Wireless access network, artificial intelligence model acquisition method and device, equipment and medium

Abstract

The invention relates to a wireless access network, an artificial intelligence model acquisition method and device, equipment and a medium. The method comprises the following steps: the wireless access network comprises a wireless network element and an artificial intelligence model service network element, and the wireless network element at least comprises a smart base station iNB; the artificial intelligence model service network element is used for training an artificial intelligence model according to the data acquired by the wireless network element and providing the artificial intelligence model for the wireless network element; and the wireless network element is used for sending the collected data to the artificial intelligence model service network element, so that the artificial intelligence model service network element trains an artificial intelligence model according to the collected data, and is used for reasoning according to the artificial intelligence model provided by the artificial intelligence model service network element to obtain a target communication parameter. According to the invention, a new wireless access network architecture design and a new flow design are provided for endogenous intelligent characteristics of the wireless access network.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a wireless access network, an artificial intelligence model acquisition method, device, equipment and medium. Background Art

[0002] Endogenous intelligence is an important feature of the next generation of mobile wireless networks (such as 6G). This feature emphasizes that AI (Artificial Intelligence) capabilities are inherent in and used by wireless access networks, helping to continuously optimize and evolve wireless access networks and form smart wireless access networks.

[0003] Currently, there is an urgent need to provide new wireless access network architecture design and process design based on the inherent intelligent characteristics of wireless access networks. Summary of the invention

[0004] The embodiments of the present application provide a wireless access network, an artificial intelligence model acquisition method, an apparatus, a device and a medium, and in particular provide a new wireless access network architecture design and a process design for the intrinsic intelligent characteristics of the wireless access network.

[0005] In a first aspect, an embodiment of the present application provides a wireless access network. The wireless access network includes: a wireless network element and an artificial intelligence model service network element, wherein the wireless network element includes at least an intelligent base station iNB;

[0006] The artificial intelligence model service network element is used to train the artificial intelligence model according to the data collected by the wireless network element, and to provide the artificial intelligence model to the wireless network element;

[0007] The wireless network element is used to send the collected data to the artificial intelligence model service network element so that the artificial intelligence model service network element can train the artificial intelligence model according to the collected data, and to perform model inference according to the artificial intelligence model provided by the artificial intelligence model service network element to obtain target communication parameters.

[0008] In one of the embodiments, the wireless network element further includes an intelligent terminal iWT, and the iWT communicates with the iNB via an air interface.

[0009] In one of the embodiments, the artificial intelligence model service network element includes a data dedicated unit DDU and a model providing network element;

[0010] The DDU is used to receive and store data collected by the wireless network element, and to train an artificial intelligence model based on the data collected by the wireless network element;

[0011] The model providing network element is used to provide the artificial intelligence model obtained by DDU training to the wireless network element.

[0012] In one of the embodiments, the DDU is also used to store the trained artificial intelligence model, model parameters of the artificial intelligence model and / or inference results of the artificial intelligence model.

[0013] In one of the embodiments, the DDU is further used to trigger the wireless network element to collect data;

[0014] The wireless network element is further configured to collect data under the triggering of the DDU and send the collected data to the DDU.

[0015] In one of the embodiments, the DDU includes a data storage unit DSU, a data processing unit DPU and a data management unit DMU;

[0016] The DSU is at least used to store data collected by the wireless network element;

[0017] The DPU is used for artificial intelligence model training;

[0018] The DMU is used to trigger the DPU to perform artificial intelligence model training and to trigger data transmission between the DSU and the DPU.

[0019] In one of the embodiments, the DSU is also used to store the trained artificial intelligence model, model parameters of the artificial intelligence model and / or inference results of the artificial intelligence model.

[0020] In one of the embodiments, the DMU is further used to trigger the wireless network element to collect data, and to trigger the DSU to receive data collected by the wireless network element;

[0021] The wireless network element is further used to collect data under the triggering of the DMU and send the collected data to the DSU;

[0022] The DSU is further configured to receive data collected by the wireless network element under the triggering of the DMU.

[0023] In one of the embodiments, the model providing network element is used to locally store the trained artificial intelligence model and provide the artificial intelligence model to the wireless network element based on the model information of the locally stored artificial intelligence model; or,

[0024] The model providing network element is used to obtain model information of the trained artificial intelligence model from the DDU, and provide the artificial intelligence model to the wireless network element based on the model information obtained from the DDU.

[0025] In one embodiment, the model information includes at least one of the following:

[0026] Model files for AI models;

[0027] The storage address of the model file of the artificial intelligence model;

[0028] Identification of the artificial intelligence model.

[0029] In one of the embodiments, the model provides a network element and is also used to trigger the DDU to execute an artificial intelligence model training process when the artificial intelligence model requested by the wireless network element does not exist in the DDU.

[0030] In one embodiment, the model providing network element includes: an artificial intelligence service server AISS; or

[0031] The AISS and the artificial intelligence service agent AISP, wherein the AISP is arranged between the AISS and the wireless network element.

[0032] In one of the embodiments, the network architecture of the wireless access network is a reference point architecture; or, the network architecture of the wireless access network is a partially service-based architecture.

[0033] In one of the embodiments, when the network architecture of the wireless access network is a reference point architecture, the network element setting method of the wireless access network includes at least one of the following:

[0034] There are multiple levels of AISP;

[0035] There are multiple levels of AISS;

[0036] The AISP is jointly provided with the iNB;

[0037] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISP is jointly provided with the DSU;

[0038] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISP is jointly provided with the DMU;

[0039] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DSU and the DPU are jointly provided;

[0040] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DPU;

[0041] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DSU;

[0042] The iNB is divided into multiple units.

[0043] In one of the embodiments, when the network architecture of the wireless access network is a partially service-oriented architecture, the wireless access network further comprises: an access network profile function network element and an access network communication proxy network element;

[0044] The access network archive function network element is used for each service-oriented network element in the wireless access network to register in the access network archive function network element, and for each service-oriented network element in the wireless access network to discover other registered service-oriented network elements in the access network archive function network element;

[0045] The access network communication proxy network element is used to forward a message of a service-oriented network element in the wireless access network to another service-oriented network element in the wireless access network.

[0046] In one of the embodiments, when the network architecture of the wireless access network is a partially service-oriented architecture, the iNB includes a smart base station non-real-time function and a smart base station real-time function;

[0047] The non-real-time function of the smart base station is used to perform non-real-time processing of wireless access;

[0048] The real-time function of the smart base station is used to perform real-time processing of wireless access.

[0049] In one of the embodiments, when the network architecture of the wireless access network is a partially service-oriented architecture, the network element setting method of the wireless access network includes at least one of the following:

[0050] There are multiple levels of AISP;

[0051] There are multiple levels of AISS;

[0052] The AISP is jointly provided with the iNB;

[0053] The non-real-time function of the smart base station and the real-time function of the smart base station are combined;

[0054] The non-real-time function of the smart base station and the real-time function of the smart base station are divided into multiple units;

[0055] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISP is jointly provided with the DSU;

[0056] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISS is jointly provided with the DMU;

[0057] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DSU and the DPU are jointly provided;

[0058] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DPU;

[0059] In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DSU;

[0060] All or part of the AISS, the AISP and the DDU are arranged in a core network;

[0061] All or part of the AISS, the AISP, the DSU, the DPU and the DMU are arranged in a core network.

[0062] In one of the embodiments, when the network architecture of the wireless access network is a reference point architecture, the wireless access network includes the following interfaces: an A1 interface, an A2 interface, a D1 interface, and an air interface;

[0063] The A1 interface is an interface between the wireless network element and the AISS and an interface between the AISP and the AISS;

[0064] The A2 interface is an interface between the AISS and the DDU;

[0065] The D1 interface is an interface between the DDU and the wireless network element;

[0066] The air interface is used for wireless access control and for carrying air interface transmission of the A1 interface and the D1 interface.

[0067] In one of the embodiments, when the DDU includes the DSU, the DPU and the DMU, ​​the D1 interface includes a D1-C interface and a D1-D interface;

[0068] The D1-C interface is an interface between the DMU and the wireless network element;

[0069] The D1-D interface is an interface between the DSU and the wireless network element.

[0070] In one of the embodiments, when the DDU includes the DSU, the DPU and the DMU, ​​the radio access network further includes the following interfaces: a D2 interface, a D3 interface and a D4 interface;

[0071] The D2 interface is an interface between the DMU and the DSU;

[0072] The D3 interface is an interface between the DMU and the DPU;

[0073] The D4 interface is an interface between the DSU and the DPU.

[0074] In one embodiment, when the network architecture of the wireless access network is a partially service-oriented architecture, the wireless access network includes the following interfaces: AISS Interface, N DDU Interface, N DCU Interface, iF interface and air interface;

[0075] The N AISS The interface is an interface for the AISS and the AISP to provide services externally;

[0076] The N DDU The interface is the interface through which the DDU provides services externally;

[0077] The N DCU The interface is an interface for the wireless network element to provide services externally;

[0078] The iF interface is an interface between the non-real-time function of the smart base station and the real-time function of the smart base station;

[0079] The air interface is used for wireless access control and for carrying the N AISS Interface, the N DDU interface and the N DCU Air interface transmission of the interface.

[0080] In one of the embodiments, when the DDU includes the DSU, the DPU and the DMU, ​​the wireless access network further includes the following interface: DMU Interface, N DSU Interface and N DPU interface;

[0081] The N DMU The interface is an interface for the DMU to provide services externally;

[0082] The N DSU The interface is an interface for the DSU to provide services externally;

[0083] The NDPU Interface The DPU provides services externally.

[0084] In one of the embodiments, the wireless network element is further used to optimize the corresponding communication process according to the target communication parameters.

[0085] In a second aspect, an embodiment of the present application provides an artificial intelligence model acquisition method. The artificial intelligence model acquisition method is used for the model providing network element of the wireless access network described in the first aspect above, and the method includes:

[0086] Receiving a model request message sent by an artificial intelligence model client network element in the wireless access network;

[0087] Obtain a model response message according to the model request message;

[0088] The model response message is sent to the artificial intelligence model client network element.

[0089] In one embodiment, the model request message includes at least one of the following parameters:

[0090] Functional identification;

[0091] Scene identification;

[0092] Optional inputs to AI models;

[0093] Constraints of AI models.

[0094] In one of the embodiments, if the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes an identifier of the first artificial intelligence model, a model file of the first artificial intelligence model, or a storage address of the model file of the first artificial intelligence model.

[0095] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

[0096] In one embodiment, the identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0097] The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0098] In one embodiment, the method further comprises:

[0099] If the first artificial intelligence model is not obtained by the local search of the model providing network element, a model acquisition request message is sent to the DDU in the radio access network;

[0100] A model acquisition response message is received from the DDU based on the model acquisition request message, wherein the model acquisition response message is used by the model providing network element to obtain the model response message according to the model acquisition response message.

[0101] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0102] Functional identification;

[0103] Scene identification;

[0104] Optional inputs to AI models;

[0105] Constraints of AI models.

[0106] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

[0107] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0108] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0109] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0110] In one of the embodiments, the model acquisition request message further includes indication information for allowing automatic model generation, and the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0111] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0112] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0113] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0114] In one embodiment, the method further comprises:

[0115] If the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, a model generation request message is sent to the DDU, where the model generation request message is used to instruct the DDU to execute a training process for a third artificial intelligence model.

[0116] In one embodiment, the model generation request message includes at least one of the following parameters:

[0117] Functional identification;

[0118] Scene identification;

[0119] Optional inputs to AI models;

[0120] Constraints of AI models.

[0121] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0122] In one embodiment, the method further comprises:

[0123] receiving a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model;

[0124] The model response message is obtained according to the model generation response message.

[0125] In one of the embodiments, if the DDU successfully trains to obtain the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0126] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0127] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0128] In one of the embodiments, if the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no available model indication information.

[0129] In a third aspect, an embodiment of the present application provides an artificial intelligence model acquisition method. The artificial intelligence model acquisition method is used for the DDU of the wireless access network described in the first aspect above, and the method includes:

[0130] Receiving a model acquisition request message sent by a model providing network element of the wireless access network;

[0131] A model acquisition response message is sent to the model providing network element based on the model acquisition request message.

[0132] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0133] Functional identification;

[0134] Scene identification;

[0135] Optional inputs to AI models;

[0136] Constraints of AI models.

[0137] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes an identifier of the second artificial intelligence model, a model file of the second artificial intelligence model, or a storage address of the model file of the second artificial intelligence model.

[0138] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0139] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0140] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0141] In one of the embodiments, the model acquisition request message further includes indication information for allowing automatic model generation, and the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0142] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0143] In one embodiment, the method further comprises:

[0144] If the DDU retrieval fails to obtain the second artificial intelligence model, the training process of the third artificial intelligence model is automatically executed.

[0145] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0146] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0147] In one embodiment, the method further comprises:

[0148] Receiving a model generation request message sent by the model providing network element after receiving the model acquisition response message;

[0149] A training process of a third artificial intelligence model is executed based on the model generation request message.

[0150] In one embodiment, the model generation request message includes at least one of the following parameters:

[0151] the function identifier;

[0152] the scene identifier;

[0153] optional inputs to the artificial intelligence model;

[0154] The constraints of the artificial intelligence model.

[0155] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0156] In one embodiment, the training process of executing the third artificial intelligence model includes:

[0157] Sending a data collection request message to the data collection unit DCU;

[0158] Receiving a data acquisition response message and collected data sent by the DCU based on the data acquisition request message;

[0159] The third artificial intelligence model is trained according to the data sent by the DCU.

[0160] In one embodiment, the data collection request message includes at least one of the following parameters:

[0161] the function identifier;

[0162] the scene identifier;

[0163] optional inputs to the artificial intelligence model;

[0164] The address for receiving the collected data;

[0165] The amount of data to be collected.

[0166] In one embodiment, the data collection response message includes at least one of the following parameters:

[0167] the function identifier;

[0168] the scene identifier;

[0169] Optional input to the artificial intelligence model.

[0170] In one embodiment, the method further comprises:

[0171] After executing the training process of the third artificial intelligence model, a model generation response message is sent to the model providing network element.

[0172] In one of the embodiments, if the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0173] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0174] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0175] In one of the embodiments, if the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

[0176] In a fourth aspect, an embodiment of the present application provides an artificial intelligence model acquisition device. The artificial intelligence model acquisition device is used for the model providing network element of the wireless access network described in the first aspect above, and the device includes:

[0177] A receiving unit, configured to receive a model request message sent by an artificial intelligence model client network element in the wireless access network;

[0178] An acquiring unit, configured to acquire a model response message according to the model request message;

[0179] A sending unit is used to send the model response message to the artificial intelligence model client network element.

[0180] In a fifth aspect, an embodiment of the present application provides an artificial intelligence model acquisition device. The artificial intelligence model acquisition device is used for the DDU of the wireless access network described in the first aspect above, and the device includes:

[0181] A receiving unit, configured to receive a model acquisition request message sent by a model providing network element of the radio access network;

[0182] A sending unit is used to send a model acquisition response message to the model providing network element based on the model acquisition request message.

[0183] In a sixth aspect, an embodiment of the present application provides a model providing network element. The model providing network element includes a memory, a transceiver, and a processor:

[0184] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0185] Controlling the transceiver to receive a model request message sent by an artificial intelligence model client network element in the wireless access network;

[0186] Obtain a model response message according to the model request message;

[0187] Control the transceiver to send the model response message to the artificial intelligence model client network element.

[0188] In one embodiment, the model request message includes at least one of the following parameters:

[0189] Functional identification;

[0190] Scene identification;

[0191] Optional inputs to AI models;

[0192] Constraints of AI models.

[0193] In one of the embodiments, if the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes an identifier of the first artificial intelligence model, a model file of the first artificial intelligence model, or a storage address of the model file of the first artificial intelligence model.

[0194] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

[0195] In one embodiment, the identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0196] The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0197] In one embodiment, the processor is further configured to perform the following operations:

[0198] If the first artificial intelligence model is not obtained by the local search of the model providing network element, controlling the transceiver to send a model acquisition request message to the DDU in the radio access network;

[0199] The transceiver is controlled to receive a model acquisition response message sent by the DDU based on the model acquisition request message, wherein the model acquisition response message is used for the model providing network element to obtain the model response message according to the model acquisition response message.

[0200] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0201] Functional identification;

[0202] Scene identification;

[0203] Optional inputs to AI models;

[0204] Constraints of AI models.

[0205] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

[0206] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0207] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0208] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0209] In one of the embodiments, the model acquisition request message further includes indication information for allowing automatic model generation, and the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0210] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0211] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0212] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0213] In one embodiment, the processor is further configured to perform the following operations:

[0214] If the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, the transceiver is controlled to send a model generation request message to the DDU, and the model generation request message is used to instruct the DDU to execute the training process of the third artificial intelligence model.

[0215] In one embodiment, the model generation request message includes at least one of the following parameters:

[0216] Functional identification;

[0217] Scene identification;

[0218] Optional inputs to AI models;

[0219] Constraints of AI models.

[0220] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0221] In one embodiment, the processor is further configured to perform the following operations:

[0222] Controlling the transceiver to receive a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model;

[0223] The model response message is obtained according to the model generation response message.

[0224] In one of the embodiments, if the DDU successfully trains to obtain the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0225] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0226] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0227] In one of the embodiments, if the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no available model indication information.

[0228] In a seventh aspect, an embodiment of the present application provides a data dedicated unit. The data dedicated unit includes a memory, a transceiver, and a processor:

[0229] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0230] Controlling the transceiver to receive a model acquisition request message sent by a model providing network element of the wireless access network;

[0231] The transceiver is controlled to send a model acquisition response message to the model providing network element based on the model acquisition request message.

[0232] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0233] Functional identification;

[0234] Scene identification;

[0235] Optional inputs to AI models;

[0236] Constraints of AI models.

[0237] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes an identifier of the second artificial intelligence model, a model file of the second artificial intelligence model, or a storage address of the model file of the second artificial intelligence model.

[0238] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0239] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0240] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0241] In one of the embodiments, the model acquisition request message further includes indication information for allowing automatic model generation, and the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0242] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0243] In one embodiment, the processor is further configured to perform the following operations:

[0244] If the DDU retrieval fails to obtain the second artificial intelligence model, the training process of the third artificial intelligence model is automatically executed.

[0245] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0246] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0247] In one embodiment, the processor is further configured to perform the following operations:

[0248] Controlling the transceiver to receive a model generation request message sent by the model providing network element after receiving the model acquisition response message;

[0249] A training process of a third artificial intelligence model is executed based on the model generation request message.

[0250] In one embodiment, the model generation request message includes at least one of the following parameters:

[0251] the function identifier;

[0252] the scene identifier;

[0253] optional inputs to the artificial intelligence model;

[0254] The constraints of the artificial intelligence model.

[0255] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0256] In one embodiment, the processor is specifically configured to perform the following operations:

[0257] Controlling the transceiver to send a data acquisition request message to a data acquisition unit DCU;

[0258] Controlling the transceiver to receive the data acquisition response message and the collected data sent by the DCU based on the data acquisition request message;

[0259] The third artificial intelligence model is trained according to the data sent by the DCU.

[0260] In one embodiment, the data collection request message includes at least one of the following parameters:

[0261] the function identifier;

[0262] the scene identifier;

[0263] optional inputs to the artificial intelligence model;

[0264] The address for receiving the collected data;

[0265] The amount of data to be collected.

[0266] In one embodiment, the data collection response message includes at least one of the following parameters:

[0267] the function identifier;

[0268] the scene identifier;

[0269] Optional input to the artificial intelligence model.

[0270] In one embodiment, the processor is further configured to perform the following operations:

[0271] After executing the training process of the third artificial intelligence model, the transceiver is controlled to send a model generation response message to the model providing network element.

[0272] In one of the embodiments, if the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0273] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0274] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0275] In one of the embodiments, if the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

[0276] In an eighth aspect, an embodiment of the present application provides a processor-readable storage medium. The processor-readable storage medium stores a program, and the program is used to enable a processor to execute the method described in the second aspect.

[0277] In a ninth aspect, an embodiment of the present application provides a processor-readable storage medium, wherein the processor-readable storage medium stores a program, and the program is used to enable the processor to execute the method described in the third aspect.

[0278] In a tenth aspect, an embodiment of the present application provides a chip, wherein the chip includes a programmable logic circuit and / or program instructions, and when the chip is running, the steps of the method described in the second aspect or the third aspect are implemented.

[0279] The above-mentioned wireless access network, artificial intelligence model acquisition method, device, equipment and medium, the wireless access network includes a wireless network element and an artificial intelligence model service network element, the wireless network element at least includes an intelligent base station iNB, wherein the artificial intelligence model service network element is used to train the artificial intelligence model according to the data collected by the wireless network element, and to provide the artificial intelligence model to the wireless network element; the wireless network element is used to send the collected data to the artificial intelligence model service network element, so that the artificial intelligence model service network element can train the artificial intelligence model according to the collected data, and to perform model reasoning according to the artificial intelligence model provided by the artificial intelligence model service network element to obtain the target communication parameters. In this way, by designing an artificial intelligence model service network element in the wireless access network, the artificial intelligence model service network element can cooperate with the wireless network element, and the artificial intelligence model service network element can train the artificial intelligence model and provide the trained artificial intelligence model to the wireless network element. The embodiment of the present application provides a new wireless access network architecture design and process design for the intrinsic intelligent characteristics of the wireless access network. BRIEF DESCRIPTION OF THE DRAWINGS

[0280] Figure 1 A schematic diagram of an exemplary wireless access network architecture with external intelligent features;

[0281] Figure 2-1A flowchart of training an AI model through NG-RAN OAM and performing AI model reasoning through NG-RAN node;

[0282] Figure 2-2 A schematic diagram of the process of training and reasoning an artificial intelligence model through an NG-RAN node;

[0283] Figure 3 A schematic diagram of a deployment mode of RAN-DCA distributed deployment;

[0284] Figure 4 A schematic diagram of a deployment mode of RAN-DCA centralized deployment;

[0285] Figure 5 A schematic diagram of a deployment mode of RAN-DCA hybrid deployment;

[0286] Figure 6 This is a schematic diagram of the architecture of a wireless access network in which the network architecture is a reference point architecture in one embodiment of the present application;

[0287] Figure 7 This is a schematic diagram of the architecture of a wireless access network in which the network architecture is a reference point architecture in another embodiment of the present application;

[0288] Figure 8 This is a schematic diagram of the architecture of a wireless access network in which the network architecture is a partially service-based architecture in another embodiment of the present application;

[0289] Fig. 9 This is a schematic diagram of the architecture of a wireless access network in which the network architecture is a partially service-based architecture in another embodiment of the present application;

[0290] Fig.10 A flowchart of an artificial intelligence model acquisition method in one embodiment of the present application;

[0291] Fig.11 This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0292] Fig.12 This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0293] Fig.13 This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0294] Fig.14 This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0295] Fig.15This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0296] Fig.16 This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0297] Fig.17 This is a flowchart of a method for obtaining an artificial intelligence model in another embodiment of the present application;

[0298] Fig.18 This is a structural block diagram of an artificial intelligence model acquisition device in one embodiment of the present application;

[0299] Fig.19 This is a structural block diagram of an artificial intelligence model acquisition device in one embodiment of the present application;

[0300] Fig. 20 A schematic diagram of the structure of a network element is provided for a model in one embodiment of the present application;

[0301] Fig.21 This is a schematic diagram of the structure of a data dedicated unit in one embodiment of the present application;

[0302] Fig. 22 This is a schematic diagram of the structure of a chip in one embodiment of the present application. DETAILED DESCRIPTION

[0303] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0304] Endogenous intelligence is an important feature of the next generation of mobile wireless networks (such as 6G). This feature emphasizes that AI capabilities are inherent in and used by wireless access networks, helping to continuously optimize and evolve wireless access networks and form smart wireless access networks.

[0305] Currently, there is an urgent need to provide new wireless access network architecture design and process design based on the inherent intelligent characteristics of wireless access networks.

[0306] See also Figure 1 , Figure 1The figure is a schematic diagram of an exemplary wireless access network architecture with external intelligent features. In O-RAN (Open Radio Access Network), Near-RT RIC (Near RealTime RAN Intelligent Controller) collects RAN (Radio Access Network) information through the E2 interface and obtains A1-P information (i.e., policy information) from the A1 interface, and then intelligently processes the RAN information, A1-P information, and A1 interface auxiliary information to obtain signaling for guiding RAN behavior (RAN behavior such as configuring wireless parameters, mobility events, etc.).

[0307] like Figure 1 As shown, Near-RT RIC consists of Near-RT RIC Platform and xApp (eXtended APPlication), where intelligent processing is located in xApp, and intelligent processing includes processes such as reasoning of artificial intelligence models. Near-RT RIC Platform connects Non-RT RIC (Non Real-Time RAN Intelligent Controller) and RAN, and is responsible for forwarding the inputs about intelligent processing provided by Non-RT RIC and RAN to xApp. After xApp completes the reasoning of the artificial intelligence model based on these inputs, Near-RT RIC Platform passes the intelligent output of the artificial intelligence model to RAN.

[0308] In the related technology, in the 3GPP (3rd Generation Partnership Project) NR (New Radio) network, NG-RAN (Next Generation Radio Access Network) OAM (Operation Administration and Maintenance) or NG-RAN node is reused as a place for artificial intelligence model training and reasoning.

[0309] See also Figure 2-1 and Figure 2-2 , Figure 2-1 The figure is a flowchart of an exemplary process of training an artificial intelligence model through NG-RAN OAM and performing artificial intelligence model reasoning through NG-RAN node.

[0310] Figure 2-1 The process shown is as follows:

[0311] 1. NG-RAN node 1 sends Measurement Configuration to UE (User Equipment);

[0312] 2. UE performs Measurement;

[0313] 3. The UE reports a UE Measurement report to NG-RAN node 1;

[0314] 4. NG-RAN node 1 and NG-RAN node 2 send Input data for Model Training to OAM;

[0315] 5. OAM conducts Model Training;

[0316] 6. OAM sends AI / ML Model Deployment / Update to NG-RAN node1.

[0317] 7. The UE reports UE Measurement location to NG-RAN node1;

[0318] 8. NG-RAN node 2 sends Input data for load balancing inference to NG-RAN node 1.

[0319] 9. NG-RAN node 1 performs Load Balancing predictions (Model Inference).

[0320] 10. Optionally, NG-RAN node 1 sends Model Performance Feedback to OAM;

[0321] 11. UE and NG-RAN node 1 perform Load Balancing action;

[0322] 12. NG-RAN node 1 and NG-RAN node 2 send Feedback to OAM.

[0323] Figure 2-2 The figure is a flowchart of an exemplary process of artificial intelligence model training and reasoning through NG-RAN node.

[0324] Figure 2-2 The process shown is as follows:

[0325] 1. NG-RAN node 1 sends Measurement Configuration to UE;

[0326] 2. UE performs Measurement;

[0327] 3. The UE reports a UE Measurement report to NG-RAN node 1;

[0328] 4. NG-RAN node2 sends Input data for Model Training to NG-RAN node1;

[0329] 5. NG-RAN node 1 performs Model Training;

[0330] 6. The UE reports the UE Measurement location to NG-RAN node1;

[0331] 7. NG-RAN node2 sends Input data for load balancing inference to NG-RAN node1;

[0332] 8. NG-RAN node1 performs Load Balancing predictions;

[0333] 9. UE, NG-RAN node1 and NG-RAN node2 Load Balancing action;

[0334] 10. NG-RAN node 2 sends Feedback to NG-RAN node 1.

[0335] In the related art, the RAN-DCA (Radio Access Network-Data Collection and Analysis) module is also discussed to be responsible for the AI ​​model training, reasoning and other processes. DCA can be deployed in RAN in various forms. For example, see Figure 3-Figure 5 , which shows a schematic diagram of three deployment modes of DCA in a wireless access network, where:

[0336] Figure 3 A deployment method of RAN-DCA distributed deployment is shown, in which DCA is deployed in the CU (Centralized Unit) and DU (Distributed Unit) of gNB (gNodeB, next-generation base station), and each DCA jointly performs federated learning to train the artificial intelligence model.

[0337] Figure 4 A deployment method of RAN-DCA centralized deployment is shown, where DCA is deployed separately to train artificial intelligence models. Figure 5 Another deployment mode of RAN-DCA hybrid deployment is shown, in which DCA is deployed in both distributed and centralized modes.

[0338] However, the intelligent features of the wireless access network architecture mentioned above are not inherent in the wireless access network architecture, but are added to the existing architecture as additional features. For networks that have been deployed on a large scale, such as 5G, it is very difficult to add intelligent features.

[0339] In addition, the "high computing power" and "high storage power" characteristics of the wireless network elements in the wireless access network architecture mentioned above are significant, which leads to an increase in the design, development, deployment, operation and maintenance costs of the wireless network elements.

[0340] Furthermore, reliable wireless access and wireless data transmission capabilities remain the primary design goals of wireless access networks, and the computing power, storage capacity, and wireless capabilities of wireless network elements need to be emphasized depending on different scenarios.

[0341] In view of this, the embodiments of the present application provide a new wireless access network (which may be referred to as a smart wireless access network) architecture design and process design targeting the inherent intelligent characteristics of the wireless access network.

[0342] Hereinafter, the new wireless access network architecture design provided by the embodiment of the present application for the intrinsic intelligent characteristics of the wireless access network is first introduced.

[0343] An embodiment of the present application provides a wireless access network, which includes: a wireless network element and an artificial intelligence model service network element.

[0344] The wireless network element includes at least an iNB (Intelligent Node B), which is responsible for wireless access processing and may also have certain computing and storage capabilities.

[0345] Among them, the wireless network element is used to send the collected data to the artificial intelligence model service network element, so that the artificial intelligence model service network element can train the artificial intelligence model based on the data collected by the wireless network element.

[0346] Exemplarily, the data collected by the wireless network element may include RAN data, etc. The artificial intelligence model may be an AI (Artificial Intelligence) model or an ML (Machine Learning) model, and of course, a DL (Deep Learning) model or a neural network model, etc., and the model type of the artificial intelligence model is not limited here.

[0347] The artificial intelligence model service network element is used to train the artificial intelligence model based on the data collected by the wireless network element to obtain a trained artificial intelligence model, and the artificial intelligence model service network element is used to provide the trained artificial intelligence model to the wireless network element.

[0348] In the embodiments of the present application, artificial intelligence models include base station RRM (Radio Resource Management) model, base station MAC (Medium Access Control) scheduling model, base station uplink channel assessment model, terminal RRM model, terminal downlink channel assessment model, and the like.

[0349] After the artificial intelligence model service network element provides the trained artificial intelligence model to the wireless network element, the wireless network element is also used to perform model inference based on the artificial intelligence model provided by the artificial intelligence model service network element to obtain target communication parameters.

[0350] In an embodiment of the present application, the wireless network element is also used to optimize the corresponding communication process according to the target communication parameters, that is, the wireless network element can use the artificial intelligence model to achieve its own performance optimization.

[0351] Exemplarily, taking the wireless network element being iNB as an example and the above-mentioned corresponding communication process being PDSCH (Physical Downlink Shared Channel) scheduling as an example, the artificial intelligence model service network element provides the iNB with a trained artificial intelligence model, which is used for PDSCH scheduling.

[0352] The iNB performs model inference based on the artificial intelligence model to obtain the target communication parameters. As an implementation method, the target communication parameters may also be a PDSCH scheduling optimization strategy. The iNB may optimize the PDSCH scheduling result initially generated by the iNB based on the PDSCH scheduling optimization strategy, for example, adjusting the PDSCH resource allocation ratio in the PDSCH scheduling result initially generated by the iNB, and so on.

[0353] Of course, the above-mentioned corresponding communication process can also be PDCCH (Physical Downlink Control Channel) scheduling, then the artificial intelligence model service network element provides the iNB with the trained artificial intelligence model for PDCCH scheduling.

[0354] For example, taking the wireless network element as iNB, after receiving the analog communication signal, the iNB needs to perform analog-to-digital conversion on the analog communication signal to obtain a digital communication signal. The analog-to-digital conversion process involves the signal detection and estimation process. The above-mentioned corresponding communication process can be the signal detection and estimation process. The iNB can optimize the results of the iNB signal detection and estimation process based on the inference results of the artificial intelligence model corresponding to the communication process.

[0355] In the embodiments of the present application, the network elements that can use artificial intelligence models to optimize their own performance are not limited to wireless network elements, but can also be any other network elements. There is no restriction on the types of network elements that can use artificial intelligence models to optimize their own performance.

[0356] In this way, by designing an artificial intelligence model service network element in the wireless access network, the artificial intelligence model service network element can train the artificial intelligence model by cooperating with the wireless network element, and provide the trained artificial intelligence model to the wireless network element. The embodiment of the present application provides a new wireless access network architecture design and process design for the intrinsic intelligent characteristics of the wireless access network.

[0357] In a possible implementation, the wireless network element also includes an iWT (intelligent Wireless Terminal), and the iWT communicates with the iNB through an air interface. Both the iWT and the iNB can collect data and send the collected data to the artificial intelligence model service network element for training the artificial intelligence model. Both the iWT and the iNB can also optimize their own performance according to the artificial intelligence model provided by the artificial intelligence model service network element.

[0358] Exemplarily, taking the wireless network element as iWT, iWT needs to perform uplink channel detection during uplink data transmission. The communication process for optimization processing can also be an uplink channel detection process. iWT can optimize the result of the uplink channel detection process of iWT according to the inference result of the artificial intelligence model corresponding to the communication process, and so on.

[0359] In other words, both iWT and iNB have AISC (AI Service Client) and DCU (Data Collection Unit) functions. AISC can use artificial intelligence models to optimize its own performance, and DCU is used to perform data collection.

[0360] In an embodiment of the present application, the artificial intelligence model service network element may include a DDU (Data Dedicated Unit) and a model providing network element.

[0361] First, the DDU is introduced. The DDU is used to receive and store data collected by wireless network elements (such as the above-mentioned iWT and / or iNB), and is used to train an artificial intelligence model based on the data collected by the wireless network elements.

[0362] In a possible implementation, the DDU is also used to store the trained artificial intelligence model, the model parameters of the artificial intelligence model and / or the inference results of the artificial intelligence model. That is, the DDU is used to perform data storage and data processing, the data includes RAN data, analytical result, model framework, etc., the data processing includes model training, association analysis, etc., the DDU trains the artificial intelligence model, and returns the trained artificial intelligence model to the model providing network element.

[0363] In some implementations, the DDU can also organize data collection, storage, calculation, etc. according to the needs of the network elements provided by the model, that is, the DDU is also used to trigger the wireless network elements to collect data. In this way, the wireless network elements are also used to collect data under the triggering of the DDU and send the collected data to the DDU.

[0364] The method process of the DDU triggering the wireless network element to collect data, the wireless network element collecting data under the triggering of the DDU and sending the collected data to the DDU will be introduced in the following embodiments.

[0365] The model providing network element is introduced below. In the embodiment of the present application, the model providing network element is used to provide the artificial intelligence model obtained by DDU training to the wireless network element.

[0366] As mentioned above, the artificial intelligence model trained by DDU will return the trained artificial intelligence model to the model providing network element, so that the model providing network element can provide the artificial intelligence model trained by DDU to the wireless network element.

[0367] In one possible implementation, the model providing network element is used to locally store the trained artificial intelligence model, and provide the artificial intelligence model to the wireless network element based on the model information of the locally stored artificial intelligence model.

[0368] In another possible implementation, the model providing network element is used to obtain model information of the trained artificial intelligence model from the DDU, and provide the artificial intelligence model to the wireless network element based on the model information obtained from the DDU.

[0369] Exemplarily, if the wireless network element needs to obtain an artificial intelligence model, it can send a model request message to the model providing network element. After receiving the model request message, the model providing network element locally searches whether there is an artificial intelligence model that meets the model request message. If the model providing network element stores the artificial intelligence model locally, it provides the artificial intelligence model to the wireless network element based on the model information of the locally stored artificial intelligence model; if the model providing network element does not store the artificial intelligence model locally, the model providing network element obtains the model information of the trained artificial intelligence model from the DDU, and provides the artificial intelligence model to the wireless network element based on the model information obtained from the DDU.

[0370] Of course, in some possible implementations, the artificial intelligence model requested by the wireless network element may not exist in the DDU. The model providing network element is also used to trigger the DDU to execute the artificial intelligence model training process when the artificial intelligence model requested by the wireless network element does not exist in the DDU. In this way, the DDU executes the artificial intelligence model training process. After training the artificial intelligence model that meets the request of the wireless network element, the model providing network element can obtain the model information of the trained artificial intelligence model from the DDU, and provide the artificial intelligence model to the wireless network element based on the model information obtained from the DDU.

[0371] In an embodiment of the present application, the model information includes at least one of the following: a model file of the artificial intelligence model; a storage address of the model file of the artificial intelligence model; an identifier of the artificial intelligence model. For example, the model information includes the identifier of the artificial intelligence model and the model file of the artificial intelligence model, or the model information includes the identifier of the artificial intelligence model and the storage address of the model file of the artificial intelligence model.

[0372] Among them, the identification of the artificial intelligence model is a combination of a function identification and a scene identification; or, the identification of the artificial intelligence model includes a function identification, a scene identification and other information. No specific restriction is made on the content contained in the identification of the artificial intelligence model.

[0373] In the embodiment of the present application, as an implementation method, the model providing network element includes AISS (AI Service Server), and AISS provides an available AI artificial intelligence model; or, the model providing network element includes AISS and AISP (AI Service Proxy), wherein AISP is arranged between AISS and wireless network element, and by setting AISP, the delay of AISC in acquiring artificial intelligence model and the processing burden of AISS can be reduced. For AISC, the functions of AISP and AISS are equivalent.

[0374] The architecture of the wireless access network according to the embodiment of the present application is described below with reference to the drawings.

[0375] In an embodiment of the present application, the network architecture of the wireless access network is a reference point architecture; or, the network architecture of the wireless access network is a partially service-oriented architecture.

[0376] First, we introduce the reference point architecture. For example, see Figure 6 , Figure 6 The diagram is a schematic diagram of an architecture of an intelligent radio access network (iRAN) with an exemplary network architecture as a reference point architecture.

[0377] like Figure 6 As shown in the above introduction, we can see that:

[0378] 1) AISC can use artificial intelligence models to optimize its own performance. AISCs include iWT, iNB, etc. Of course, AISCs can also be any other network elements.

[0379] 2) AISS, used to provide available artificial intelligence models, such as base station RRM model, base station MAC scheduling model, base station uplink channel assessment model, terminal RRM model, terminal downlink channel assessment model, etc.

[0380] 3) AISP: Introducing AISP as an agent of AISS between AISC and AISS can reduce the delay of AISC in acquiring artificial intelligence models and alleviate the processing burden of AISS.

[0381] 4) DCU, used to perform data collection, which includes RAN data, etc. DCU includes iWT, iNB, etc. Of course, DCU can also be other network elements.

[0382] 5) DDU, which is used to perform data storage, perform data processing, organize data collection, storage, calculation, etc. according to the needs of AISS, and return the available model to AISS.

[0383] Among them, the data includes RAN data, analytical results, model framework, etc.; data processing includes model training, association analysis, etc.

[0384] 6) iNB, responsible for wireless access processing, and may also have certain computing and storage capabilities.

[0385] 7)iWT: intelligent wireless terminal.

[0386] Figure 6 All network elements in the iRAN architecture shown belong to iRAN, among which iNB and iWT are the wireless part of iRAN, and AISP, AISS, and DDU are the AI ​​part of iRAN. iNB, iWT, AISP, AISS, and DDU are logically independent of each other.

[0387] It should be noted that AISC and DCU are additional features of iRAN network elements: AISC means that the corresponding iRAN network element (such as iNB or iWT) can request an artificial intelligence model from AISS, and DCU means that the corresponding iRAN network element can provide data collection to DDU.

[0388] Below, we introduce Figure 6 The interface of the radio access network is shown.

[0389] Please combine Figure 6 When the network architecture of the wireless access network is a reference point architecture, the wireless access network includes the following interfaces: an A1 interface, an A2 interface, a D1 interface, and an air interface.

[0390] Among them, the A1 interface is the interface between the wireless network element and the AISS and the interface between the AISP and the AISS. For example, the interface between the AISC (such as iNB) and the AISS is the A1 interface, the interface between the AISC and the AISP is the A1 interface, and the interface between the AISP and the AISS is the A1 interface. The A1 interface is used for requesting and obtaining artificial intelligence models.

[0391] The A2 interface is the interface between AISS and DDU. The A2 interface is used by AISS to trigger DDU to collect and store data, and train artificial intelligence models to obtain the trained artificial intelligence model.

[0392] The D1 interface is the interface between the DDU and the wireless network element. When the wireless network element has the DCU function, the D1 interface is used by the DDU to trigger the DCU to collect and report data, and by the DCU to report the collected data to the DDU.

[0393] The air interface is the interface between iWT and iNB. The air interface is used for wireless access control and for carrying air interface transmission of A1 interface and D1 interface.

[0394] exist Figure 6 Based on the network architecture (reference point architecture) of the wireless access network shown in FIG. Figure 7 , Figure 7 FIG. 4 is a schematic diagram of an architecture of a smart wireless access network with another exemplary network architecture being a reference point architecture.

[0395] like Figure 7 As shown, Figure 6 The DDU in the wireless access network shown can be further divided into DMU (Data Management Unit), DSU (Data Storage Unit) and DPU (Data Process Unit), that is, DDU includes DSU, DPU and DMU.

[0396] Figure 7 For the related introduction of network elements such as AISC, AISS, AISP, DCU, iNB and iWT shown, please refer to the related description above and will not be repeated here. DSU, DPU and DMU are introduced here.

[0397] Among them, DSU is at least used to store data collected by wireless network elements. DSU is also used to store the trained artificial intelligence model, model parameters of the artificial intelligence model and / or the inference results of the artificial intelligence model, that is, DSU is used to perform data storage. The data stored here includes RAN data, model, analytical result, model framework, etc.

[0398] DPU is used to perform data processing, such as artificial intelligence model training, association analysis, etc.

[0399] The DMU is used to trigger the DPU to perform AI model training and to trigger data transmission between the DSU and DPU.

[0400] Optionally, the DMU is also used to trigger the wireless network element to collect data, and to trigger the DSU to receive data collected by the wireless network element. In this way, the wireless network element is also used to collect data under the triggering of the DMU and send the collected data to the DSU, and the DSU is also used to receive data collected by the wireless network element under the triggering of the DMU.

[0401] That is, DMU can organize data collection, storage, calculation, etc. according to the needs of AISS. In addition, DMU is also used to return the usable artificial intelligence model obtained by DPU through artificial intelligence model training to AISS.

[0402] Below, we introduce Figure 7 The interface of the radio access network is shown.

[0403] in, Figure 7 For the related introduction of the A1 interface, A2 interface, and air interface shown, please refer to the related description above and will not be repeated here. The following introduces the related interfaces of DSU, DPU, and DMU.

[0404] like Figure 7 As shown, in the case where the DDU includes the DSU, the DPU and the DMU, ​​the above-mentioned D1 interface includes the D1-C interface and the D1-D interface.

[0405] The D1-C interface is an interface between the DMU and a wireless network element (eg, iNB). When the wireless network element has a DCU function, the D1-C interface is used by the DMU to trigger the DCU to collect and report data.

[0406] The D1-D interface is an interface between the DSU and a wireless network element (eg, iNB). When the wireless network element has a DCU function, the D1-D interface is used by the DCU to report the collected data to the DSU.

[0407] Please continue to see Figure 7 In the case where the DDU includes the DSU, the DPU and the DMU, ​​the wireless access network further includes the following interfaces: a D2 interface, a D3 interface and a D4 interface.

[0408] The D2 interface is an interface between the DMU and the DSU. The D2 interface is used by the DMU to trigger the DSU to receive data collected by the wireless network element, and by the DMU to trigger the DSU to submit the specified data to the DPU for processing.

[0409] The D3 interface is the interface between the DMU and the DPU. The D3 interface is used by the DMU to trigger the DPU to perform data processing, such as training an artificial intelligence model.

[0410] The D4 interface is the interface between DSU and DPU. The D4 interface is used for DSU to submit specified data to DPU, and DPU to transmit the trained artificial intelligence model to DSU.

[0411] In the above Figure 6 and Figure 7 Based on the reference point architecture of the wireless access network shown, the architecture design of the wireless access network can also have more varied architectures, which are introduced below.

[0412] When the network architecture of the wireless access network is a reference point architecture, the network element setting method of the wireless access network includes at least one of the following:

[0413] Cancel AISP;

[0414] Set up multiple levels of AISP;

[0415] There are multiple levels of AISS;

[0416] AISP and iNB are co-located;

[0417] In the case where the DDU includes the DSU, DPU and DMU, ​​the AISP is co-located with the DSU;

[0418] When the DDU includes the DSU, DPU and DMU, ​​the AISP is co-located with the DMU;

[0419] When the DDU includes the DSU, DPU and DMU, ​​the AISS is co-located with the DMU;

[0420] When the DDU includes the DSU, DPU and DMU, ​​the DSU and DPU are co-located;

[0421] When the DDU includes the DSU, DPU and DMU, ​​the DMU is co-located with the DPU;

[0422] When the DDU includes the DSU, DPU and DMU, ​​the DMU is co-located with the DSU;

[0423] The iNB is divided into multiple units, such as a centralized unit iNB-CU and a distributed unit iNB-DU.

[0424] Among them, "combined design" refers to combined design or combined setting.

[0425] The above introduces the network architecture of the wireless access network with a reference point architecture. The relationship between the network elements in the reference point architecture is clear. However, the connection between the network elements in the reference point architecture needs to be maintained for a long time. For services that are not used frequently or have many counterparts (such as an AISS may face thousands of AISCs), network resources may be wasted and the use is not convenient. Therefore, the embodiment of the present application proposes a network architecture of a wireless access network with a partial service-oriented architecture to avoid the above-mentioned shortcomings.

[0426] The following introduces the network architecture of the wireless access network as a partially service-oriented architecture.

[0427] See also Figure 8 , Figure 8 The diagram is a schematic diagram of an architecture of an intelligent wireless access network in which an exemplary network architecture is a partially service-oriented architecture. Figure 8 For the related introduction of the AISC, DCU, AISS, AISP and DDU network elements shown, please refer to the related description above and will not be repeated here.

[0428] like Figure 8 As shown, in the case where the network architecture of the radio access network is a partially service-oriented architecture, the radio access network further includes: an access network archive function (RAN Repository Function) network element and an access network communication proxy (RAN Communication Proxy, RAN communication proxy) network element;

[0429] Among them, the access network archive function network element is used for each service-oriented network element in the wireless access network to register in the access network archive function network element, and for each service-oriented network element in the wireless access network to discover other registered service-oriented network elements in the access network archive function network element. Service-oriented network elements include the above-mentioned wireless network elements, artificial intelligence model service network elements, etc., that is, the access network archive function network element accepts RAN function registration and provides RAN function discovery to support service-oriented RAN function communication.

[0430] The access network communication proxy network element is used to forward the message of a service-oriented network element in the wireless access network to another service-oriented network element in the wireless access network, that is, the access network communication proxy network element is used to proxy the communication between service-oriented network elements. It can receive messages from service-oriented network elements in the wireless access network, and forward the received messages from the service-oriented network elements to another suitable service-oriented network element in the wireless access network to support service-oriented RAN function communication.

[0431] In the embodiment of the present application, when the network architecture of the wireless access network is a partially service-oriented architecture, the iNB includes an intelligent base station non-real-time function (iNB Non-RT Function) and an intelligent base station real-time function (iNB RT Function).

[0432] Among them, the non-real-time functions of smart base stations are used for non-real-time processing of wireless access, such as radio access control (RAC), radio bearer control (RBC), mobility management (MM), radio measurement control (RMC), dual connectivity control (DCC), beamforming control (BC), automatic request for retransmission of data (ARQ), etc.

[0433] The real-time function of the smart base station is used for real-time processing of wireless access, such as MAC scheduling, PDSCH scheduling, uplink channel estimation, PUSCH (Physical Uplink Shared Channel) scheduling, downlink channel estimation, timing advance, power control, hybrid automatic repeat request (Hybrid Automatic Repeat reQuest, HARQ) of data, etc.

[0434] Figure 8 All network elements in the iRAN architecture shown belong to iRAN, among which iNB Non-RT Function, iNB RTFunction, and iWT are the radio part of iRAN, and AISP, AISS, and DDU are the AI ​​part of iRAN. iNB Non-RT Function, iNB RT Function, iWT, AISP, AISS, and DDU are logically independent of each other.

[0435] It should be noted that AISC and DCU are additional features of iRAN network elements: AISC means that the corresponding iRAN network element can request an artificial intelligence model from AISS, and DCU means that the corresponding iRAN network element can provide data collection to DDU. Typical AISC and DCU include iNB Non-RT Function, iNB RT Function, and iWT.

[0436] Below, we introduce Figure 8 The interface of the radio access network is shown.

[0437] Please combine Figure 8 In the case where the network architecture of the wireless access network is a partially service-oriented architecture, the wireless access network includes the following interfaces: N AISS Interface, NDDU Interface, N DCU Interface, iF interface and air interface. Among them:

[0438] N AISS The interface is used by AISS and AISP to provide services to the outside. AISS The interface is used to provide available artificial intelligence models, such as base station RRM model, base station MAC scheduling model, base station uplink channel assessment model, terminal RRM model, terminal downlink channel assessment model, etc. AISS and AISP can provide N AISS Interface, iWT, iNB Non-RT Function, iNB RTFunction can all use this N AISS Interface to obtain the services provided by AISS and AISP.

[0439] N DDU The interface is used by DDU to provide services to the outside. DDU The interface is used to receive data collected by DCU and provide data collection management, data storage, and artificial intelligence model training services. The data includes RAN data, analytical results, model framework, etc. DDU can provide N DDU Interface, AISS, iWT, iNB Non-RT Function, iNB RTFunction can all use N DDU interface.

[0440] N DCU The interface is the interface that the wireless network element provides services to the outside. DCU The interface can provide data collection services, including RAN data, etc. iWT, iNB Non-RT Function, and iNB RT Function can all provide N DCU Interface, DDU can use N DCU interface.

[0441] The iF interface is the interface between the non-real-time function of the smart base station and the real-time function of the smart base station. The iF interface is used to transmit the non-real-time processing signaling of wireless access and carry N AISS 、N DCU 、N DDU iF interface transmission.

[0442] The air interface is the interface between the iWT and the iNB RT Function. The air interface is used for wireless access control and for carrying N AISS Interface, N DDU Interface and N DCU Air interface transmission of the interface.

[0443] exist Figure 8 Based on the network architecture (partial service-oriented architecture) of the wireless access network shown in FIG. Fig. 9 , Fig. 9 The diagram is a schematic diagram of another exemplary network architecture of an intelligent wireless access network with a partial service-oriented architecture.

[0444] With the above Figure 7 The network architecture of the wireless access network of the reference point architecture shown is similar to that of Fig. 9 As shown, Figure 7 The DDU in the wireless access network shown can be further divided into DMU, ​​DSU and DPU, that is, the DDU can include DSU, DPU and DMU.

[0445] Fig. 9 For the related introduction of network elements such as AISS, AISP, DMU, ​​DSU, DPU, iNB Non-RT Function, iNB RT Function, iWT, RAN Repository Function, and RAN Communication Proxy shown in the figure, please refer to the related description above and will not be repeated here. Here we introduce the related interface implementation of DSU, DPU, and DMU under some service-oriented architecture.

[0446] Please continue to see Fig. 9 In the case where the DDU includes the DSU, DPU and DMU, ​​the wireless access network includes N AISS Interface, N DDU Interface, N DCU In addition to the interface, iF interface and air interface, the wireless access network also includes the following interfaces: N DMU Interface, N DSU Interface and N DPU interface.

[0447] About N AISS Interface, N DDU Interface, N DCU For an introduction to interfaces, iF interfaces, and air interfaces, see Figure 8 The relevant description of the embodiment shown is not repeated here. DMU Interface, N DSU Interface and N DPU interface.

[0448] N DMU Interface is D MU Interface that provides services to the outside world, N DMU The interface can provide artificial intelligence models. DMU can provide N DMU Interface, AISS can use N DMUinterface.

[0449] N DSU The interface is the interface that DSU provides services to the outside. DSU The interface is used to receive data collected by DCU (such as iNB Non-RTFunction, iNB RT Function, iWT), N DSU The interface provides data storage and access services, and data storage and access management services. The collected data includes RAN data, etc. The data that can be stored includes RAN data, model, analytical result, model framework, etc. DSU can provide N DSU Interface, iWT, iNB Non-RTFunction, iNB RT Function, DMU can all use N DSU interface.

[0450] N DPU The interface is used by the DPU to provide services to the outside. DPU The interface is used to receive training data of artificial intelligence models, provide artificial intelligence model training services, and provide artificial intelligence model training management services. The training data includes RAN data, model framework, etc. DPU can provide N DPU Interface, DSU, DMU can use N DPU interface.

[0451] Regarding DSU, DPU and DMU, ​​through N DSU Interface, N DPU Interface and N DMU The method process of interface implementation will be introduced in the following embodiment.

[0452] In the above Figure 8 and Fig. 9 Based on the partial service-oriented architecture of the wireless access network shown in the figure, the architecture design of the wireless access network can also have more varied architectures, which are introduced below.

[0453] When the network architecture of the wireless access network is a partially service-oriented architecture, the network element setting method of the wireless access network includes at least one of the following:

[0454] Cancel AISP;

[0455] Set up multiple levels of AISP;

[0456] There are multiple levels of AISS;

[0457] AISP and iNB are co-located;

[0458] The non-real-time function of the smart base station and the real-time function of the smart base station are integrated;

[0459] The non-real-time function and real-time function of the smart base station are divided into multiple units, such as a centralized unit iNB-CU and a distributed unit iNB-DU;

[0460] In the case where the DDU includes the DSU, DPU and DMU, ​​the AISP is co-located with the DSU;

[0461] When the DDU includes the DSU, DPU and DMU, ​​the AISS is co-located with the DMU;

[0462] When the DDU includes the DSU, DPU and DMU, ​​the DSU and DPU are co-located;

[0463] When the DDU includes the DSU, DPU and DMU, ​​the DMU is co-located with the DPU;

[0464] When the DDU includes the DSU, DPU and DMU, ​​the DMU is co-located with the DSU;

[0465] All or part of the AISS, AISP and DDU are arranged in the core network, for example, it may be a 6G (sixth generation mobile communication standard) core network;

[0466] All or part of the AISS, AISP, DSU, DPU and DMU are set in the core network, for example, it can be a 6G core network.

[0467] The above embodiments provide a variety of new wireless access network architecture designs for the endogenous intelligent characteristics of the wireless access network. In the wireless access network architecture provided by the embodiments of the present application, the intelligent characteristics (such as the generation, acquisition, and use of artificial intelligence models) are endogenous to the wireless access network architecture, rather than being supplemented to the existing architecture as additional characteristics; in addition, in the embodiments of the present application, the wireless network element is decoupled from the "high computing characteristics" and "high storage characteristics" to reduce the development and deployment costs; furthermore, the wireless capabilities, computing capabilities, and storage capabilities of the wireless network element can be flexibly arranged according to the needs of the scenario.

[0468] The following introduces a new process design of a wireless access network provided by an embodiment of the present application for the intrinsic intelligent characteristics of the wireless access network.

[0469] Based on the wireless access network provided in any of the above embodiments, an embodiment of the present application provides an artificial intelligence model acquisition method, which is used in the model providing network element of the wireless access network in any of the above embodiments.

[0470] See also Fig.10 , the method includes Fig.10Step 1001, step 1002 and step 1003 shown:

[0471] Step 1001: The model providing network element receives a model request message sent by an artificial intelligence model client network element in a wireless access network.

[0472] The artificial intelligence model client network element can be an AISC, which can be, for example, an iWT, an iNB or other network element. When the artificial intelligence model client network element needs to obtain an artificial intelligence model with a specific function for a specific scenario, the artificial intelligence model client network element sends a model request message to the model providing network element, and the model providing network element receives the model request message sent by the artificial intelligence model client network element in the wireless access network.

[0473] In a possible implementation, the model request message includes at least one of the following parameters: function ID, scenario ID, optional inputs of the artificial intelligence model, and restrictions of the artificial intelligence model.

[0474] Among them, the Function ID and Scenario ID together determine the minimum input parameter set and minimum output parameter set of the artificial intelligence model required by the artificial intelligence model customer network element. For example, the Function ID is moving bus connected mobility management, the Scenario ID is ZS1 DL (ZS1 bus downlink line), the minimum input parameter set of the artificial intelligence model IpS = {iWT measurement report, iWT bearer profile, target cell profile} and the minimum output parameter set OpS = {HO command} have been uniquely determined.

[0475] In addition, AISC can use optional inputs of the artificial intelligence model through the Optional Inputs declaration to obtain better model performance, such as Optional Inputs = {iWT velocity, iWT position, ZS1 roadside iNB topology}.

[0476] AISC can also use Model Limits to declare the restrictions that the required artificial intelligence model should meet, such as the size of RAM (Random Access Memory), the size of ROM (Read-Only Memory), the operating system OS model, etc., to meet the operating environment restrictions of AISC.

[0477] Step 1002: The model providing network element obtains a model response message according to the model request message.

[0478] After receiving the model request message, the model providing network element searches whether the artificial intelligence model requested by the model request message is stored locally. It can be understood that there are two situations:

[0479] 1) The model providing network element locally stores the artificial intelligence model requested by the model request message

[0480] If the model providing network element locally retrieves and obtains the first artificial intelligence model that meets the requirements of the model request message, the model providing network element uses the identifier of the first artificial intelligence model and the model file of the first artificial intelligence model (or the storage address of the model file of the first artificial intelligence model) as the model response message, or uses the model file of the first artificial intelligence model (or the storage address of the model file of the first artificial intelligence model) as the model response message, that is, the model response message includes the identifier of the first artificial intelligence model, the model file of the first artificial intelligence model or the storage address of the model file of the first artificial intelligence model.

[0481] It should be noted that, in this case, the function identifier and scenario identifier in the model request message sent by the artificial intelligence model client network element are used to jointly determine the minimum input parameter set and minimum output parameter set of the first artificial intelligence model.

[0482] In an embodiment of the present application, the identifier of the first artificial intelligence model is a combination of a function identifier and a scene identifier; or, the identifier of the first artificial intelligence model includes a function identifier, a scene identifier and other information.

[0483] When the identifier of the first artificial intelligence model is a combination of a function identifier and a scenario identifier, the identifier of the first artificial intelligence model can be omitted in the model response message, that is, in this case, the model response message includes the model file of the first artificial intelligence model (or the storage address of the model file of the first artificial intelligence model); when the identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information (such as the manufacturer identifier, Optional Inputs, etc.), the identifier of the first artificial intelligence model cannot be omitted in the model response message, that is, in this case, the model response message includes the identifier of the first artificial intelligence model, the model file of the first artificial intelligence model (or the storage address of the model file of the first artificial intelligence model).

[0484] In an embodiment of the present application, the storage address of the model file of the first artificial intelligence model can be in the form of URI (Uniform Resource Identifier), or in the form of IPv4 (Internet Protocol Version 4) + URI or IPv6 (Internet Protocol Version 6) + URI, etc.

[0485] Among them, the model providing network element can be an AISS or an AISP that stores the artificial intelligence model requested by the model request message.

[0486] 2) The model providing network element does not store the artificial intelligence model requested by the model request message locally

[0487] If the model providing network element does not locally store the artificial intelligence model requested by the model request message, the model providing network element further inquires whether the DDU has the artificial intelligence model requested by the model request message.

[0488] If the model providing network element fails to obtain the first artificial intelligence model through local search, the model providing network element sends a model acquisition request message to the DDU in the wireless access network, and the DDU receives the model acquisition request message sent by the model providing network element in the wireless access network.

[0489] The DDU obtains a model acquisition response message based on the model acquisition request message, and sends the model acquisition response message to the model providing network element.

[0490] Among them, the model acquisition request message sent by the model providing network element to the DDU includes at least one of the following parameters: function identifier, scenario identifier, optional input of the artificial intelligence model, and restriction conditions of the artificial intelligence model. The function identifier, scenario identifier, optional input of the artificial intelligence model, and restriction conditions of the artificial intelligence model will be introduced in the following embodiments.

[0491] If the DDU retrieves a second artificial intelligence model that meets the requirements of the model acquisition request message, the DDU sends the identifier of the second artificial intelligence model and the model file of the second artificial intelligence model (or the storage address of the model file of the second artificial intelligence model) as a model acquisition response message to the model providing network element, that is, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model or the storage address of the model file of the second artificial intelligence model.

[0492] In this way, the model providing network element receives the model acquisition response message sent by the DDU based on the model acquisition request message. The model acquisition response message is used for the model providing network element to obtain the model response message according to the model acquisition response message. For example, the model providing network element can extract the content in the model acquisition response message and add it to the model response message. The model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

[0493] In an embodiment of the present application, similar to the above model request message, the function identifier and scenario identifier in the model acquisition request message sent by the model providing network element to the DDU are used to jointly determine the minimum input parameter set and minimum output parameter set of the second artificial intelligence model.

[0494] In an embodiment of the present application, the identifier of the second artificial intelligence model is a combination of a function identifier and a scene identifier; or, the identifier of the second artificial intelligence model includes a function identifier, a scene identifier and other information.

[0495] Similar to the identifier of the first artificial intelligence model, when the identifier of the second artificial intelligence model is a combination of a function identifier and a scene identifier, the identifier of the second artificial intelligence model can be omitted in the model response message, that is, in this case, the model response message includes the model file of the second artificial intelligence model (or the storage address of the model file of the second artificial intelligence model); when the identifier of the second artificial intelligence model includes a function identifier, a scene identifier and other information, the identifier of the second artificial intelligence model cannot be omitted in the model response message, that is, in this case, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model (or the storage address of the model file of the second artificial intelligence model).

[0496] Step 1003: The model providing network element sends the model response message to the artificial intelligence model client network element.

[0497] In this way, when the model providing network element locally stores the artificial intelligence model requested by the model request message, the model providing network element directly obtains the model response message and sends it to the artificial intelligence model client network element; when the model providing network element does not locally store the artificial intelligence model requested by the model request message, the model providing network element requests the DDU to obtain the model acquisition response message, the model providing network element obtains the model response message based on the model acquisition response message, and sends the model response message to the artificial intelligence model client network element.

[0498] The artificial intelligence model client network element can obtain the artificial intelligence model it needs based on the model response message.

[0499] The network element provided by the above model can be AISS or AISP.

[0500] It should be noted that in the above, when the model providing network element does not locally store the artificial intelligence model requested by the model request message, the explanation is based on the example that the artificial intelligence model requested by the model request message sent by the artificial intelligence model client network element has been trained, and the DDU locally stores an artificial intelligence model that meets the request of the model request message. The situation where the artificial intelligence model requested by the model request message sent by the artificial intelligence model client network element has not been trained will be explained in the following embodiments.

[0501] Hereinafter, the implementation methods of the above-mentioned embodiments of the present application will be described again with reference to illustrations and examples.

[0502] like Fig.11 As shown, Fig.11 A flowchart of an exemplary artificial intelligence model acquisition method.

[0503] Fig.11 In the example, the model providing network element is AISS, and the artificial intelligence model acquisition method includes the following steps:

[0504] 1. AISC sends a model request message Model REQ to AISS.

[0505] Among them, Model REQ includes function ID, scenario ID, optional inputs of the artificial intelligence model, and Model Limits of the artificial intelligence model.

[0506] 2. After receiving the Model REQ, AISS performs a model search locally and retrieves the first artificial intelligence model requested by the Model REQ stored locally.

[0507] 3. AISS sends a model response message Model RESP to AISC.

[0508] Among them, Model RESP includes the identifier Model ID of the first artificial intelligence model, the model file Model File of the first artificial intelligence model (or the storage address Model File addr. of the model file of the first artificial intelligence model).

[0509] like Fig.12 As shown, Fig.12 A flowchart of another exemplary artificial intelligence model acquisition method.

[0510] Fig.12 In the method for obtaining the artificial intelligence model, the method comprises the following steps:

[0511] 1. AISC sends a model request message Model REQ to AISS.

[0512] Among them, Model REQ includes function ID, scenario ID, optional inputs of the artificial intelligence model, and Model Limits of the artificial intelligence model.

[0513] 2. After receiving the Model REQ, AISS performs a model search locally and retrieves the artificial intelligence model requested by the Model REQ that is not stored locally.

[0514] 3. AISS sends a model fetch request message Model Fetch REQ to DDU.

[0515] Model Fetch REQ includes function ID, scenario ID, optional inputs of the artificial intelligence model, and Model Limits of the artificial intelligence model.

[0516] 4. The DDU retrieves locally and obtains the second artificial intelligence model that meets the requirements of the model acquisition request message. The DDU sends a model acquisition response message Model Fetch RESP to the AISS.

[0517] Model Fetch RESP includes the identifier Model ID of the second artificial intelligence model and the model file Model File of the second artificial intelligence model (or the storage address Model File addr. of the model file of the second artificial intelligence model).

[0518] 5. AISS sends a model response message Model RESP to AISC.

[0519] Among them, Model RESP includes the identification Model ID of the second artificial intelligence model, the model file Model File of the second artificial intelligence model (or the storage address Model File addr. of the model file of the second artificial intelligence model).

[0520] It should be noted that the above-mentioned first artificial intelligence model and second artificial intelligence model are both artificial intelligence models that satisfy the model request message Model REQ sent by AISC to AISS. If AISS performs Model Search locally and retrieves the artificial intelligence model requested by the Model REQ stored locally, then the artificial intelligence model is called the first artificial intelligence model. If DDU performs Model Search locally and retrieves the artificial intelligence model requested by the Model REQ stored locally, then the artificial intelligence model is called the second artificial intelligence model, for the purpose of distinguishing them.

[0521] The implementation of the above-mentioned embodiment of the present application is illustrated below by taking an example.

[0522] Assume that AISC is iNB, the model provider is AISS (of course it can also be AISP), iNB needs to obtain the PDSCH scheduling model from 7 pm to 11 pm of the YLZC cell, there is no requirement for Optional Inputs, Model Limits are that the RAM occupied is not more than 621M bytes, the ROM occupied is not more than 8M bytes, and the operating system is MC 2.11.

[0523] See also Fig.13 In a possible implementation, the artificial intelligence model acquisition method includes the following steps:

[0524] 1. AISC sends a model request message Model REQ to AISS.

[0525] Among them, Model REQ includes function identifier Function ID = {PDSCH Schedule}, scenario identifier Scenario ID = {YLZC at 1900-2300CST}, optional inputs of the artificial intelligence model Optional Inputs = NULL, and restrictions of the artificial intelligence model Model Limits = {Occupied RAM≤621MB, Occupied ROM≤8MB, OS = MC 2.11}.

[0526] 2. After receiving the Model REQ, AISS performs a model search locally and retrieves the first artificial intelligence model requested by the Model REQ stored locally.

[0527] 3. AISS sends a model response message Model RESP to AISC.

[0528] Wherein, Model RESP includes the identifier of the first artificial intelligence model Model ID = {2852309021

[0529] 699L0213}, where 2852309 corresponds to PDSCH Schedule, 021699 corresponds to YLZC at 1900-2300CST, L0213 corresponds to occupied RAM ≤ 621MB, occupied ROM ≤ 8MB, and operating system OS = MC 2.11.

[0530] Model RESP also includes the storage address of the model file of the first artificial intelligence model, Model File addr.={www.PDSCHSchedule.server651 / YLZC-1900-2300CST / L0213}.

[0531] See also Fig.14 In a possible implementation, the artificial intelligence model acquisition method includes the following steps:

[0532] 1. AISC sends a model request message Model REQ to AISS.

[0533] Among them, Model REQ includes function identifier Function ID = {PDSCH Schedule}, scenario identifier Scenario ID = {YLZC at 1900-2300CST}, optional inputs of the artificial intelligence model Optional Inputs = NULL, and restrictions of the artificial intelligence model Model Limits = {Occupied RAM≤621MB, Occupied ROM≤8MB, OS = MC 2.11}.

[0534] 2. After receiving the Model REQ, AISS performs a model search locally and finds that the artificial intelligence model requested by the Model REQ is not stored locally.

[0535] 3. AISS sends a model fetch request message Model Fetch REQ to DDU.

[0536] Model Fetch REQ includes Function ID = {PDSCH Schedule}, Scenario ID = {YLZC at 1900-2300CST}, Optional Inputs of the AI ​​model =

[0537] NULL, AI model constraints Model Limits = {Occupied RAM ≤ 621MB, Occupied ROM ≤ 8MB, OS = MC 2.11}.

[0538] 4. The DDU retrieves locally and obtains the second artificial intelligence model that meets the requirements of the model acquisition request message. The DDU sends a model acquisition response message Model Fetch RESP to the AISS.

[0539] Model Fetch RESP includes the identifier Model ID of the second artificial intelligence model and the model file Model File of the second artificial intelligence model (or the storage address Model File addr. of the model file of the second artificial intelligence model).

[0540] Wherein, Model Fetch RESP includes the identifier of the second artificial intelligence model Model ID = {28523

[0541] 09021699L0213}, where 2852309 corresponds to PDSCH Schedule, 021699 corresponds to YLZC at 7pm to 11pm, and L0213 corresponds to occupied RAM ≤ 621MB, occupied ROM ≤ 8MB, and operating system OS = MC 2.11.

[0542]

[0543] addr.={www.PDSCHSchedule.server909 / YLZC-1900-2300CST / L0213}.

[0544] 5. AISS sends a model response message Model RESP to AISC.

[0545] Wherein, Model RESP includes the identifier of the second artificial intelligence model Model ID = {28523090216

[0546] 99L0213}、The storage address of the model file of the second artificial intelligence model Model File addr.={www.

[0547] PDSCHSchedule.server909 / YLZC-1900-2300CST / L0213}.

[0548] The following introduces the implementation method of the artificial intelligence model acquisition method when the artificial intelligence model requested by the model request message sent by the artificial intelligence model client network element has not been trained.

[0549] Based on the above embodiment, the model providing network element receives the model request message sent by the artificial intelligence model client network element. If the model providing network element fails to obtain the first artificial intelligence model that meets the requirements of the model request message through local retrieval, it sends a model acquisition request message to the DDU in the wireless access network.

[0550] In an embodiment of the present application, the model acquisition request message also includes AutoModel Generation Allowed, which is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained in the local retrieval.

[0551] Alternatively, the model acquisition request message does not include indication information allowing automatic model generation, and the model acquisition request message implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval.

[0552] That is, the model acquisition request message can also instruct the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0553] Among them, allowing the model to automatically generate indication information can have many forms of expression, such as being represented by other names; for example, the model automatically generates indication information can also be a default feature of the system and does not need to be indicated in the model acquisition request message; for example, allowing the model to automatically generate indication information can also be an optional feature of the system, and not indicating it in the model acquisition request message means that it is not requested.

[0554] In this way, after the DDU receives the model acquisition request message, if the DDU fails to retrieve the second artificial intelligence model, and the model acquisition request message instructs the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval, the DDU will automatically execute the training process of the third artificial intelligence model. Of course, if the model acquisition request message does not instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval, the DDU will not automatically execute the training process of the third artificial intelligence model, but will wait until the model providing network element sends a model generation request message to the DDU before executing the training process of the third artificial intelligence model.

[0555] The training process of the DDU executing the third artificial intelligence model will be described in the following embodiments. Here, before the DDU automatically executes the training process of the third artificial intelligence model, the model provides a possible interaction process between the network element and the DDU.

[0556] In one possible implementation, if the DDU fails to retrieve the second artificial intelligence model and the DDU automatically executes the training process of the third artificial intelligence model, the model acquisition response message sent by the DDU to the model providing network element based on the model acquisition request message includes model generating indication information Model Generating Indication, and the model generating indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0557] In this way, after the model providing network element receives the model acquisition response message, it knows that the DDU retrieval did not obtain the second artificial intelligence model, and the DDU automatically executes the training process of the third artificial intelligence model. The model providing network element continues to wait for the next response from the DDU.

[0558] In another possible implementation, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message sent by the DDU to the model providing network element based on the model acquisition request message only indicates that the DDU fails to retrieve the second artificial intelligence model, that is, the model acquisition response message does not indicate that the DDU is executing the training process of the third artificial intelligence model.

[0559] After the model providing network element receives the model acquisition response message, if the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, the model providing network element sends a model generation request message Model GenerationREQ to the DDU, and the model generation request message is used to instruct the DDU to execute the training process of the third artificial intelligence model.

[0560] In an embodiment of the present application, the model generation request message includes at least one of the following parameters: function identifier, scenario identifier, optional input of the artificial intelligence model, and restriction conditions of the artificial intelligence model.

[0561] Similar to the content included in the model request message, the function identifier and scenario identifier included in the model generation request message are used to jointly determine the minimum input parameter set and minimum output parameter set of the third artificial intelligence model.

[0562] In this way, the DDU receives the model generation request message sent by the model providing network element after receiving the model acquisition response message, and the DDU executes the training process of the third artificial intelligence model based on the model generation request message.

[0563] The following is an introduction to the training process of the DDU executing the third artificial intelligence model. The training process of the DDU executing the third artificial intelligence model includes the following steps A1 to A3:

[0564] Step A1: DDU sends a data collection request message to the data collection unit DCU.

[0565] According to the relevant requirements generated by the artificial intelligence model, the DDU sends a data collection request message DataCollection REQ to the DCU, where the data collection request message includes at least one of the following parameters: function identifier FunctionID, scenario identifier Scenario ID, optional inputs of the artificial intelligence model Optional Inputs, address DDU C-tunnel Info for receiving collected data, and number of data items to be collected Number of Collected Data Items.

[0566] Among them, Function ID, Scenario ID, and Optional Inputs in the data collection request message indicate the data that the DCU should collect, DDU C-tunnel Info indicates the address to which the data collected by the DCU should be sent, and Number of Collected Data Items indicates the number of data items that the DCU needs to collect.

[0567] In an embodiment of the present application, the DDU may send a data collection request message to multiple DCUs, that is, instruct multiple DCUs to perform data collection to speed up the data collection progress.

[0568] DCU performs data collection and can feedback a data collection response message Data Collection RESP to DDU. The data collection response message is used to indicate that DCU has started data collection. After data collection is completed, DCU can report the collected data to DDU.

[0569] Step A2: The DDU receives the data collection response message and the collected data sent by the DCU based on the data collection request message.

[0570] Among them, the data collection response message may include at least one of the following parameters: function identification, scenario identification, and optional input of the artificial intelligence model.

[0571] Step A3: DDU trains the third artificial intelligence model based on the data sent by DCU.

[0572] The DDU trains the third artificial intelligence model based on the data sent by the DCU. After the model training is successful, the DDU stores the trained third artificial intelligence model.

[0573] Of course, when DDU trains the third artificial intelligence model based on the data sent by DCU, model training may fail.

[0574] Furthermore, after executing the training process of the third artificial intelligence model, the DDU sends a model generation response message to the model providing network element.

[0575] It can be understood that the model generation response message Model Generation RESP may indicate that the model training is successful or that the model training fails.

[0576] In an embodiment of the present application, if the DDU successfully trains to obtain the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model; if the DDU fails to successfully train to obtain the third artificial intelligence model, the model generation response message includes model generation failure indication information.

[0577] The model providing network element receives the model generation response message sent by DDU after executing the training process of the third artificial intelligence model. Then, the model providing network element obtains the model response message according to the model generation response message, and the model providing network element sends the model response message to the artificial intelligence model client network element.

[0578] As mentioned above, there are two situations when the model generates a response message:

[0579] 1) If DDU successfully trains the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. In this case, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0580] The identifier of the third artificial intelligence model is a combination of a function identifier and a scene identifier; or, the identifier of the third artificial intelligence model includes a function identifier, a scene identifier and other information.

[0581] Similar to the above, when the identifier of the third artificial intelligence model is a combination of a function identifier and a scenario identifier, the identifier of the third artificial intelligence model can be omitted in the model response message, that is, at this time the model response message includes the model file of the third artificial intelligence model (or the storage address of the model file of the third artificial intelligence model); when the identifier of the third artificial intelligence model includes a function identifier, a scenario identifier and other information, the identifier of the third artificial intelligence model cannot be omitted in the model response message, that is, at this time the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model (or the storage address of the model file of the third artificial intelligence model).

[0582] In an embodiment of the present application, the storage address of the model file of the third artificial intelligence model can be in the form of URI (Uniform Resource Identifier), or in the form of IPv4 (Internet Protocol Version 4) + URI or IPv6 (Internet Protocol Version 6) + URI, etc.

[0583] 2) If the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information. In this case, the model response message includes no available model indication information.

[0584] Below, with reference to diagrams and examples, the implementation method of the above-mentioned embodiment of the present application is explained again in the case where the artificial intelligence model requested by the model request message sent by the artificial intelligence model client network element has not been trained.

[0585] like Fig.15 As shown, Fig.15 The flowchart of an exemplary artificial intelligence model acquisition method is shown in FIG. 1 , which is illustrated by taking the model providing network element as AISS as an example. The artificial intelligence model acquisition method includes the following steps:

[0586] 1. AISC sends a model request message Model REQ to AISS.

[0587] Among them, Model REQ includes function ID, scenario ID, optional inputs of the artificial intelligence model, and Model Limits of the artificial intelligence model.

[0588] 2. After receiving the Model REQ, AISS performs a model search locally and retrieves the artificial intelligence model requested by the Model REQ that is not stored locally.

[0589] 3. AISS sends a model fetch request message Model Fetch REQ to DDU.

[0590] Model Fetch REQ includes function ID, scenario ID, optional inputs of the artificial intelligence model, Model Limits, and Auto Model Generation Allowed.

[0591] 4. DDU sends a model fetch response message Model Fetch RESP to AISS.

[0592] Model Fetch RESP includes model generation indication information Model Generating Indication, and the model generation indication information is used to indicate that the DDU failed to retrieve the second artificial intelligence model, and that the DDU is executing the training process of the third artificial intelligence model.

[0593] Optionally, Model Fetch RESP only instructs the DDU to retrieve the second artificial intelligence model.

[0594] 5. Optionally, when Model Fetch RESP only instructs the DDU to retrieve but fails to obtain the second artificial intelligence model, the AISS sends a model generation request message Model Generation REQ to the DDU.

[0595] The model generation request message Model Generation REQ includes a function identifier Function ID, a scenario identifier Scenario ID, optional inputs of the artificial intelligence model Optional Inputs, and restrictions ModelLimits of the artificial intelligence model.

[0596] 6. DDU sends a data collection request message Data Collection REQ to DCU.

[0597] Data Collection REQ includes the function ID, the scenario ID, the optional inputs of the artificial intelligence model, the address DDU C-tunnel Info for receiving the collected data, and the number of data items to be collected.

[0598] 7. The DCU feeds back a data collection response message Data Collection RESP to the DDU.

[0599] Data Collection RESP includes function ID, scenario ID, and optional inputs of the artificial intelligence model.

[0600] 8. DCU feeds back the collected data to DDU.

[0601] 9. DDU trains the third AI model based on the data sent by DCU.

[0602] 10. After the model training is successful, DDU stores the trained third artificial intelligence model.

[0603] Optionally, model training may also fail.

[0604] 11. DDU sends a model generation response message Model Generation RESP to AISS.

[0605] 12. AISS sends a model response message Model RESP to AISP.

[0606] Step 11 and step 12 are divided into two situations:

[0607] If the model training is successful:

[0608] Then step 11 includes step 11A: DDU sends a model generation response message Model Generation RESP to AISS, where the Model Generation RESP includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model; step 12 includes step 12A: AISS sends a model response message Model RESP to AISP, where the Model RESP includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0609] If model training fails:

[0610] Then step 11 includes step 11B: DDU sends a model generation response message Model Generation RESP to AISS, where the Model Generation RESP includes model generation failure indication information; step 12 includes step 12B: AISS sends a model response message Model RESP to AISP, where the Model RESP includes no model available indication information.

[0611] Below, through an example, Fig.15 The embodiments are illustrated by way of example.

[0612] Assume that AISC is iWT, and it is necessary to obtain the downlink detection and estimation model of PDSCH from 7 pm to 11 pm in the YLZC cell. There is no requirement for Optional Inputs. Model Limits are that the RAM occupied is no more than 13M bytes, the ROM occupied is no more than 1Mbyte, and the operating system is JN 2.14.

[0613] See also Fig.16 In a possible implementation, the artificial intelligence model acquisition method includes the following steps:

[0614] 1. AISC sends a model request message Model REQ to AISS.

[0615] Among them, Model REQ includes function identifier Function ID = {PDSCH Detection&Evaluation}, scenario identifier Scenario ID = {YLZC at1900-2300CST}, optional input of the artificial intelligence model OptionalInputs = NULL, and restrictions of the artificial intelligence model Model Limits = {Occupied RAM≤13MB (occupied RAM is not more than 13M byte), Occupied ROM≤1MB (occupied ROM is not more than 1M byte), OS = JN 2.14 (operating system is JN 2.14)}.

[0616] 2. After receiving the Model REQ, AISS performs a model search locally and retrieves the artificial intelligence model requested by the Model REQ that is not stored locally.

[0617] 3. AISS sends a model fetch request message Model Fetch REQ to DDU.

[0618] Model Fetch REQ includes function identifier Function ID = {PDSCH Detection&Evaluation}, scenario identifier Scenario ID = {YLZC at 1900-2300CST (YLZC cell 7pm to 11pm)}, optional inputs of the artificial intelligence model Optional Inputs = NULL, restrictions of the artificial intelligence model Model Limits = {Occupied RAM≤13MB, Occupied ROM≤1MB, OS = JN 2.14} and indication information Auto Model Generation Allowed allowing automatic model generation.

[0619] 4. DDU sends a model fetch response message Model Fetch RESP to AISS.

[0620] Model Fetch RESP includes model generation indication information Model Generating Indication, and the model generation indication information is used to indicate that the DDU failed to retrieve the second artificial intelligence model, and that the DDU is executing the training process of the third artificial intelligence model.

[0621] 5. DDU sends a data collection request message Data Collection REQ to DCU.

[0622] Data Collection REQ includes function identifier Function ID = {PDSCH Detection&Evaluation}, scenario identifier Scenario ID = {YLZC at1900-2300CST}, artificial intelligence model restriction conditions Model Limits = {Occupied RAM≤13MB, Occupied ROM≤1MB, OS = JN 2.14}, address DDU C-tunnel Info for receiving collected data = {IPv4 = 1.2.3.4, SCTP port = 432}, and number of data to be collected Number of Collected Data Items = {5000}.

[0623] 6. The DCU feeds back a data collection response message Data Collection RESP to the DDU.

[0624] Data Collection RESP includes a function ID = {PDSCH Detection & Evaluation} and a scenario ID = {YLZC at 1900-2300CST}.

[0625] 7. DCU feeds back the collected data to DDU.

[0626] 8. DDU trains the third AI model based on the data sent by DCU.

[0627] 9. After the model training is successful, DDU stores the trained third artificial intelligence model.

[0628] 10. DDU sends a model generation response message Model Generation RESP to AISS.

[0629] Model Generation RESP includes the identifier of the third artificial intelligence model Model ID = {2852317021699L0715}, the storage address of the model file of the third artificial intelligence model Model File addr. = {www.PDSCHDetectionEvaluation.server982 / YLZC-1900-2300CST / L0

[0630] 715}.

[0631] Among them, in the Model ID, 2852317 corresponds to PDSCH Detection&Evaluation, 021699 corresponds to YLZC at 1900-2300CST, L0715 corresponds to Occupied RAM≤13MB, Occupied ROM≤1MB, and OS=JN2.14.

[0632] 11. AISS sends a model response message Model RESP to AISP.

[0633] Model RESP includes the identifier of the third artificial intelligence model Model ID = {2852317021699L07

[0634] 15}. The storage address of the model file of the third artificial intelligence model Model File addr.={www.PDSCHDetectionEvaluation.server908 / YLZC-1900-2300CST / L0715}.

[0635] In one embodiment, an artificial intelligence model acquisition method is provided, which is used in the DDU of the wireless access network in any of the above embodiments, such as Fig.17 As shown, the method includes:

[0636] Step 1701: The DDU receives a model acquisition request message sent by a model providing network element of a wireless access network.

[0637] Step 1702: The DDU sends a model acquisition response message to the model providing network element based on the model acquisition request message.

[0638] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0639] Function identification; scenario identification; optional inputs to the AI ​​model; constraints on the AI ​​model.

[0640] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes an identifier of the second artificial intelligence model, a model file of the second artificial intelligence model, or a storage address of the model file of the second artificial intelligence model.

[0641] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0642] In one of the embodiments, the identifier of the second artificial intelligence model is a combination of a function identifier and a scene identifier; or, the identifier of the second artificial intelligence model includes a function identifier, a scene identifier and other information.

[0643] In one of the embodiments, the model acquisition request message also includes indication information allowing automatic model generation, and the indication information allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval; or, the model acquisition request message does not include indication information allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval.

[0644] In one of the embodiments, the method further includes: if the DDU fails to retrieve the second artificial intelligence model, the DDU automatically executes the training process of the third artificial intelligence model.

[0645] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0646] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message indicates that the DDU fails to retrieve the second artificial intelligence model.

[0647] In one of the embodiments, the method also includes: the DDU receives a model generation request message sent by the model providing network element after receiving a model acquisition response message; and the DDU executes a training process of the third artificial intelligence model based on the model generation request message.

[0648] In one embodiment, the model generation request message includes at least one of the following parameters:

[0649] Function identification; scenario identification; optional inputs to the AI ​​model; constraints on the AI ​​model.

[0650] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of a third artificial intelligence model.

[0651] In one embodiment, the DDU performs a training process of the third artificial intelligence model, including:

[0652] The DDU sends a data acquisition request message to the data acquisition unit DCU; the DDU receives the data acquisition response message and the collected data sent by the DCU based on the data acquisition request message; the DDU trains the third artificial intelligence model based on the data sent by the DCU.

[0653] In one embodiment, the data collection request message includes at least one of the following parameters:

[0654] Function identification; scenario identification; optional inputs to the AI ​​model; address for receiving collected data;

[0655] The amount of data to be collected.

[0656] In one embodiment, the data collection response message includes at least one of the following parameters:

[0657] Function identification; scenario identification; optional input to the artificial intelligence model.

[0658] In one embodiment, the method further comprises:

[0659] After executing the training process of the third artificial intelligence model, the DDU sends a model generation response message to the model providing network element.

[0660] In one of the embodiments, if the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0661] In one of the embodiments, the identifier of the third artificial intelligence model is a combination of a function identifier and a scene identifier; or, the identifier of the third artificial intelligence model includes a function identifier, a scene identifier and other information.

[0662] In one of the embodiments, if the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

[0663] For the specific limitations on the artificial intelligence model acquisition method for DDU, please refer to the limitations on the artificial intelligence model acquisition method for model providing network elements mentioned above, which will not be repeated here.

[0664] It should be understood that, although the various steps in the above flowchart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above flowchart may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0665] In one embodiment, Fig.18 As shown, an artificial intelligence model acquisition device is provided, which is used in a model providing network element of a wireless access network in any of the above embodiments, and the device includes:

[0666] The receiving unit 1801 is used to receive a model request message sent by the artificial intelligence model client network element in the wireless access network;

[0667] An acquiring unit 1802 is configured to acquire a model response message according to the model request message;

[0668] The sending unit 1803 is used to send the model response message to the artificial intelligence model client network element.

[0669] In one embodiment, the model request message includes at least one of the following parameters:

[0670] Functional identification;

[0671] Scene identification;

[0672] Optional inputs to AI models;

[0673] Constraints of AI models.

[0674] In one of the embodiments, if the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes an identifier of the first artificial intelligence model, a model file of the first artificial intelligence model, or a storage address of the model file of the first artificial intelligence model.

[0675] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

[0676] In one embodiment, the identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0677] The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0678] In one of the embodiments, the sending unit 1803 is further configured to send a model acquisition request message to the DDU in the radio access network if the model providing network element fails to obtain the first artificial intelligence model through a local search;

[0679] The receiving unit 1801 is further used to receive a model acquisition response message sent by the DDU based on the model acquisition request message, and the model acquisition response message is used for the model providing network element to obtain the model response message according to the model acquisition response message.

[0680] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0681] Functional identification;

[0682] Scene identification;

[0683] Optional inputs to AI models;

[0684] Constraints of AI models.

[0685] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

[0686] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0687] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0688] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0689] In one of the embodiments, the model acquisition request message further includes indication information of allowing automatic model generation, and the indication information of allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0690] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0691] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0692] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0693] In one of the embodiments, the sending unit 1803 is also used to send a model generation request message to the DDU if the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, and the model generation request message is used to instruct the DDU to execute the training process of the third artificial intelligence model.

[0694] In one embodiment, the model generation request message includes at least one of the following parameters:

[0695] Functional identification;

[0696] Scene identification;

[0697] Optional inputs to AI models;

[0698] Constraints of AI models.

[0699] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0700] In one of the embodiments, the receiving unit 1801 is further used to receive a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model;

[0701] The acquiring unit 1802 is further configured to acquire the model response message according to the model generation response message.

[0702] In one of the embodiments, if the DDU successfully trains to obtain the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0703] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0704] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0705] In one of the embodiments, if the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no available model indication information.

[0706] For the specific limitations of the artificial intelligence model acquisition device for the model providing network element, please refer to the limitations of the artificial intelligence model acquisition method for the model providing network element above, which will not be repeated here. Each module in the above-mentioned artificial intelligence model acquisition device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the model providing network element in the form of hardware, or can be stored in the memory in the model providing network element in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0707] In one embodiment, Fig.19 As shown, an artificial intelligence model acquisition device is provided, which is used in a DDU of a wireless access network in any of the above embodiments, and the device includes:

[0708] The receiving unit 1901 is configured to receive a model acquisition request message sent by a model providing network element of the radio access network;

[0709] The sending unit 1902 is used to send a model acquisition response message to the model providing network element based on the model acquisition request message.

[0710] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0711] Functional identification;

[0712] Scene identification;

[0713] Optional inputs to AI models;

[0714] Constraints of AI models.

[0715] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes an identifier of the second artificial intelligence model, a model file of the second artificial intelligence model, or a storage address of the model file of the second artificial intelligence model.

[0716] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0717] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0718] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0719] In one of the embodiments, the model acquisition request message further includes indication information of allowing automatic model generation, and the indication information of allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0720] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0721] In one embodiment, the device further comprises:

[0722] The first execution unit is used to automatically execute the training process of the third artificial intelligence model if the DDU retrieval fails to obtain the second artificial intelligence model.

[0723] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0724] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0725] In one of the embodiments, the receiving unit 1901 is further configured to receive a model generation request message sent by the model providing network element after receiving the model acquisition response message;

[0726] The first execution unit is specifically used to execute the training process of the third artificial intelligence model based on the model generation request message.

[0727] In one embodiment, the model generation request message includes at least one of the following parameters:

[0728] the function identifier;

[0729] the scene identifier;

[0730] optional inputs to the artificial intelligence model;

[0731] The constraints of the artificial intelligence model.

[0732] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0733] In one of the embodiments, the first execution unit is specifically used to send a data acquisition request message to a data acquisition unit DCU; receive a data acquisition response message and collected data sent by the DCU based on the data acquisition request message; and train the third artificial intelligence model according to the data sent by the DCU.

[0734] In one embodiment, the data collection request message includes at least one of the following parameters:

[0735] the function identifier;

[0736] the scene identifier;

[0737] optional inputs to the artificial intelligence model;

[0738] The address for receiving the collected data;

[0739] The amount of data to be collected.

[0740] In one embodiment, the data collection response message includes at least one of the following parameters:

[0741] the function identifier;

[0742] the scene identifier;

[0743] Optional input to the artificial intelligence model.

[0744] In one of the embodiments, the sending unit 1902 is also used to send a model generation response message to the model providing network element after executing the training process of the third artificial intelligence model.

[0745] In one of the embodiments, if the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0746] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0747] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0748] In one of the embodiments, if the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

[0749] For the specific definition of the artificial intelligence model acquisition device for DDU, please refer to the definition of the artificial intelligence model acquisition method for DDU above, which will not be repeated here. Each module in the above-mentioned artificial intelligence model acquisition device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the DDU in hardware form, or stored in the memory in the DDU in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0750] Fig. 20 The model provided in the embodiment of the present application provides a schematic diagram of the structure of a network element. The network element provided by the model may include a processor 2000, a transceiver 2010 and a memory 2020. The transceiver 2010 is used to receive and send data under the control of the processor 2000.

[0751] Among them, Fig. 20 In the embodiment of the present invention, the bus architecture may include any number of interconnected buses and bridges, specifically linking together various circuits of one or more processors represented by processor 2000 and memory represented by memory 2020. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface.

[0752] The transceiver 2010 may be a plurality of components, i.e., include a transmitter and a receiver, and provide a unit for communicating with various other devices on a transmission medium, such as a wireless channel, a wired channel, an optical cable, etc. The processor 2000 is responsible for managing the bus architecture and general processing, and the memory 2020 may store data used by the processor 2000 when performing operations.

[0753] The processor 2000 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD). The processor 2000 may also adopt a multi-core architecture.

[0754] The processor 2000 calls the program stored in the memory 2020 to execute the following steps according to the obtained executable instructions:

[0755] Control the transceiver 2010 to receive a model request message sent by an artificial intelligence model client network element in the wireless access network;

[0756] Obtain a model response message according to the model request message;

[0757] Control the transceiver 2010 to send the model response message to the artificial intelligence model client network element.

[0758] In one embodiment, the model request message includes at least one of the following parameters:

[0759] Functional identification;

[0760] Scene identification;

[0761] Optional inputs to AI models;

[0762] Constraints of AI models.

[0763] In one of the embodiments, if the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes an identifier of the first artificial intelligence model, a model file of the first artificial intelligence model, or a storage address of the model file of the first artificial intelligence model.

[0764] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

[0765] In one embodiment, the identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0766] The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0767] In one embodiment, the processor 2000 is further configured to perform the following operations:

[0768] If the model providing network element local search fails to obtain the first artificial intelligence model, controlling the transceiver 2010 to send a model acquisition request message to the DDU in the radio access network;

[0769] The transceiver 2010 is controlled to receive the model acquisition response message sent by the DDU based on the model acquisition request message, wherein the model acquisition response message is used for the model providing network element to obtain the model response message according to the model acquisition response message.

[0770] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0771] Functional identification;

[0772] Scene identification;

[0773] Optional inputs to AI models;

[0774] Constraints of AI models.

[0775] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

[0776] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0777] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0778] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0779] In one of the embodiments, the model acquisition request message further includes indication information of allowing automatic model generation, and the indication information of allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0780] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0781] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0782] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0783] In one embodiment, the processor 2000 is further configured to perform the following operations:

[0784] If the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, the transceiver 2010 is controlled to send a model generation request message to the DDU, and the model generation request message is used to instruct the DDU to execute the training process of the third artificial intelligence model.

[0785] In one embodiment, the model generation request message includes at least one of the following parameters:

[0786] Functional identification;

[0787] Scene identification;

[0788] Optional inputs to AI models;

[0789] Constraints of AI models.

[0790] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0791] In one embodiment, the processor 2000 is further configured to perform the following operations:

[0792] Control the transceiver 2010 to receive a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model;

[0793] The model response message is obtained according to the model generation response message.

[0794] In one of the embodiments, if the DDU successfully trains to obtain the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0795] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0796] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0797] In one of the embodiments, if the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no available model indication information.

[0798] Fig.21 The structure diagram of the data dedicated unit provided in the embodiment of the present application is as follows. The data dedicated unit may include a processor 2100, a transceiver 2110 and a memory 2120. The transceiver 2110 is used to receive and send data under the control of the processor 2100.

[0799] Among them, Fig.21 In the embodiment of the present invention, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 2100 and memory represented by memory 2120. The bus architecture may also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface.

[0800] The transceiver 2110 may be a plurality of components, i.e., a transmitter and a receiver, and provides a unit for communicating with various other devices on a transmission medium, such as a wireless channel, a wired channel, an optical cable, etc. The processor 2100 is responsible for managing the bus architecture and general processing, and the memory 2120 may store data used by the processor 2100 when performing operations.

[0801] The processor 2100 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD). The processor 2100 may also adopt a multi-core architecture.

[0802] The processor 2100 calls the program stored in the memory 2120 to execute the following steps according to the obtained executable instructions:

[0803] Control the transceiver 2110 to receive a model acquisition request message sent by a model providing network element of the radio access network;

[0804] The transceiver 2110 is controlled to send a model acquisition response message to the model providing network element based on the model acquisition request message.

[0805] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0806] Functional identification;

[0807] Scene identification;

[0808] Optional inputs to AI models;

[0809] Constraints of AI models.

[0810] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes an identifier of the second artificial intelligence model, a model file of the second artificial intelligence model, or a storage address of the model file of the second artificial intelligence model.

[0811] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0812] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0813] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0814] In one of the embodiments, the model acquisition request message further includes indication information for allowing automatic model generation, and the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0815] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0816] In one embodiment, the processor 2100 is further configured to perform the following operations:

[0817] If the DDU retrieval fails to obtain the second artificial intelligence model, the training process of the third artificial intelligence model is automatically executed.

[0818] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0819] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0820] In one embodiment, the processor 2100 is further configured to perform the following operations:

[0821] Controlling the transceiver 2110 to receive the model generation request message sent by the model providing network element after receiving the model acquisition response message;

[0822] A training process of a third artificial intelligence model is executed based on the model generation request message.

[0823] In one embodiment, the model generation request message includes at least one of the following parameters:

[0824] the function identifier;

[0825] the scene identifier;

[0826] optional inputs to the artificial intelligence model;

[0827] The constraints of the artificial intelligence model.

[0828] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0829] In one embodiment, the processor 2100 is specifically configured to perform the following operations:

[0830] Controlling the transceiver 2110 to send a data acquisition request message to a data acquisition unit DCU;

[0831] Control the transceiver 2110 to receive the data acquisition response message and the collected data sent by the DCU based on the data acquisition request message;

[0832] The third artificial intelligence model is trained according to the data sent by the DCU.

[0833] In one embodiment, the data collection request message includes at least one of the following parameters:

[0834] the function identifier;

[0835] the scene identifier;

[0836] optional inputs to the artificial intelligence model;

[0837] The address for receiving the collected data;

[0838] The amount of data to be collected.

[0839] In one embodiment, the data collection response message includes at least one of the following parameters:

[0840] the function identifier;

[0841] the scene identifier;

[0842] Optional input to the artificial intelligence model.

[0843] In one embodiment, the processor 2100 is further configured to perform the following operations:

[0844] After executing the training process of the third artificial intelligence model, the transceiver 2110 is controlled to send a model generation response message to the model providing network element.

[0845] In one of the embodiments, if the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0846] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0847] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0848] In one of the embodiments, if the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

[0849] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0850] Receiving a model request message sent by an artificial intelligence model client network element in the wireless access network;

[0851] Obtain a model response message according to the model request message;

[0852] The model response message is sent to the artificial intelligence model client network element.

[0853] In one embodiment, the model request message includes at least one of the following parameters:

[0854] Functional identification;

[0855] Scene identification;

[0856] Optional inputs to AI models;

[0857] Constraints of AI models.

[0858] In one of the embodiments, if the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes an identifier of the first artificial intelligence model, a model file of the first artificial intelligence model, or a storage address of the model file of the first artificial intelligence model.

[0859] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

[0860] In one embodiment, the identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0861] The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0862] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0863] If the first artificial intelligence model is not obtained by the local search of the model providing network element, a model acquisition request message is sent to the DDU in the radio access network;

[0864] A model acquisition response message is received from the DDU based on the model acquisition request message, wherein the model acquisition response message is used by the model providing network element to obtain the model response message according to the model acquisition response message.

[0865] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0866] Functional identification;

[0867] Scene identification;

[0868] Optional inputs to AI models;

[0869] Constraints of AI models.

[0870] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

[0871] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0872] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0873] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0874] In one of the embodiments, the model acquisition request message further includes indication information for allowing automatic model generation, and the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0875] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0876] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0877] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0878] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0879] If the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, a model generation request message is sent to the DDU, where the model generation request message is used to instruct the DDU to execute a training process for a third artificial intelligence model.

[0880] In one embodiment, the model generation request message includes at least one of the following parameters:

[0881] Functional identification;

[0882] Scene identification;

[0883] Optional inputs to AI models;

[0884] Constraints of AI models.

[0885] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0886] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0887] receiving a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model;

[0888] The model response message is obtained according to the model generation response message.

[0889] In one of the embodiments, if the DDU successfully trains to obtain the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0890] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0891] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0892] In one of the embodiments, if the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no available model indication information.

[0893] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0894] Receiving a model acquisition request message sent by a model providing network element of the wireless access network;

[0895] A model acquisition response message is sent to the model providing network element based on the model acquisition request message.

[0896] In one embodiment, the model acquisition request message includes at least one of the following parameters:

[0897] Functional identification;

[0898] Scene identification;

[0899] Optional inputs to AI models;

[0900] Constraints of AI models.

[0901] In one of the embodiments, if the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes an identifier of the second artificial intelligence model, a model file of the second artificial intelligence model, or a storage address of the model file of the second artificial intelligence model.

[0902] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

[0903] In one embodiment, the identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or

[0904] The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0905] In one of the embodiments, the model acquisition request message further includes indication information of allowing automatic model generation, and the indication information of allowing automatic model generation is used to instruct the DDU to automatically execute the training process of the third artificial intelligence model when the second artificial intelligence model is not obtained by local retrieval; or,

[0906] The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

[0907] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0908] If the DDU retrieval fails to obtain the second artificial intelligence model, the training process of the third artificial intelligence model is automatically executed.

[0909] In one of the embodiments, if the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

[0910] In one of the embodiments, if the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

[0911] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0912] Receiving a model generation request message sent by the model providing network element after receiving the model acquisition response message;

[0913] A training process of a third artificial intelligence model is executed based on the model generation request message.

[0914] In one embodiment, the model generation request message includes at least one of the following parameters:

[0915] the function identifier;

[0916] the scene identifier;

[0917] optional inputs to the artificial intelligence model;

[0918] The constraints of the artificial intelligence model.

[0919] In one of the embodiments, the function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

[0920] In one embodiment, when the computer program is executed by a processor, the following steps are specifically implemented:

[0921] Sending a data collection request message to the data collection unit DCU;

[0922] Receiving a data acquisition response message and collected data sent by the DCU based on the data acquisition request message;

[0923] The third artificial intelligence model is trained according to the data sent by the DCU.

[0924] In one embodiment, the data collection request message includes at least one of the following parameters:

[0925] the function identifier;

[0926] the scene identifier;

[0927] optional inputs to the artificial intelligence model;

[0928] The address for receiving the collected data;

[0929] The amount of data to be collected.

[0930] In one embodiment, the data collection response message includes at least one of the following parameters:

[0931] the function identifier;

[0932] the scene identifier;

[0933] Optional input to the artificial intelligence model.

[0934] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0935] After executing the training process of the third artificial intelligence model, a model generation response message is sent to the model providing network element.

[0936] In one of the embodiments, if the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

[0937] In one embodiment, the identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or,

[0938] The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

[0939] In one of the embodiments, if the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

[0940] Fig. 22 It is a schematic structural diagram of the chip of an embodiment of the present application. Fig. 22 The chip 2200 shown includes a processor 2210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0941] Alternatively, if Fig. 22 As shown, the chip 2200 may further include a memory 2220. The processor 2210 may call and run a computer program from the memory 2220 to implement the method in the embodiment of the present application.

[0942] The memory 2220 may be a separate device independent of the processor 2210 , or may be integrated into the processor 2210 .

[0943] Optionally, the chip 2200 may further include an input interface 2230. The processor 2210 may control the input interface 2230 to communicate with other devices or chips, and specifically, may obtain information or data sent by other devices or chips.

[0944] Optionally, the chip 2200 may further include an output interface 2240. The processor 2210 may control the output interface 2240 to communicate with other devices or chips, and specifically, may output information or data to other devices or chips.

[0945] Optionally, the chip 2200 can be applied to the network element device in the embodiment of the present application, and the chip 2200 can implement the corresponding processes implemented in the various methods in the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0946] It should be understood that the chip 2200 mentioned in the embodiment of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0947] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0948] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0949] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A wireless access network, characterized in that: The wireless access network includes: a wireless network element and an artificial intelligence model service network element, and the wireless network element includes at least an intelligent base station iNB; The artificial intelligence model service network element is used to train the artificial intelligence model according to the data collected by the wireless network element, and to provide the artificial intelligence model to the wireless network element; The wireless network element is used to send the collected data to the artificial intelligence model service network element so that the artificial intelligence model service network element can train the artificial intelligence model according to the collected data, and to perform model inference according to the artificial intelligence model provided by the artificial intelligence model service network element to obtain target communication parameters.

2. The wireless access network according to claim 1, characterized in that: The wireless network element also includes an intelligent terminal iWT, and the iWT communicates with the iNB through an air interface.

3. The wireless access network according to claim 2, characterized in that: The artificial intelligence model service network element includes a data dedicated unit DDU and a model providing network element; The DDU is used to receive and store data collected by the wireless network element, and to train an artificial intelligence model based on the data collected by the wireless network element; The model providing network element is used to provide the artificial intelligence model obtained by DDU training to the wireless network element.

4. The wireless access network according to claim 3, characterized in that: The DDU is also used to store the trained artificial intelligence model, model parameters of the artificial intelligence model and / or inference results of the artificial intelligence model.

5. The wireless access network according to claim 3, characterized in that: The DDU is also used to trigger the wireless network element to collect data; The wireless network element is further configured to collect data under the triggering of the DDU and send the collected data to the DDU.

6. The wireless access network according to claim 3, characterized in that: The DDU includes a data storage unit DSU, a data processing unit DPU and a data management unit DMU; The DSU is at least used to store data collected by the wireless network element; The DPU is used for artificial intelligence model training; The DMU is used to trigger the DPU to perform artificial intelligence model training and to trigger data transmission between the DSU and the DPU.

7. The wireless access network according to claim 6, characterized in that: The DSU is also used to store the trained artificial intelligence model, model parameters of the artificial intelligence model and / or inference results of the artificial intelligence model.

8. The wireless access network according to claim 6, characterized in that: The DMU is further used to trigger the wireless network element to collect data, and to trigger the DSU to receive the data collected by the wireless network element; The wireless network element is further used to collect data under the triggering of the DMU and send the collected data to the DSU; The DSU is further configured to receive data collected by the wireless network element under the triggering of the DMU.

9. The wireless access network according to claim 3, characterized in that: The model providing network element is used to locally store the trained artificial intelligence model and provide the artificial intelligence model to the wireless network element based on the model information of the locally stored artificial intelligence model; or, The model providing network element is used to obtain model information of the trained artificial intelligence model from the DDU, and provide the artificial intelligence model to the wireless network element based on the model information obtained from the DDU.

10. The wireless access network according to claim 9, characterized in that: The model information includes at least one of the following: Model files for AI models; The storage address of the model file of the artificial intelligence model; Identification of the artificial intelligence model.

11. The wireless access network according to claim 3, characterized in that: The model provides the network element and is also used to trigger the DDU to execute the artificial intelligence model training process when the artificial intelligence model requested by the wireless network element does not exist in the DDU.

12. The wireless access network according to claim 3 or 6, characterized in that: The model providing network element includes: an artificial intelligence service server AISS; or The AISS and the artificial intelligence service agent AISP, wherein the AISP is arranged between the AISS and the wireless network element.

13. The wireless access network according to claim 12, characterized in that: The network architecture of the wireless access network is a reference point architecture; or, the network architecture of the wireless access network is a partially service-oriented architecture.

14. The wireless access network according to claim 13, characterized in that: In the case where the network architecture of the wireless access network is a reference point architecture, the network element setting method of the wireless access network includes at least one of the following: There are multiple levels of AISP; There are multiple levels of AISS; The AISP is jointly provided with the iNB; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISP is jointly provided with the DSU; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISP is jointly provided with the DMU; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DSU and the DPU are jointly provided; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DPU; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DSU; The iNB is divided into multiple units.

15. The wireless access network according to claim 13, characterized in that: In the case where the network architecture of the wireless access network is a partially service-oriented architecture, the wireless access network further comprises: an access network profile function network element and an access network communication proxy network element; The access network archive function network element is used for each service-oriented network element in the wireless access network to register in the access network archive function network element, and for each service-oriented network element in the wireless access network to discover other registered service-oriented network elements in the access network archive function network element; The access network communication proxy network element is used to forward a message of a service-oriented network element in the wireless access network to another service-oriented network element in the wireless access network.

16. The wireless access network according to claim 15, characterized in that: In the case where the network architecture of the wireless access network is a partially service-oriented architecture, the iNB includes a smart base station non-real-time function and a smart base station real-time function; The non-real-time function of the smart base station is used to perform non-real-time processing of wireless access; The real-time function of the smart base station is used to perform real-time processing of wireless access.

17. The wireless access network according to claim 16, characterized in that: In the case where the network architecture of the wireless access network is a partially service-oriented architecture, the network element setting method of the wireless access network includes at least one of the following: There are multiple levels of AISP; There are multiple levels of AISS; The AISP is jointly provided with the iNB; The non-real-time function of the smart base station and the real-time function of the smart base station are combined; The non-real-time function of the smart base station and the real-time function of the smart base station are divided into multiple units; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISP is jointly provided with the DSU; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the AISS is jointly provided with the DMU; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DSU and the DPU are jointly provided; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DPU; In the case where the DDU includes the DSU, the DPU and the DMU, ​​the DMU is jointly provided with the DSU; All or part of the AISS, the AISP and the DDU are arranged in a core network; All or part of the AISS, the AISP, the DSU, the DPU and the DMU are arranged in the core network.

18. The wireless access network according to claim 13, characterized in that: In the case where the network architecture of the wireless access network is a reference point architecture, the wireless access network includes the following interfaces: an A1 interface, an A2 interface, a D1 interface, and an air interface; The A1 interface is an interface between the wireless network element and the AISS and an interface between the AISP and the AISS; The A2 interface is an interface between the AISS and the DDU; The D1 interface is an interface between the DDU and the wireless network element; The air interface is used for wireless access control and for carrying air interface transmission of the A1 interface and the D1 interface.

19. The wireless access network according to claim 18, characterized in that: In the case where the DDU includes the DSU, the DPU and the DMU, ​​the D1 interface includes a D1-C interface and a D1-D interface; The D1-C interface is an interface between the DMU and the wireless network element; The D1-D interface is an interface between the DSU and the wireless network element.

20. The wireless access network according to claim 18, characterized in that: In the case where the DDU includes the DSU, the DPU and the DMU, ​​the radio access network further includes the following interfaces: D2 interface, D3 interface and D4 interface; The D2 interface is an interface between the DMU and the DSU; The D3 interface is an interface between the DMU and the DPU; The D4 interface is an interface between the DSU and the DPU.

21. The wireless access network according to claim 16, characterized in that: In the case where the network architecture of the wireless access network is a partially service-oriented architecture, the wireless access network includes the following interfaces: AISS Interface, N DDU Interface, N DCU Interface, iF interface and air interface; The N AISS The interface is an interface for the AISS and the AISP to provide services externally; The N DDU The interface is the interface for the DDU to provide services externally; The N DCU The interface is an interface for the wireless network element to provide services externally; The iF interface is an interface between the non-real-time function of the smart base station and the real-time function of the smart base station; The air interface is used for wireless access control and for carrying the N AISS Interface, the N DDU interface and the N DCU Air interface transmission of the interface.

22. The wireless access network according to claim 21, characterized in that: In the case where the DDU includes the DSU, the DPU and the DMU, ​​the wireless access network further includes the following interface: DMU Interface, N DSU Interface and N DPU interface; The N DMU The interface is an interface for the DMU to provide services externally; The N DSU The interface is an interface for the DSU to provide services externally; The N DPU Interface The DPU provides services externally.

23. The wireless access network according to claim 1, characterized in that: The wireless network element is further used to optimize the corresponding communication process according to the target communication parameters.

24. A method for acquiring an artificial intelligence model, characterized in that: A model providing network element for a wireless access network as claimed in any one of claims 1 to 23, the method comprising: Receiving a model request message sent by an artificial intelligence model client network element in the wireless access network; Obtain a model response message according to the model request message; The model response message is sent to the artificial intelligence model client network element.

25. The method according to claim 24, characterized in that The model request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

26. The method according to claim 25, characterized in that If the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes the identifier of the first artificial intelligence model, the model file of the first artificial intelligence model, or the storage address of the model file of the first artificial intelligence model.

27. The method according to claim 26, characterized in that The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

28. The method according to claim 26, characterized in that The identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

29. The method according to claim 26, characterized in that The method further comprises: If the first artificial intelligence model is not obtained by the local search of the model providing network element, a model acquisition request message is sent to the DDU in the radio access network; A model acquisition response message is received from the DDU based on the model acquisition request message, wherein the model acquisition response message is used by the model providing network element to obtain the model response message according to the model acquisition response message.

30. The method according to claim 29, characterized in that The model acquisition request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

31. The method according to claim 30, characterized in that If the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

32. The method according to claim 31, characterized in that The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

33. The method according to claim 31, characterized in that The identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

34. The method according to claim 31, characterized in that The model acquisition request message further includes indication information for allowing automatic model generation, where the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute a training process of a third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval; or, The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

35. The method according to claim 34, characterized in that If the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

36. The method according to claim 31, characterized in that If the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

37. The method according to claim 36, characterized in that The method further comprises: If the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, a model generation request message is sent to the DDU, where the model generation request message is used to instruct the DDU to execute a training process for a third artificial intelligence model.

38. The method according to claim 37, characterized in that The model generation request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

39. The method according to claim 35 or 37, characterized in that The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

40. The method according to claim 35 or 37, characterized in that The method further comprises: receiving a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model; The model response message is obtained according to the model generation response message.

41. The method according to claim 40, characterized in that If the DDU successfully trains the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

42. The method according to claim 41, characterized in that The identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

43. The method according to claim 40, characterized in that If the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no model available indication information.

44. A method for acquiring an artificial intelligence model, characterized in that: A DDU for a wireless access network according to any one of claims 1 to 23, the method comprising: Receiving a model acquisition request message sent by a model providing network element of the wireless access network; A model acquisition response message is sent to the model providing network element based on the model acquisition request message.

45. The method according to claim 44, characterized in that The model acquisition request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

46. ​​The method according to claim 45, characterized in that If the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

47. The method according to claim 46, characterized in that The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

48. The method according to claim 46, characterized in that The identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

49. The method according to claim 46, characterized in that The model acquisition request message further includes indication information for allowing automatic model generation, where the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute a training process of a third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval; or, The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

50. The method according to claim 49, characterized in that The method further comprises: If the DDU retrieval fails to obtain the second artificial intelligence model, the training process of the third artificial intelligence model is automatically executed.

51. The method according to claim 50, characterized in that If the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

52. The method according to claim 46, characterized in that If the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

53. The method according to claim 52, characterized in that The method further comprises: Receiving a model generation request message sent by the model providing network element after receiving the model acquisition response message; A training process of a third artificial intelligence model is executed based on the model generation request message.

54. The method according to claim 53, characterized in that The model generation request message includes at least one of the following parameters: the function identifier; the scene identifier; optional inputs to the artificial intelligence model; The constraints of the artificial intelligence model.

55. The method according to claim 50 or 53, characterized in that The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

56. The method according to claim 50 or 53, characterized in that The training process of executing the third artificial intelligence model includes: Sending a data collection request message to the data collection unit DCU; Receiving a data acquisition response message and collected data sent by the DCU based on the data acquisition request message; The third artificial intelligence model is trained according to the data sent by the DCU.

57. The method according to claim 56, characterized in that The data collection request message includes at least one of the following parameters: the function identifier; the scene identifier; optional inputs to the artificial intelligence model; The address for receiving the collected data; The quantity of data to be collected.

58. The method according to claim 56, characterized in that The data collection response message includes at least one of the following parameters: the function identifier; the scene identifier; Optional input to the artificial intelligence model.

59. The method according to claim 50 or 53, characterized in that The method further comprises: After executing the training process of the third artificial intelligence model, a model generation response message is sent to the model providing network element.

60. The method according to claim 59, characterized in that If the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

61. The method according to claim 60, characterized in that The identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

62. The method according to claim 59, characterized in that If the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

63. An artificial intelligence model acquisition device, characterized in that: A model providing network element for a wireless access network as claimed in any one of claims 1 to 23, the device comprising: A receiving unit, configured to receive a model request message sent by an artificial intelligence model client network element in the wireless access network; An acquiring unit, configured to acquire a model response message according to the model request message; A sending unit is used to send the model response message to the artificial intelligence model client network element.

64. An artificial intelligence model acquisition device, characterized in that: A DDU for a wireless access network according to any one of claims 1 to 23, the device comprising: A receiving unit, configured to receive a model acquisition request message sent by a model providing network element of the radio access network; A sending unit is used to send a model acquisition response message to the model providing network element based on the model acquisition request message.

65. A model provides a network element, characterized in that Including memory, transceiver, processor: A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations: Controlling the transceiver to receive a model request message sent by an artificial intelligence model client network element in the wireless access network; Obtain a model response message according to the model request message; Control the transceiver to send the model response message to the artificial intelligence model client network element.

66. The model providing network element according to claim 65, characterized in that The model request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

67. The model providing network element according to claim 66, characterized in that If the model providing network element locally retrieves a first artificial intelligence model that meets the requirements of the model request message, the model response message includes the identifier of the first artificial intelligence model, the model file of the first artificial intelligence model, or the storage address of the model file of the first artificial intelligence model.

68. The model providing network element according to claim 67, characterized in that: The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the first artificial intelligence model.

69. The model providing network element according to claim 67, characterized in that: The identifier of the first artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the first artificial intelligence model includes the function identifier, the scenario identifier and other information.

70. The model providing network element according to claim 67, characterized in that: The processor is further configured to perform the following operations: If the first artificial intelligence model is not obtained by the local search of the model providing network element, controlling the transceiver to send a model acquisition request message to the DDU in the radio access network; The transceiver is controlled to receive a model acquisition response message sent by the DDU based on the model acquisition request message, wherein the model acquisition response message is used for the model providing network element to obtain the model response message according to the model acquisition response message.

71. The model providing network element according to claim 70, characterized in that: The model acquisition request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

72. The model providing network element according to claim 71, characterized in that: If the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model. Correspondingly, the model response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

73. The model providing network element according to claim 72, characterized in that: The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

74. The model providing network element according to claim 72, characterized in that: The identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

75. The model providing network element according to claim 72, characterized in that: The model acquisition request message further includes indication information for allowing automatic model generation, where the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute a training process of a third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval; or, The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

76. The model providing network element according to claim 75, characterized in that If the DDU fails to retrieve the second artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

77. The model providing network element according to claim 72, characterized in that: If the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

78. The model providing network element according to claim 77, characterized in that: The processor is further configured to perform the following operations: If the model acquisition response message indicates that the DDU has not retrieved the second artificial intelligence model, the transceiver is controlled to send a model generation request message to the DDU, and the model generation request message is used to instruct the DDU to execute the training process of the third artificial intelligence model.

79. The model providing network element according to claim 78, characterized in that The model generation request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

80. The model providing network element according to claim 76 or 78, characterized in that: The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

81. The model providing network element according to claim 76 or 78, characterized in that: The processor is further configured to perform the following operations: Controlling the transceiver to receive a model generation response message sent by the DDU after executing the training process of the third artificial intelligence model; The model response message is obtained according to the model generation response message.

82. The model providing network element according to claim 81, characterized in that: If the DDU successfully trains the third artificial intelligence model, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model. Correspondingly, the model response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

83. The model providing network element according to claim 82, characterized in that: The identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

84. The model providing network element according to claim 81, characterized in that: If the DDU fails to successfully train the third artificial intelligence model, the model generation response message includes model generation failure indication information, and correspondingly, the model response message includes no model available indication information.

85. A data dedicated unit, characterized in that Including memory, transceiver, processor: A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations: Controlling the transceiver to receive a model acquisition request message sent by a model providing network element of the wireless access network; The transceiver is controlled to send a model acquisition response message to the model providing network element based on the model acquisition request message.

86. The data dedicated unit according to claim 85, characterized in that The model acquisition request message includes at least one of the following parameters: Functional identification; Scene identification; Optional inputs to AI models; Constraints of AI models.

87. The data dedicated unit according to claim 86, characterized in that If the DDU retrieves and obtains a second artificial intelligence model that meets the requirements of the model acquisition request message, the model acquisition response message includes the identifier of the second artificial intelligence model, the model file of the second artificial intelligence model, or the storage address of the model file of the second artificial intelligence model.

88. The data dedicated unit according to claim 87, characterized in that: The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the second artificial intelligence model.

89. The data dedicated unit according to claim 87, characterized in that: The identifier of the second artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the second artificial intelligence model includes the function identifier, the scenario identifier and other information.

90. The data dedicated unit according to claim 87, characterized in that: The model acquisition request message further includes indication information for allowing automatic model generation, where the indication information for allowing automatic model generation is used to instruct the DDU to automatically execute a training process of a third artificial intelligence model when the second artificial intelligence model is not obtained through local retrieval; or, The model acquisition request message does not include the indication information for allowing automatic model generation, and implicitly instructs the DDU to automatically execute the training process of the third artificial intelligence model if the second artificial intelligence model is not obtained through local retrieval.

91. The data dedicated unit according to claim 90, characterized in that: The processor is further configured to perform the following operations: If the DDU retrieval fails to obtain the second artificial intelligence model, the training process of the third artificial intelligence model is automatically executed.

92. The data dedicated unit according to claim 91, characterized in that: If the DDU fails to retrieve the second artificial intelligence model and automatically executes the training process of the third artificial intelligence model, the model acquisition response message includes model generation indication information, and the model generation indication information is used to indicate that the DDU fails to retrieve the second artificial intelligence model and that the DDU is executing the training process of the third artificial intelligence model.

93. The data dedicated unit according to claim 87, characterized in that: If the DDU retrieval fails to obtain the second artificial intelligence model, the model acquisition response message indicates that the DDU retrieval fails to obtain the second artificial intelligence model.

94. The data dedicated unit according to claim 93, characterized in that: The processor is further configured to perform the following operations: Controlling the transceiver to receive a model generation request message sent by the model providing network element after receiving the model acquisition response message; A training process of a third artificial intelligence model is executed based on the model generation request message.

95. The data dedicated unit according to claim 94, characterized in that: The model generation request message includes at least one of the following parameters: the function identifier; the scene identifier; optional inputs to the artificial intelligence model; The constraints of the artificial intelligence model.

96. The data dedicated unit according to claim 91 or 94, characterized in that: The function identifier and the scenario identifier are used to jointly determine a minimum input parameter set and a minimum output parameter set of the third artificial intelligence model.

97. The data dedicated unit according to claim 91 or 94, characterized in that: The processor is specifically configured to perform the following operations: Controlling the transceiver to send a data acquisition request message to a data acquisition unit DCU; Controlling the transceiver to receive the data acquisition response message and the collected data sent by the DCU based on the data acquisition request message; The third artificial intelligence model is trained according to the data sent by the DCU.

98. The data dedicated unit according to claim 97, characterized in that: The data collection request message includes at least one of the following parameters: the function identifier; the scene identifier; optional inputs to the artificial intelligence model; The address for receiving the collected data; The quantity of data to be collected.

99. The data dedicated unit according to claim 97, characterized in that: The data collection response message includes at least one of the following parameters: the function identifier; the scene identifier; Optional input to the artificial intelligence model.

100. The data dedicated unit according to claim 91 or 94, characterized in that: The processor is further configured to perform the following operations: After executing the training process of the third artificial intelligence model, the transceiver is controlled to send a model generation response message to the model providing network element.

101. The data dedicated unit according to claim 100, characterized in that: If the third artificial intelligence model is successfully trained, the model generation response message includes the identifier of the third artificial intelligence model, the model file of the third artificial intelligence model, or the storage address of the model file of the third artificial intelligence model.

102. The data dedicated unit according to claim 101, characterized in that: The identifier of the third artificial intelligence model is a combination of the function identifier and the scenario identifier; or The identifier of the third artificial intelligence model includes the function identifier, the scenario identifier and other information.

103. The data dedicated unit according to claim 101, characterized in that: If the third artificial intelligence model is not successfully trained, the model generation response message includes model generation failure indication information.

104. A processor-readable storage medium, characterized in that: The processor-readable storage medium stores a program, and the program is used to enable the processor to execute the method according to any one of claims 24 to 43.

105. A processor-readable storage medium, characterized in that: The processor-readable storage medium stores a program, and the program is used to enable the processor to execute the method according to any one of claims 44 to 62.

Images