可以包括存储器1530。其中，存储器1530可以设置于DSF实体内部，还可以设置于DSF实体外部。上述图14中所示的处理模块1402均可以由处理器1520实现。发送模块1403、接收模块1401可以由通信接口1510。处理器1520通过通信接口1510接收信息或者消息，并用于实现图4～图7 中所述的DSF实体或者第二功能实体所执行的方法。在实现过程中，处理流程的各步骤可以通过处理器1520中的硬件的集成逻辑电路或者软件形式的指令完成图4～图7中所述的DSF实体所执行的方法。
本申请实施例中通信接口1510可以是电路、总线、收发器或者其它任意可以用于进行信息交互的装置。其中，示例性地，该其它装置可以是与该装置1500相连的设备，比如，该其它装置可以是DSF实体或者AMF实体或者MEF实体等。
本申请实施例中处理器1520可以是通用处理器、数字信号处理器、专用集成电路、现场可编程门阵列或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件，可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件单元组合执行完成。处理器1520用于实现上述方法所执行的程序代码可以存储在存储器1530中。存储器1530和处理器1520耦合。
本申请实施例中的耦合是装置、单元或模块之间的间接耦合或通信连接，可以是电性，机械或其它的形式，用于装置、单元或模块之间的信息交互。
处理器1520可能和存储器1530协同操作。存储器1530可以是非易失性存储器，比如硬盘(hard disk drive，HDD)或固态硬盘(solid-state drive，SSD)等，还可以是易失性存储器(volatile memory)，例如随机存取存储器(random-access memory，RAM)。存储器1530是能够用于携带或存储具有指令或数据结构形式的期望的程序代码并能够由计算机存取的任何其他介质，但不限于此。
本申请实施例中不限定上述通信接口1510、处理器1520以及存储器1530之间的具体连接介质。本申请实施例在图15中以存储器1530、处理器1520以及通信接口1510之间通过总线连接，总线在图15中以粗线表示，其它部件之间的连接方式，仅是进行示意性说明，并不引以为限。所述总线可以分为地址总线、数据总线、控制总线等。为便于表示，图15中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
基于以上实施例，本申请实施例还提供了一种计算机存储介质，该存储介质中存储软件程序，该软件程序在被一个或多个处理器读取并执行时可实现上述任意一个或多个实施例提供的方法。所述计算机存储介质可以包括：U盘、移动硬盘、只读存储器、随机存取存储器、磁碟或者光盘等各种可以存储程序代码的介质。
基于以上实施例，本申请实施例还提供了一种芯片，该芯片包括处理器，用于实现上述任意一个或多个实施例所涉及的功能，例如获取或处理上述方法中所涉及的信息或者数据。可选地，所述芯片还包括存储器，所述存储器，用于处理器所执行必要的程序指令和数据。该芯片，可以由芯片构成，也可以包含芯片和其他分立器件。
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本申请是参照根据本申请的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/ 或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
显然，本领域的技术人员可以对本申请进行各种改动和变型而不脱离本申请的精神和范围。这样，倘若本申请的这些修改和变型属于本申请权利要求及其等同技术的范围之内，则本申请也意图包含这些改动和变型在内。

Claims (30)

  1. 一种模型更新方法，其特征在于，包括：
    第一功能实体向第二功能实体发送模型更新策略，所述模型更新策略包括第一网元的模型的更新条件，所述第一网元的模型用于指导对所述第一网元的参数调整；
    所述第一功能实体在接收到更新请求时，执行更新所述第一网元的模型的流程；
    其中，所述更新请求是所述第二功能实体在确定所述模型满足所述更新条件时触发的。
  2. 如权利要求1所述的方法，其特征在于，所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件。
  3. 如权利要求1或2所述的方法，其特征在于，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；
    所述第一功能实体执行更新所述第一网元的模型的流程之前，所述方法还包括：
    所述AMF实体向模型执行功能MEF实体发送所述第一网元的模型；
    所述第一功能实体执行更新所述模型的流程，包括：
    所述AMF实体向数据服务功能DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述第一网元对应的第二数据；
    所述AMF实体接收所述DSF实体发送的所述第一网元对应的第二数据，所述第二数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，以及所述APF实体根据所述输出结果确定的所述第一网元的参数调整动作；
    所述AMF实体根据所述第二数据重新训练模型，并将重新训练的模型发送给所述MEF。
  4. 如权利要求1或2所述的方法，其特征在于，所述方法应用的系统中包括所述第一网元在内的多个网元，所述多个网元均对应所述第一网元的模型，且所述多个网元由至少两个APF实体管理，所述第一功能实体为AMF实体，所述第二功能实体为所述至少两个APF实体中的任一个，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中；
    所述AMF实体在分别向每个APF实体发送所述模型更新策略后，所述第一功能实体执行更新所述模型的流程时，包括：
    所述AMF实体确定在预设时长内接收到发送所述更新请求的APF实体的数量达到预设阈值；
    所述AMF实体分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的第三数据；
    所述AMF实体接收所述至少两个DSF实体发来的所述多个网元分别对应的第三数据；
    所述AMF实体根据所述多个网元分别对应的第三数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF。
  5. 如权利要求1或2所述的方法，其特征在于，所述第一功能实体为AMF实体，第二功能实体为MEF实体；
    所述方法还包括：
    所述AMF实体向所述MEF实体发送所述第一网元的模型；
    所述第一功能实体执行更新所述第一网元的模型的流程，包括：
    所述AMF实体向DSF实体发送数据请求，所述数据请求用于请求进行所述第一网元的模型更新所需的所述第一网元对应的第四数据；
    所述AMF实体接收所述DSF实体发送的所述第四数据，所述第四数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，产生所述输出结果的数据以及所述输出结果对应的所述第一网元参数的调整动作；
    所述AMF实体根据所述四数据重新训练所述第一网元的模型，并将重新训练的模型发送给所述MEF。
  6. 一种模型更新方法，其特征在于，包括：
    第二功能实体接收第一功能实体发送的模型更新策略，所述模型更新策略包括模型的更新条件，所述模型用于指导对网元的参数调整；
    所述第二功能实体根据所述模型更新策略确定所述模型是否满足所述更新条件；
    所述第二功能实体在确定所述模型满足所述更新条件时，向所述第一功能实体发送更新请求，所述更新请求用于触发所述第一功能实体执行更新所述模型的流程。
  7. 如权利要求6所述的方法，其特征在于，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；
    所述第二功能实体在确定所述模型满足所述更新条件之前，所述方法还包括：
    所述APF实体向数据服务功能DSF实体发送数据请求，所述数据请求包括所述第一数据的索引；
    所述APF实体接收所述DSF实体根据所述第一数据的索引发送的所述第一数据；
    所述第二功能实体确定所述模型是否满足所述更新条件，包括：
    所述APF实体根据所述第一数据确定所述模型是否满足所述更新条件。
  8. 如权利要求7所述的方法，其特征在于，所述方法还包括：
    所述APF实体接收模型执行功能MEF实体发送的所述模型的输出结果；
    所述APF实体根据所述输出结果确定所述网元的参数调整动作；
    所述APF实体将确定所述网元的参数调整动作发送给所述网元，用于指示所述网元根据所述参数调整动作调整所述网元的参数。
  9. 如权利要求8所述的方法，其特征在于，所述方法还包括：
    所述APF实体将所述输出结果以及所述输出结果对应的所述网元的参数调整动作发送给所述DSF实体，以便于所述DSF实体对产生所述输出结果的数据进行标记，用于更新所述模型。
  10. 如权利要求6所述的方法，其特征在于，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为模型执行功能MEF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；
    所述第二功能实体在确定所述模型是否满足所述更新条件之前，所述方法还包括：
    所述MEF实体向数据服务功能DSF实体发送第一数据请求，所述第一数据请求包括所述第一数据的索引；
    所述MEF实体接收所述DSF实体发送的根据所述第一数据的索引确定的所述第一数据；
    所述第二功能实体确定所述模型是否满足所述更新条件，包括：
    所述MEF实体根据所述第一数据确定所述模型是否满足所述更新条件。
  11. 如权利要求10所述的方法，其特征在于，所述方法还包括：
    所述MEF接收所述AMF发送的所述模型；
    所述MEF实体向所述DSF实体发送第二数据请求，所述第二数据请求包括用于表征所述网元所在的当前网络环境状态的第二数据的索引；
    所述MEF实体接收所述DSF实体发送的根据所述第二数据的索引确定的所述第二数据；
    所述MEF实体将所述第二数据输入所述模型得到输出结果，所述结果中包括所述网元在所述当前网络环境状态下的参数调整动作；
    所述MEF实体将所述输出结果发送给所述APF实体。
  12. 一种模型更新方法，其特征在于，包括：
    数据服务功能DSF实体接收第一功能实体发送的数据请求，所述数据请求包括第一数据的索引，所述第一数据用于确定模型执行功能MEF所运行的模型是否满足更新条件，所述模型用于指导对网元的参数调整；
    所述DSF实体根据所述第一数据的索引向所述第一功能实体发送所述第一数据；
    其中，所述第一功能实体为智能协同功能APF实体或者为所述MEF实体。
  13. 如权利要求12所述的方法，其特征在于，所述第一功能实体为智能协同功能APF实体时，所述方法还包括：
    所述DSF实体接收所述APF实体发送的所述模型的输出结果以及所述输出结果对应的所述网元的参数调整动作；
    所述DSF对产生所述输出结果的数据进行标记，用于更新所述模型。
  14. 如权利要求13所述的方法，其特征在于，所述方法还包括：
    所述DSF实体接收分析和建模功能AMF实体发送的数据请求，所述数据请求用于请求更新所述模型所需的第二数据；
    所述DSF实体将更新所述模型所需的第二数据发送给所述AMF实体，所述第二数据中包括所述输出结果、产生所述输出结果的数据以及所述输出结果对应的所述网元的参数调整动作。
  15. 一种模型更新装置，其特征在于，所述装置应用于第一功能实体，包括：
    通信接口，用于向第二功能实体发送模型更新策略，所述模型更新策略包括第一网元的模型的更新条件，所述第一网元的模型用于指导对所述第一网元的参数调整；以及接收更新请求；其中，所述更新请求是所述第二功能实体在确定所述模型满足所述更新条件时触发的；
    处理器，用于在通过所述通信接口接收到所述更新请求时，执行更新所述第一网元的模型的流程。
  16. 如权利要求15所述的装置，其特征在于，所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件。
  17. 如权利要求15或16所述的装置，其特征在于，所述第一功能实体为分析和建模功能AMF实体，所述第二功能实体为智能协同功能APF实体；
    所述通信接口，还用于在所述处理器执行更新所述第一网元的模型的流程之前，向模型执行功能MEF实体发送所述第一网元的模型；
    所述处理器，在执行更新所述第一网元的模型的流程时，具体用于：
    通过所述通信接口向数据服务功能DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述第一网元对应的第二数据；
    通过所述通信接口接收所述DSF实体发送的所述第一网元对应的第二数据，所述第二数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，以及所述APF实体根据所述输出结果确定的所述第一网元的参数调整动作；
    根据所述第二数据重新训练模型，并将重新训练的模型发送给所述MEF。
  18. 如权利要求15或16所述的装置，其特征在于，所述装置应用的系统中包括所述第一网元在内的多个网元，所述多个网元均对应所述第一网元的模型，且所述多个网元由至少两个APF实体管理，所述第一功能实体为AMF实体，所述第二功能实体为所述至少两个APF实体中的任一个，所述多个网元的相关数据由至少两个DSF实体管理，所述多个网元分别对应的模型配置于至少两个MEF中；
    所述处理器，在通过所述通信接口分别向每个APF实体发送所述模型更新策略后，在执行更新所述模型的流程时，确定在预设时长内接收到发送所述更新请求的APF实体的数量达到预设阈值；
    通过所述通信接口分别向所述至少两个DSF实体发送数据请求，所述数据请求用于请求进行模型更新所需的所述多个网元分别对应的第三数据；
    通过所述通信接口接收所述至少两个DSF实体发来的所述多个网元分别对应的第三数据；
    根据所述多个网元分别对应的第三数据重新训练模型，并将重新训练的模型分别发送给所述至少两个MEF。
  19. 如权利要求15或16所述的装置，其特征在于，所述第一功能实体为AMF实体，第二功能实体为MEF实体；
    所述通信接口，还用于向所述MEF实体发送所述第一网元的模型；
    所述处理器，在执行更新所述第一网元的模型的流程时，具体用于：
    通过所述通信接口向DSF实体发送数据请求，所述数据请求用于请求进行所述第一网元的模型更新所需的所述第一网元对应的第四数据；
    通过所述通信接口接收所述DSF实体发送的所述第四数据，所述第四数据中包括所述MEF实体确定的所述第一网元的模型的输出结果，产生所述输出结果的数据以及所述输出结果对应的所述第一网元参数的调整动作；
    根据所述四数据重新训练所述第一网元的模型，并将重新训练的模型发送给所述MEF。
  20. 一种模型更新装置，其特征在于，所述装置应用于第二功能实体，包括：
    通信接口，用于接收第一功能实体发送的模型更新策略，所述模型更新策略包括模型的更新条件，所述模型用于指导对网元的参数调整；
    处理器，用于根据所述模型更新策略确定所述模型是否满足所述更新条件；在确定所述模型满足所述更新条件时，通过所述通信接口向所述第一功能实体发送更新请求，所述更新请求用于触发所述第一功能实体执行更新所述模型的流程。
  21. 如权利要求20所述的装置，其特征在于，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为智能协同功能APF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；
    所述处理器，还用于：
    在确定所述模型满足所述更新条件之前，通过所述通信接口向数据服务功能DSF实体发送数据请求，所述数据请求包括所述第一数据的索引；
    通过所述通信接口接收所述DSF实体根据所述第一数据的索引发送的所述第一数据；
    在确定所述模型是否满足所述更新条件时，具体用于根据所述第一数据确定所述模型是否满足所述更新条件。
  22. 如权利要求21所述的装置，其特征在于，所述通信接口，还用于接收模型执行功能MEF实体发送的所述模型的输出结果；
    所述处理器，还用于根据所述输出结果确定所述网元的参数调整动作；
    所述通信接口，还用于将确定所述网元的参数调整动作发送给所述网元，用于指示所述网元根据所述参数调整动作调整所述网元的参数。
  23. 如权利要求22所述的装置，其特征在于，所述通信接口，还用于将所述输出结果以及所述输出结果对应的所述网元的参数调整动作发送给所述DSF实体，以便于所述DSF实体对产生所述输出结果的数据进行标记，用于更新所述模型。
  24. 如权利要求20所述的装置，其特征在于，所述第一功能实体为分析和建模功能AMF实体，第二功能实体为模型执行功能MEF实体；所述模型更新策略中还包括第一数据的索引，所述第一数据用于确定所述模型是否满足所述更新条件；
    所述处理器，还用于：
    在确定所述模型是否满足所述更新条件之前，通过所述通信接口向数据服务功能DSF实体发送第一数据请求，所述第一数据请求包括所述第一数据的索引；
    通过所述通信接口接收所述DSF实体发送的根据所述第一数据的索引确定的所述第一数据；
    所述处理器，在确定所述模型是否满足所述更新条件时，具体用于：
    根据所述第一数据确定所述模型是否满足所述更新条件。
  25. 如权利要求24所述的装置，其特征在于，所述通信接口，还用于：
    接收所述AMF发送的所述模型；
    向所述DSF实体发送第二数据请求，所述第二数据请求包括用于表征所述网元所在的当前网络环境状态的第二数据的索引；
    接收所述DSF实体发送的根据所述第二数据的索引确定的所述第二数据；
    所述处理器，还用于将所述第二数据输入所述模型得到输出结果，所述结果中包括所述网元在所述当前网络环境状态下的参数调整动作；
    所述通信接口，还用于将所述输出结果发送给所述APF实体。
  26. 一种模型更新装置，其特征在于，所述装置应用于数据服务功能DSF实体，包括：
    通信接口，用于接收第一功能实体发送的数据请求，所述数据请求包括第一数据的索引，所述第一数据用于确定模型执行功能MEF所运行的模型是否满足更新条件，所述模型用于指导对网元的参数调整；
    处理器，用于获取所述数据请求中的第一数据的索引，并根据所述第一数据的索引确定所述第一数据；
    所述通信接口，还用于向所述第一功能实体发送所述第一数据；
    其中，所述第一功能实体为智能协同功能APF实体或者为所述MEF实体。
  27. 如权利要求26所述的装置，其特征在于，所述第一功能实体为智能协同功能APF实体时，所述通信接口，还用于接收所述APF实体发送的所述模型的输出结果以及所述输出结果对应的所述网元的参数调整动作；
    所述处理器，还用于对产生所述输出结果的数据进行标记，用于更新所述模型。
  28. 如权利要求27所述的装置，其特征在于，所述通信接口，还用于接收分析和建模功能AMF实体发送的数据请求，所述数据请求用于请求更新所述模型所需的第二数据；
    所述处理器，还用于根据所述数据请求确定所述第二数据，通过所述通信接口将所述第二数据发送给所述AMF实体，所述第二数据中包括所述输出结果、产生所述输出结果的数据以及所述输出结果对应的所述网元的参数调整动作。
  29. 一种系统，其特征在于，包括：
    分析和建模功能AMF实体，智能协同功能APF实体；
    所述AMF实体，用于向所述APF实体发送模型更新策略，所述模型更新策略包括网元的模型的更新条件，所述模型用于指导对所述网元的参数调整；
    所述APF实体，用于接收所述AMF实体发送的所述模型更新策略，在确定所述模型满足所述模型更新策略包括的更新条件时，向所述AMF实体发送更新请求，所述更新请求用于触发所述AMF实体执行更新所述模型的流程；
    所述AMF实体，用于在接收到所述更新请求时，执行更新所述模型的流程。
  30. 一种系统，其特征在于，包括：
    分析和建模功能AMF实体以及模型执行功能MEF实体；
    所述AMF实体，用于向所述MEF实体发送模型更新策略，所述模型更新策略包括网元的模型的更新条件，所述模型用于指导对所述网元的参数调整；
    所述MEF实体，用于接收所述AMF实体发送的所述模型更新策略，在确定所述模型满足所述模型更新策略包括的更新条件时，向所述AMF实体发送更新请求，所述更新请求用于触发所述AMF实体执行更新所述模型的流程；
    所述AMF实体，用于在接收到所述MEF实体发送的更新请求时，执行更新所述模型的流程。
PCT/CN2019/084148 2018-04-27 2019-04-24 一种模型更新方法、装置及系统 Ceased WO2019206196A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
JP2020560331A JP7159347B2 (ja) 2018-04-27 2019-04-24 モデル更新方法および装置、ならびにシステム
EP19791815.4A EP3780495A4 (en) 2018-04-27 2019-04-24 METHOD, DEVICE AND SYSTEM FOR MODEL UPDATE
EP25172910.9A EP4625925A3 (en) 2018-04-27 2019-04-24 Model update method and apparatus, and system
US17/081,570 US11451452B2 (en) 2018-04-27 2020-10-27 Model update method and apparatus, and system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810393730.0A CN110417565B (zh) 2018-04-27 2018-04-27 一种模型更新方法、装置及系统
CN201810393730.0 2018-04-27

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/081,570 Continuation US11451452B2 (en) 2018-04-27 2020-10-27 Model update method and apparatus, and system

Publications (1)

Publication Number Publication Date
WO2019206196A1 true WO2019206196A1 (zh) 2019-10-31

Family

ID=68293478

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/084148 Ceased WO2019206196A1 (zh) 2018-04-27 2019-04-24 一种模型更新方法、装置及系统

Country Status (5)

Country Link
US (1) US11451452B2 (zh)
EP (2) EP4625925A3 (zh)
JP (1) JP7159347B2 (zh)
CN (2) CN110417565B (zh)
WO (1) WO2019206196A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022161615A1 (en) * 2021-01-29 2022-08-04 Nokia Technologies Oy Machine learning model renewal
JP2023504103A (ja) * 2019-11-30 2023-02-01 華為技術有限公司 モデル更新システム、モデル更新方法及び関連装置
US12177926B2 (en) 2020-02-04 2024-12-24 Telefonaktiebolaget Lm Ericsson (Publ) Managing service function chains

Families Citing this family (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111652379B (zh) * 2020-05-29 2024-04-16 京东城市(北京)数字科技有限公司 模型管理方法、装置、电子设备及存储介质
CN113839797B (zh) 2020-06-23 2022-11-25 华为技术有限公司 数据处理方法和装置
CN113259147B (zh) * 2020-06-28 2022-07-26 中兴通讯股份有限公司 网元管理方法、装置、计算机设备、介质
CN112085208B (zh) * 2020-07-30 2024-08-20 北京聚云科技有限公司 一种利用云端进行模型训练的方法及装置
CN111898706A (zh) * 2020-08-24 2020-11-06 深圳市富之富信息科技有限公司 模型的智能迭代部署方法、装置、计算机设备及存储介质
CN114765771B (zh) * 2021-01-08 2025-03-14 展讯通信（上海）有限公司 模型更新方法及装置、存储介质、终端、网络侧设备
CN115617621B (zh) * 2021-07-12 2025-08-26 中移(苏州)软件技术有限公司 一种告警方法，告警系统及存储介质
CN116056069B (zh) * 2021-08-12 2025-07-04 大唐移动通信设备有限公司 一种ue能力更新方法、装置及设备
CN116017422A (zh) * 2021-10-21 2023-04-25 华为技术有限公司 一种通信方法及装置
EP4449765A4 (en) * 2021-12-13 2025-01-29 Telefonaktiebolaget LM Ericsson (publ) Managing a wireless device which has available a machine learning model that is operable to connect to a communication network
CN116629340A (zh) * 2022-02-08 2023-08-22 北京邮电大学 模型更新的方法、装置、设备以及存储介质
CN116776987A (zh) * 2022-03-07 2023-09-19 维沃移动通信有限公司 模型的准确度确定方法、装置及网络侧设备
EP4488886A4 (en) * 2022-03-07 2025-06-11 Vivo Mobile Communication Co., Ltd. Method and device for determining model accuracy and network-side device
CN116781516A (zh) * 2022-03-08 2023-09-19 华为技术有限公司 一种更新知识的方法和通信装置
WO2023181139A1 (ja) * 2022-03-22 2023-09-28 日本電信電話株式会社 通信制御装置、通信システム及び通信制御方法
JP7646594B2 (ja) * 2022-03-23 2025-03-17 Kddi株式会社 移動通信ネットワークの制御装置
WO2023187793A1 (en) * 2022-03-31 2023-10-05 Telefonaktiebolaget Lm Ericsson (Publ) First node, second node, third node and methods performed thereby for handling predictive models
EP4539418A4 (en) * 2022-07-05 2025-08-06 Huawei Tech Co Ltd COMMUNICATION METHOD AND APPARATUS
CN119547387A (zh) * 2022-08-10 2025-02-28 中兴通讯股份有限公司 无线通信系统中的人工智能辅助配置方法
CN118509905A (zh) * 2023-02-16 2024-08-16 华为技术有限公司 一种通信方法及装置
EP4690910A1 (en) * 2023-04-06 2026-02-11 NEC Corporation Devices and methods for communication

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102244679A (zh) * 2011-06-30 2011-11-16 南京邮电大学 一种环境感知模型中的环境信息更新方法
CN102339390A (zh) * 2011-09-16 2012-02-01 天津市亚安科技股份有限公司 一种视频监控系统目标模板的更新方法及系统
US8140589B2 (en) * 2007-03-28 2012-03-20 International Business Machines Corporation Autonomic updating of templates in a content management system
CN102571400A (zh) * 2010-12-29 2012-07-11 中国移动通信集团设计院有限公司 一种通信网络资源模型的更新方法、系统及装置
CN106599836A (zh) * 2016-12-13 2017-04-26 北京智慧眼科技股份有限公司 多人脸跟踪方法及跟踪系统

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6058260A (en) * 1995-06-12 2000-05-02 The United States Of America As Represented By The Secretary Of The Army Methods and apparatus for planning and managing a communications network
CN101452386B (zh) * 2007-11-28 2013-05-15 国际商业机器公司 使用正向推理链执行基于上下文模型的策略的方法及策略引擎
SG192002A1 (en) * 2011-01-26 2013-08-30 Google Inc Dynamic predictive modeling platform
CN103139008A (zh) * 2011-11-23 2013-06-05 中兴通讯股份有限公司 检测报文心跳周期的自适应方法和装置
EP3192213A1 (en) * 2014-09-12 2017-07-19 Voellmy, Andreas R. Managing network forwarding configurations using algorithmic policies
GB2539977A (en) * 2015-06-30 2017-01-04 British Telecomm Communications Network
WO2017123674A1 (en) * 2016-01-11 2017-07-20 Equinix, Inc. Architecture for data center infrastructure monitoring
US10404727B2 (en) * 2016-03-25 2019-09-03 Cisco Technology, Inc. Self organizing learning topologies
US10460255B2 (en) * 2016-07-29 2019-10-29 Splunk Inc. Machine learning in edge analytics
US10623932B2 (en) * 2016-12-23 2020-04-14 Accenture Global Solutions Limited Detecting trigger conditions to dynamically update a model based on sensor data in a fog computing environment
US11886823B2 (en) * 2018-02-01 2024-01-30 International Business Machines Corporation Dynamically constructing and configuring a conversational agent learning model

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8140589B2 (en) * 2007-03-28 2012-03-20 International Business Machines Corporation Autonomic updating of templates in a content management system
CN102571400A (zh) * 2010-12-29 2012-07-11 中国移动通信集团设计院有限公司 一种通信网络资源模型的更新方法、系统及装置
CN102244679A (zh) * 2011-06-30 2011-11-16 南京邮电大学 一种环境感知模型中的环境信息更新方法
CN102339390A (zh) * 2011-09-16 2012-02-01 天津市亚安科技股份有限公司 一种视频监控系统目标模板的更新方法及系统
CN106599836A (zh) * 2016-12-13 2017-04-26 北京智慧眼科技股份有限公司 多人脸跟踪方法及跟踪系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3780495A4

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2023504103A (ja) * 2019-11-30 2023-02-01 華為技術有限公司 モデル更新システム、モデル更新方法及び関連装置
JP7401677B2 (ja) 2019-11-30 2023-12-19 華為技術有限公司 モデル更新システム、モデル更新方法及び関連装置
US12177926B2 (en) 2020-02-04 2024-12-24 Telefonaktiebolaget Lm Ericsson (Publ) Managing service function chains
WO2022161615A1 (en) * 2021-01-29 2022-08-04 Nokia Technologies Oy Machine learning model renewal

Also Published As

Publication number Publication date
EP3780495A4 (en) 2021-09-01
EP4625925A2 (en) 2025-10-01
JP7159347B2 (ja) 2022-10-24
JP2021520748A (ja) 2021-08-19
US20210044488A1 (en) 2021-02-11
EP4625925A3 (en) 2026-03-18
CN110417565A (zh) 2019-11-05
CN112883024A (zh) 2021-06-01
EP3780495A1 (en) 2021-02-17
US11451452B2 (en) 2022-09-20
CN110417565B (zh) 2021-01-29

Similar Documents

Publication Publication Date Title
US11451452B2 (en) Model update method and apparatus, and system
EP3855842A1 (en) Method and apparatus for dynamically allocating radio resources in a wireless communication system
US20230370181A1 (en) Communication device predicted future interference information
CN110139325B (zh) 一种网络参数调优方法及装置
JP2025510956A (ja) 機械学習モデル訓練のための通信方法および装置
CN117014120A (zh) 测量反馈方法、装置及存储介质
CN117597969A (zh) Ai数据的传输方法、装置、设备及存储介质
US20240236713A9 (en) Signalling support for split ml-assistance between next generation random access networks and user equipment
US20250344079A1 (en) Managing a plurality of wireless devices that are operable to connect to a communication network
CN118282452A (zh) 一种通信方法及装置
US20230403652A1 (en) Graph-based systems and methods for controlling power switching of components
US20230289655A1 (en) Distributed training in communication networks
US20260094059A1 (en) AI/ML Model Training Using Context Information in Wireless Networks
WO2021152629A1 (en) Method and apparatus for dynamically allocating radio resources in a wireless communication system
WO2024066620A1 (zh) 波束指示方法、装置及存储介质
CN116887327A (zh) 一种QoS预测方法及装置
WO2025156910A1 (zh) 性能监控方法、装置、设备及存储介质
CN119865831A (zh) 人工智能服务的处理方法及装置
CN117580057B (zh) 分布式无线网络升级方法、系统、设备及介质
WO2025076728A1 (en) Timely real-time aiml inference and troubleshooting in telecommunication
CN120825746A (zh) 一种体验质量QoE调整方法、装置、介质及程序产品
WO2025232532A1 (zh) 信息传输方法及通信装置
WO2025139213A1 (zh) 一种通信方法通信装置
WO2025167428A1 (zh) 波束管理中ai/ml模型的数据请求方法及装置
EP4690712A1 (en) Model functionality monitoring

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: 19791815

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020560331

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019791815

Country of ref document: EP

Effective date: 20201112

WWW Wipo information: withdrawn in national office

Ref document number: 2019791815

Country of ref document: EP