Information processing method, device and equipment

Abstract

The invention provides an information processing method, device and equipment, and the information processing method comprises the steps: decomposing an artificial intelligence (AI) task of a terminal into at least two different AI subtasks; according to the at least two different AI subtasks, sending primary service function chain arrangement information to cloud computing center equipment; receiving first information sent by the cloud computing center equipment; the first information comprises final service function chain forming decision information or confirmation information; and determining a final service function chain according to the first information. According to the scheme, the first base station and the cloud computing center equipment can be supported to determine the AI service function chain arrangement based on the global view angle (the second information), so that the arrangement effect is improved, and the problem that in the prior art, an information processing scheme for AI service function chain arrangement lacks the global view angle, so that the arrangement effect is poor is well solved.

Description

Technical Field

[0001] The present invention relates to the field of wireless technology, and in particular to an information processing method, device and equipment. Background Art

[0002] With the continuous emergence of new applications, UE (terminal) tasks often require edge networks to provide multiple AI (artificial intelligence) services nearby. An edge base station is limited by its computing, storage and model resources, and it is difficult to guarantee complete AI services. Therefore, collaboration across multiple edge base stations is required to form an AI service function chain. The training of AI models and the reasonable deployment of models, as well as the efficient and reasonable arrangement of AI service function chains are important factors in ensuring service quality.

[0003] At present, edge base stations use local optimization heuristic algorithms to make decisions when arranging AI service function chains. However, AI service function chain orchestration will connect multiple edge base stations in the network topology in series, which requires comprehensive base station resources across the entire network; and the existing service function chain orchestration based on heuristic algorithms is to obtain feasible solutions to combinatorial optimization problems through experience, lacking a global perspective, and the degree of deviation between the feasible solution and the optimal solution cannot generally be predicted.

[0004] From the above, the information processing solutions for the orchestration of AI service function chains in the prior art have problems such as lack of a global perspective resulting in poor orchestration effects. Summary of the invention

[0005] The purpose of the present invention is to provide an information processing method, device and equipment to solve the problem that the information processing solution for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0006] In order to solve the above technical problem, an embodiment of the present invention provides an information processing method, which is applied to a first base station, including:

[0007] Decompose the terminal's artificial intelligence (AI) task into at least two different AI subtasks;

[0008] Sending primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks;

[0009] Receiving first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information;

[0010] According to the first information, a final service function chain is determined.

[0011] Optionally, sending primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks includes:

[0012] Arrange the service position of each of the AI ​​subtasks according to the second information to form a primary service function chain, and obtain primary service function chain arrangement information; wherein the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; and the global status information is related to the execution of the AI ​​task;

[0013] The primary service function chain orchestration information is sent to the cloud computing center device.

[0014] Optionally, the second information includes at least one of the following:

[0015] Information about AI models stored in each base station and cloud computing center equipment;

[0016] Information about remaining computing resources in each base station and cloud computing center equipment;

[0017] Information about remaining storage resources in each base station and cloud computing center equipment;

[0018] Information about the physical location of each base station;

[0019] Information about the interconnection status of each base station;

[0020] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0021] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0022] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0023] Optionally, also include:

[0024] Sending local status information of the first base station to the cloud computing center device;

[0025] The local status information is related to executing the AI ​​task.

[0026] Optionally, the local status information includes at least one of the following:

[0027] Information about the AI ​​model stored by the first base station;

[0028] Information about remaining computing resources in the first base station;

[0029] Information about resources stored in the first base station;

[0030] information about the physical location of the first base station;

[0031] Information about the connection between the first base station and other base stations.

[0032] Optionally, the local state information is inserted into a MAC layer transmission block as a MAC control information element and transmitted through a transmission channel.

[0033] Optionally, also include:

[0034] Utilizing the final service function chain, executing the AI ​​task and obtaining a processing result;

[0035] The processing result is sent to the terminal.

[0036] Optionally, before using the final service function chain to execute the AI ​​task and obtain the processing result, the method further includes:

[0037] Determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station;

[0038] Receiving final parameter information obtained by the cloud computing center device according to the initial parameter information;

[0039] The using the final service function chain to execute the AI ​​task and obtain the processing result includes:

[0040] According to the final service function chain and the final parameter information, the AI ​​model is used to execute the AI ​​task and obtain the processing result.

[0041] Optionally, determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes:

[0042] In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0043] When the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determine whether the AI ​​model corresponding to the AI ​​subtask exists in other base stations based on the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

[0044] The embodiment of the present invention further provides an information processing method, which is applied to a cloud computing center device, comprising:

[0045] Receiving primary service function chain arrangement information sent by at least one first base station;

[0046] According to the primary service function chain arrangement information and the second information, determining whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements, and obtaining a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of the AI ​​task;

[0047] According to the determination result, first information corresponding to each of the first base stations is obtained; the first information includes: final service function chain formation decision information or confirmation information;

[0048] Send corresponding first information to each of the first base stations.

[0049] Optionally, the acquiring, according to the determination result, first information respectively corresponding to each of the first base stations includes:

[0050] In a case where the determination result indicates that the computing resources of each base station on the service function chain can meet the current requirements of all primary service function chains, the first information including the confirmation information is used as the first information corresponding to each of the first base stations;

[0051] When the determination result indicates that the computing resources of each base station on the service function chain cannot meet the current needs of all primary service function chains, at least one of the primary service function chain orchestration information is updated to determine the first base station in which the primary service function chain is changed and the first base station in which the primary service function chain remains unchanged; the first information containing the target decision information is used as the first information corresponding to the first base station in which the primary service function chain is changed; the first information containing the confirmation information is used as the first information corresponding to the first base station in which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station in which the primary service function chain is changed.

[0052] Optionally, also include:

[0053] Sending the second information to each base station on the service function chain respectively;

[0054] The second information includes at least one of the following:

[0055] Information about AI models stored in each base station and cloud computing center equipment;

[0056] Information about remaining computing resources in each base station and cloud computing center equipment;

[0057] Information about remaining storage resources in each base station and cloud computing center equipment;

[0058] Information about the physical location of each base station;

[0059] Information about the interconnection status of each base station;

[0060] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0061] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0062] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0063] Optionally, also include:

[0064] Receiving local state information of the base station respectively sent by each base station on the service function chain; wherein the local state information is related to the execution of the AI ​​task;

[0065] The second information is obtained according to the local state information.

[0066] Optionally, the local status information includes at least one of the following:

[0067] Information about the AI ​​model stored in the base station;

[0068] Information about remaining computing resources in the base station;

[0069] Information about resources stored in the base station;

[0070] Information about the physical location of the base station;

[0071] Information about the connection between the base station and other base stations.

[0072] Optionally, the local state information is inserted into a MAC layer transmission block as a MAC control information element and transmitted through a transmission channel.

[0073] Optionally, also include:

[0074] Obtaining initial parameter information of at least one AI model;

[0075] Performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0076] Send final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0077] Optionally, the obtaining of initial parameter information of at least one AI model includes at least one of the following:

[0078] Receiving initial parameter information of the AI ​​model sent by at least one base station;

[0079] Receive the AI ​​subtask sent by the first base station, train the corresponding AI model according to the AI ​​subtask, and obtain initial parameter information of the AI ​​model.

[0080] Optionally, the training of a corresponding AI model according to the AI ​​subtask includes:

[0081] By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

[0082] The embodiment of the present invention further provides an information processing method, which is applied to a first base station, and includes:

[0083] Decompose the terminal's artificial intelligence (AI) task into at least two different AI subtasks;

[0084] Determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain initial parameter information of the AI ​​model; the training party is the first base station, a cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station;

[0085] Receive final parameter information obtained by the cloud computing center device according to the initial parameter information.

[0086] Optionally, determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes:

[0087] In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0088] In the case that the first base station does not have the AI ​​model corresponding to the AI ​​subtask locally, determine whether the other base stations have the AI ​​model corresponding to the AI ​​subtask according to the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device;

[0089] Among them, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of AI tasks.

[0090] Optionally, the second information includes at least one of the following:

[0091] Information about AI models stored in each base station and cloud computing center equipment;

[0092] Information about remaining computing resources in each base station and cloud computing center equipment;

[0093] Information about remaining storage resources in each base station and cloud computing center equipment;

[0094] Information about the physical location of each base station;

[0095] Information about the interconnection status of each base station;

[0096] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0097] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0098] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0099] The embodiment of the present invention further provides an information processing method, which is applied to a cloud computing center device, comprising:

[0100] Obtaining initial parameter information of at least one AI model;

[0101] Performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0102] Send final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0103] Optionally, the obtaining of initial parameter information of at least one AI model includes at least one of the following:

[0104] Receiving initial parameter information of the AI ​​model sent by at least one base station;

[0105] Receive the AI ​​subtask sent by the first base station, train the corresponding AI model according to the AI ​​subtask, and obtain initial parameter information of the AI ​​model.

[0106] Optionally, the training of a corresponding AI model according to the AI ​​subtask includes:

[0107] By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

[0108] Optionally, also include:

[0109] Sending second information to at least one base station respectively;

[0110] Among them, the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of AI tasks.

[0111] Optionally, the second information includes at least one of the following:

[0112] Information about AI models stored in each base station and cloud computing center equipment;

[0113] Information about remaining computing resources in each base station and cloud computing center equipment;

[0114] Information about remaining storage resources in each base station and cloud computing center equipment;

[0115] Information about the physical location of each base station;

[0116] Information about the interconnection status of each base station;

[0117] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0118] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0119] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0120] The embodiment of the present invention further provides an information processing device, applied to a first base station, including:

[0121] A first processing module, configured to decompose an artificial intelligence (AI) task of the terminal into at least two different AI subtasks;

[0122] A first sending module, configured to send primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks;

[0123] A first receiving module is used to receive first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information;

[0124] The first determining module is used to determine a final service function chain according to the first information.

[0125] Optionally, sending primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks includes:

[0126] Arrange the service position of each of the AI ​​subtasks according to the second information to form a primary service function chain, and obtain primary service function chain arrangement information; wherein the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; and the global status information is related to the execution of the AI ​​task;

[0127] The primary service function chain orchestration information is sent to the cloud computing center device.

[0128] Optionally, the second information includes at least one of the following:

[0129] Information about AI models stored in each base station and cloud computing center equipment;

[0130] Information about remaining computing resources in each base station and cloud computing center equipment;

[0131] Information about remaining storage resources in each base station and cloud computing center equipment;

[0132] Information about the physical location of each base station;

[0133] Information about the interconnection status of each base station;

[0134] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0135] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0136] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0137] Optionally, also include:

[0138] A second sending module, used to send the local state information of the first base station to the cloud computing center device;

[0139] The local status information is related to executing the AI ​​task.

[0140] Optionally, the local status information includes at least one of the following:

[0141] Information about the AI ​​model stored by the first base station;

[0142] Information about remaining computing resources in the first base station;

[0143] Information about resources stored in the first base station;

[0144] information about the physical location of the first base station;

[0145] Information about the connection between the first base station and other base stations.

[0146] Optionally, the local state information is inserted into a MAC layer transmission block as a MAC control information element and transmitted through a transmission channel.

[0147] Optionally, also include:

[0148] A second processing module is used to use the final service function chain to execute the AI ​​task and obtain a processing result;

[0149] The third sending module is used to send the processing result to the terminal.

[0150] Optionally, also include:

[0151] A third processing module is used to determine the training party of the AI ​​model corresponding to each of the AI ​​subtasks before using the final service function chain to execute the AI ​​task and obtain the processing result, and use the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station;

[0152] A second receiving module is used to receive final parameter information obtained by the cloud computing center device according to the initial parameter information;

[0153] The using the final service function chain to execute the AI ​​task and obtain the processing result includes:

[0154] According to the final service function chain and the final parameter information, the AI ​​model is used to execute the AI ​​task and obtain the processing result.

[0155] Optionally, determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes:

[0156] In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0157] When the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determine whether the AI ​​model corresponding to the AI ​​subtask exists in other base stations based on the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

[0158] The embodiment of the present invention further provides an information processing device, which is applied to a cloud computing center device, including:

[0159] A third receiving module, configured to receive primary service function chain arrangement information sent by at least one first base station;

[0160] A second determination module is used to determine whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements according to the primary service function chain arrangement information and the second information, and obtain a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of the AI ​​task;

[0161] A first acquisition module is used to acquire first information corresponding to each of the first base stations according to the determination result; the first information includes: final service function chain formation decision information or confirmation information;

[0162] The fourth sending module is used to send the corresponding first information to each of the first base stations.

[0163] Optionally, the acquiring, according to the determination result, first information respectively corresponding to each of the first base stations includes:

[0164] In a case where the determination result indicates that the computing resources of each base station on the service function chain can meet the current requirements of all primary service function chains, the first information including the confirmation information is used as the first information corresponding to each of the first base stations;

[0165] When the determination result indicates that the computing resources of each base station on the service function chain cannot meet the current needs of all primary service function chains, at least one of the primary service function chain orchestration information is updated to determine the first base station in which the primary service function chain is changed and the first base station in which the primary service function chain remains unchanged; the first information containing the target decision information is used as the first information corresponding to the first base station in which the primary service function chain is changed; the first information containing the confirmation information is used as the first information corresponding to the first base station in which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station in which the primary service function chain is changed.

[0166] Optionally, also include:

[0167] A fifth sending module, configured to send the second information to each base station on the service function chain respectively;

[0168] The second information includes at least one of the following:

[0169] Information about AI models stored in each base station and cloud computing center equipment;

[0170] Information about remaining computing resources in each base station and cloud computing center equipment;

[0171] Information about remaining storage resources in each base station and cloud computing center equipment;

[0172] Information about the physical location of each base station;

[0173] Information about the interconnection status of each base station;

[0174] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0175] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0176] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0177] Optionally, also include:

[0178] A fourth receiving module, configured to receive local state information of the base station respectively sent by each base station on the service function chain; wherein the local state information is related to the execution of the AI ​​task;

[0179] The third processing module is used to obtain the second information according to the local state information.

[0180] Optionally, the local status information includes at least one of the following:

[0181] Information about the AI ​​model stored in the base station;

[0182] Information about remaining computing resources in the base station;

[0183] Information about resources stored in the base station;

[0184] Information about the physical location of the base station;

[0185] Information about the connection between the base station and other base stations.

[0186] Optionally, the local state information is inserted into a MAC layer transmission block as a MAC control information element and transmitted through a transmission channel.

[0187] Optionally, also include:

[0188] A second acquisition module is used to obtain initial parameter information of at least one AI model;

[0189] A fourth processing module, used to perform classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0190] The sixth sending module is used to send the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0191] Optionally, the obtaining of initial parameter information of at least one AI model includes at least one of the following:

[0192] Receiving initial parameter information of the AI ​​model sent by at least one base station;

[0193] Receive the AI ​​subtask sent by the first base station, train the corresponding AI model according to the AI ​​subtask, and obtain initial parameter information of the AI ​​model.

[0194] Optionally, the training of a corresponding AI model according to the AI ​​subtask includes:

[0195] By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

[0196] The embodiment of the present invention further provides an information processing device, applied to a first base station, including:

[0197] A fifth processing module, configured to decompose the artificial intelligence (AI) task of the terminal into at least two different AI subtasks;

[0198] a sixth processing module, configured to determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and obtain initial parameter information of the AI ​​model using the training party; the training party is the first base station, a cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station;

[0199] The fifth receiving module is used to receive the final parameter information obtained by the cloud computing center device according to the initial parameter information.

[0200] Optionally, determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes:

[0201] In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0202] In the case that the first base station does not have the AI ​​model corresponding to the AI ​​subtask locally, determine whether the other base stations have the AI ​​model corresponding to the AI ​​subtask according to the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device;

[0203] Among them, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of AI tasks.

[0204] Optionally, the second information includes at least one of the following:

[0205] Information about AI models stored in each base station and cloud computing center equipment;

[0206] Information about remaining computing resources in each base station and cloud computing center equipment;

[0207] Information about remaining storage resources in each base station and cloud computing center equipment;

[0208] Information about the physical location of each base station;

[0209] Information about the interconnection status of each base station;

[0210] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0211] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0212] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0213] The embodiment of the present invention further provides an information processing device, which is applied to a cloud computing center device, including:

[0214] A third acquisition module is used to obtain initial parameter information of at least one AI model;

[0215] A seventh processing module, used to perform classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0216] The seventh sending module is used to send the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0217] Optionally, the obtaining of initial parameter information of at least one AI model includes at least one of the following:

[0218] Receiving initial parameter information of the AI ​​model sent by at least one base station;

[0219] Receive the AI ​​subtask sent by the first base station, train the corresponding AI model according to the AI ​​subtask, and obtain initial parameter information of the AI ​​model.

[0220] Optionally, the training of a corresponding AI model according to the AI ​​subtask includes:

[0221] By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

[0222] Optionally, also include:

[0223] an eighth sending module, configured to send second information to at least one base station respectively;

[0224] Among them, the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of AI tasks.

[0225] Optionally, the second information includes at least one of the following:

[0226] Information about AI models stored in each base station and cloud computing center equipment;

[0227] Information about remaining computing resources in each base station and cloud computing center equipment;

[0228] Information about remaining storage resources in each base station and cloud computing center equipment;

[0229] Information about the physical location of each base station;

[0230] Information about the interconnection status of each base station;

[0231] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0232] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0233] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0234] An embodiment of the present invention further provides an information processing device, the information processing device being a first base station, comprising: a processor and a transceiver;

[0235] The processor is used to decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0236] According to the at least two different AI subtasks, primary service function chain arrangement information is sent to a cloud computing center device through the transceiver;

[0237] Receiving, by the transceiver, first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information;

[0238] According to the first information, a final service function chain is determined.

[0239] Optionally, sending primary service function chain arrangement information to a cloud computing center device through the transceiver according to the at least two different AI subtasks includes:

[0240] Arrange the service position of each of the AI ​​subtasks according to the second information to form a primary service function chain, and obtain primary service function chain arrangement information; wherein the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; and the global status information is related to the execution of the AI ​​task;

[0241] The primary service function chain arrangement information is sent to the cloud computing center device through the transceiver.

[0242] Optionally, the second information includes at least one of the following:

[0243] Information about AI models stored in each base station and cloud computing center equipment;

[0244] Information about remaining computing resources in each base station and cloud computing center equipment;

[0245] Information about remaining storage resources in each base station and cloud computing center equipment;

[0246] Information about the physical location of each base station;

[0247] Information about the interconnection status of each base station;

[0248] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0249] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0250] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0251] Optionally, the processor is further configured to:

[0252] Sending the local status information of the first base station to the cloud computing center device through the transceiver;

[0253] The local status information is related to executing the AI ​​task.

[0254] Optionally, the local status information includes at least one of the following:

[0255] Information about the AI ​​model stored by the first base station;

[0256] Information about remaining computing resources in the first base station;

[0257] Information about resources stored in the first base station;

[0258] information about the physical location of the first base station;

[0259] Information about the connection between the first base station and other base stations.

[0260] Optionally, the local state information is inserted into a MAC layer transmission block as a MAC control information element and transmitted through a transmission channel.

[0261] Optionally, the processor is further configured to:

[0262] Utilizing the final service function chain, executing the AI ​​task and obtaining a processing result;

[0263] The processing result is sent to the terminal through the transceiver.

[0264] Optionally, the processor is further configured to:

[0265] Before using the final service function chain to execute the AI ​​task and obtain the processing result, determine the training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station;

[0266] Receiving, by the transceiver, final parameter information obtained by the cloud computing center device according to the initial parameter information;

[0267] The using the final service function chain to execute the AI ​​task and obtain the processing result includes:

[0268] According to the final service function chain and the final parameter information, the AI ​​model is used to execute the AI ​​task and obtain the processing result.

[0269] Optionally, determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes:

[0270] In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0271] When the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determine whether the AI ​​model corresponding to the AI ​​subtask exists in other base stations based on the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

[0272] The embodiment of the present invention further provides an information processing device, which is a cloud computing center device, including: a processor and a transceiver;

[0273] The processor is configured to receive primary service function chain arrangement information sent by at least one first base station through the transceiver;

[0274] According to the primary service function chain arrangement information and the second information, determining whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements, and obtaining a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of the AI ​​task;

[0275] According to the determination result, first information corresponding to each of the first base stations is obtained; the first information includes: final service function chain formation decision information or confirmation information;

[0276] The corresponding first information is sent to each of the first base stations through the transceiver.

[0277] Optionally, the acquiring, according to the determination result, first information respectively corresponding to each of the first base stations includes:

[0278] In a case where the determination result indicates that the computing resources of each base station on the service function chain can meet the current requirements of all primary service function chains, the first information including the confirmation information is used as the first information corresponding to each of the first base stations;

[0279] When the determination result indicates that the computing resources of each base station on the service function chain cannot meet the current needs of all primary service function chains, at least one of the primary service function chain orchestration information is updated to determine the first base station in which the primary service function chain is changed and the first base station in which the primary service function chain remains unchanged; the first information containing the target decision information is used as the first information corresponding to the first base station in which the primary service function chain is changed; the first information containing the confirmation information is used as the first information corresponding to the first base station in which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station in which the primary service function chain is changed.

[0280] Optionally, the processor is further configured to:

[0281] Sending the second information to each base station on the service function chain respectively through the transceiver;

[0282] The second information includes at least one of the following:

[0283] Information about AI models stored in each base station and cloud computing center equipment;

[0284] Information about remaining computing resources in each base station and cloud computing center equipment;

[0285] Information about remaining storage resources in each base station and cloud computing center equipment;

[0286] Information about the physical location of each base station;

[0287] Information about the interconnection status of each base station;

[0288] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0289] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0290] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0291] Optionally, the processor is further configured to:

[0292] Receiving, by the transceiver, local state information of the base station respectively sent by each base station on the service function chain; wherein the local state information is related to the execution of the AI ​​task;

[0293] The second information is obtained according to the local state information.

[0294] Optionally, the local status information includes at least one of the following:

[0295] Information about the AI ​​model stored in the base station;

[0296] Information about remaining computing resources in the base station;

[0297] Information about resources stored in the base station;

[0298] Information about the physical location of the base station;

[0299] Information about the connection between the base station and other base stations.

[0300] Optionally, the local state information is inserted into a MAC layer transmission block as a MAC control information element and transmitted through a transmission channel.

[0301] Optionally, the processor is further configured to:

[0302] Obtaining initial parameter information of at least one AI model;

[0303] Performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0304] The final parameter information of the AI ​​model is sent to the base station corresponding to the AI ​​model through the transceiver.

[0305] Optionally, the obtaining of initial parameter information of at least one AI model includes at least one of the following:

[0306] Receiving, by the transceiver, initial parameter information of the AI ​​model sent by at least one base station;

[0307] The AI ​​subtask sent by the first base station is received by the transceiver, and the corresponding AI model is trained according to the AI ​​subtask to obtain initial parameter information of the AI ​​model.

[0308] Optionally, the training of a corresponding AI model according to the AI ​​subtask includes:

[0309] By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

[0310] An embodiment of the present invention further provides an information processing device, the information processing device being a first base station, comprising: a processor and a transceiver;

[0311] The processor is used to decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0312] Determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain initial parameter information of the AI ​​model; the training party is the first base station, a cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station;

[0313] The final parameter information obtained by the cloud computing center device according to the initial parameter information is received through the transceiver.

[0314] Optionally, determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes:

[0315] In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0316] In the case that the first base station does not have the AI ​​model corresponding to the AI ​​subtask locally, determine whether the other base stations have the AI ​​model corresponding to the AI ​​subtask according to the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device;

[0317] Among them, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of AI tasks.

[0318] Optionally, the second information includes at least one of the following:

[0319] Information about AI models stored in each base station and cloud computing center equipment;

[0320] Information about remaining computing resources in each base station and cloud computing center equipment;

[0321] Information about remaining storage resources in each base station and cloud computing center equipment;

[0322] Information about the physical location of each base station;

[0323] Information about the interconnection status of each base station;

[0324] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0325] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0326] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0327] The embodiment of the present invention further provides an information processing device, which is a cloud computing center device, including: a processor and a transceiver;

[0328] The processor is used to obtain initial parameter information of at least one AI model;

[0329] Performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0330] The final parameter information of the AI ​​model is sent to the base station corresponding to the AI ​​model through the transceiver.

[0331] Optionally, the obtaining of initial parameter information of at least one AI model includes at least one of the following:

[0332] Receiving, by the transceiver, initial parameter information of the AI ​​model sent by at least one base station;

[0333] The AI ​​subtask sent by the first base station is received by the transceiver, and the corresponding AI model is trained according to the AI ​​subtask to obtain initial parameter information of the AI ​​model.

[0334] Optionally, the training of a corresponding AI model according to the AI ​​subtask includes:

[0335] By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

[0336] Optionally, the processor is further configured to:

[0337] Sending second information to at least one base station respectively through the transceiver;

[0338] Among them, the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of AI tasks.

[0339] Optionally, the second information includes at least one of the following:

[0340] Information about AI models stored in each base station and cloud computing center equipment;

[0341] Information about remaining computing resources in each base station and cloud computing center equipment;

[0342] Information about remaining storage resources in each base station and cloud computing center equipment;

[0343] Information about the physical location of each base station;

[0344] Information about the interconnection status of each base station;

[0345] Wherein, each base station is connected to the cloud computing center equipment respectively.

[0346] Optionally, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel;

[0347] The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

[0348] An embodiment of the present invention also provides an information processing device, including a memory, a processor, and a program stored in the memory and executable on the processor; when the processor executes the program, the information processing method on the first base station side or the cloud computing center device side is implemented.

[0349] An embodiment of the present invention further provides a readable storage medium on which a program is stored. When the program is executed by a processor, the steps in the information processing method on the first base station side or the cloud computing center device side are implemented.

[0350] The beneficial effects of the above technical solution of the present invention are as follows:

[0351] In the above scheme, the information processing method decomposes the artificial intelligence AI task of the terminal into at least two different AI subtasks; according to the at least two different AI subtasks, sends primary service function chain orchestration information to the cloud computing center device; receives the first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information; according to the first information, determines the final service function chain; can support the first base station and the cloud computing center device to collaborate based on a global perspective (second information) to determine the AI ​​service function chain orchestration, thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0352] Figure 1 The information processing method of the embodiment of the present invention is shown in FIG. Figure 1 ;

[0353] Figure 2 The information processing method of the embodiment of the present invention is shown in FIG. Figure 2 ;

[0354] Figure 3 The information processing method of the embodiment of the present invention is shown in FIG. Figure 3 ;

[0355] Figure 4 The information processing method of the embodiment of the present invention is shown in FIG. Figure 4 ;

[0356] Figure 5 This is a flow chart of a model training method based on edge federated learning according to an embodiment of the present invention;

[0357] Figure 6 A schematic diagram of a service flow based on a state feedback and orchestration collaboration mechanism according to an embodiment of the present invention;

[0358] Figure 7 It is intended to represent the local state of an embodiment of the present invention;

[0359] Figure 8 It is a schematic diagram of the global state of an embodiment of the present invention;

[0360] Fig. 9 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 1 ;

[0361] Fig.10 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 2 ;

[0362] Fig.11 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 3 ;

[0363] Fig.12 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 4 ;

[0364] Fig.13 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 1 ;

[0365] Fig.14 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 2 ;

[0366] Fig.15 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 3 ;

[0367] Fig.16 The structure of the information processing device according to the embodiment of the present invention is shown in FIG. Figure 4 . DETAILED DESCRIPTION

[0368] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0369] The following first introduces the relevant contents of this plan.

[0370] (1) The existing AI model training method is to use the powerful computing and storage resources of the cloud computing center to complete AI model training and storage in the cloud, and the UE's tasks will also be implemented on-demand in the cloud. However, completing the training of all AI models in the cloud computing center has the following problems: first, link congestion and data security issues caused by uploading training data; second, a large number of AI model training in the cloud will occupy a large amount of computing resources in the cloud center (i.e., cloud computing center); third, when the AI ​​model is in use, the communication between the UE and the cloud will increase the processing delay of the AI ​​task.

[0371] (2) Existing edge base stations interact with each other through the Xn port, and the interactive status information includes the computing resources, storage resources, and location information of the base station. However, the existing information on edge base station interactions lacks the expression of model resources, making it difficult to provide complete information for the orchestration of the AI ​​service function chain. In addition, the topological structure of the edge base station network is often not fully connected. An edge base station only interacts with the surrounding edge base stations, so it cannot have global knowledge when making decisions.

[0372] Based on the above, the present invention aims at the problem that the information processing scheme for the orchestration of the AI ​​service function chain in the existing technology lacks a global perspective, resulting in poor orchestration effect, and provides an information processing method applied to the first base station, such as Figure 1 As shown, including:

[0373] Step 11: Decompose the terminal's artificial intelligence (AI) task into at least two different AI subtasks;

[0374] Step 12: Sending primary service function chain arrangement information to the cloud computing center device according to the at least two different AI subtasks;

[0375] Step 13: receiving the first information sent by the cloud computing center device; the first information includes: final service function chaining decision information or confirmation information;

[0376] Step 14: Determine the final service function chain based on the first information.

[0377] Among them, when the first information includes the final service function chain formation decision information, the final service function chain can be the service function chain indicated or carried by the final service function chain formation decision information; when the first information includes the confirmation information, it can be the following primary service function chain, which is not limited here.

[0378] The information processing method provided by the embodiment of the present invention decomposes the artificial intelligence AI task of the terminal into at least two different AI subtasks; according to the at least two different AI subtasks, sends primary service function chain orchestration information to the cloud computing center device; receives the first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information; according to the first information, determines the final service function chain; can support the first base station and the cloud computing center device to collaborate to determine the AI ​​service function chain orchestration based on a global perspective (second information), thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0379] Among them, sending primary service function chain orchestration information to the cloud computing center device according to the at least two different AI subtasks includes: arranging the service position of each of the AI ​​subtasks according to the second information to form a primary service function chain, and obtaining the primary service function chain orchestration information; wherein the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of AI tasks; and sending the primary service function chain orchestration information to the cloud computing center device.

[0380] In this way, the primary service function chain arrangement information can be accurately obtained and transmitted based on the global status information (second information).

[0381] In an embodiment of the present invention, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0382] This makes the second information clear.

[0383] The second information is inserted into the MAC layer transmission block as a media access control MAC control element and transmitted through a transmission channel; the second information is located at the beginning of the MAC protocol data unit (ie, MAC PDU) in the MAC layer transmission block.

[0384] In this way, the second information can be accurately transmitted.

[0385] Furthermore, the information processing method also includes: sending local status information of the first base station to the cloud computing center device; wherein the local status information is related to executing the AI ​​task.

[0386] In this way, the cloud computing center device can accurately obtain the local status information of the first base station.

[0387] Among them, the local status information includes at least one of the following: information about the AI ​​model stored in the first base station; information about the remaining computing resources in the first base station; information about the storage resources in the first base station; information about the physical location of the first base station; information about the connection status of the first base station with other base stations.

[0388] In this way, the local status information of the first base station can be clarified.

[0389] In the embodiment of the present invention, the local state information is inserted into a MAC layer transmission block as a MAC control information element and is transmitted through a transmission channel.

[0390] In this way, the local state information of the first base station can be accurately transmitted.

[0391] Furthermore, the information processing method also includes: utilizing the final service function chain to execute the AI ​​task and obtain a processing result; and sending the processing result to the terminal.

[0392] In this way, the AI ​​task of the terminal can be accurately executed. Among them, "using the final service function chain to execute the AI ​​task" may include: using each base station included in the final service function chain to respectively execute the AI ​​subtask assigned to each base station (itself); specifically, the final service function chain may include the first base station, and correspondingly, "using the final service function chain to execute the AI ​​task" may include: using the first base station to execute the AI ​​subtask assigned to the first base station, that is, the first base station executes the assigned AI subtask; but it is not limited to this.

[0393] In an embodiment of the present invention, before using the final service function chain to execute the AI ​​task and obtain the processing result, it also includes: determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks, and using the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station; receiving the final parameter information obtained by the cloud computing center device according to the initial parameter information; using the final service function chain to execute the AI ​​task and obtain the processing result, including: according to the final service function chain and the final parameter information, using the AI ​​model to execute the AI ​​task and obtain the processing result.

[0394] In this way, the AI ​​tasks can be accurately executed, and the problems of completing the training of all AI models in the cloud computing center, such as link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources in the cloud computing center, and long AI task processing delays, can be avoided; thereby further supporting the improvement of orchestration effects and supporting better solutions to the problem that the information processing solutions for the orchestration of AI service function chains in the existing technology lack a global perspective, resulting in poor orchestration effects.

[0395] Wherein, determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks includes: (1) when the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; (2) when the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determining whether the other base stations have the AI ​​model corresponding to the AI ​​subtask based on the second information; if so, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

[0396] In this way, the training method of the AI ​​model corresponding to the AI ​​subtask can be accurately determined.

[0397] The embodiment of the present invention also provides an information processing method, which is applied to a cloud computing center device, such as Figure 2 As shown, including:

[0398] Step 21: receiving primary service function chain arrangement information sent by at least one first base station;

[0399] Step 22: Determine whether the computing resources of each base station on the service function chain meet all current requirements of the primary service function chain according to the primary service function chain arrangement information and the second information, and obtain a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of the AI ​​task;

[0400] Step 23: acquiring first information corresponding to each of the first base stations according to the determination result; the first information includes: final service function chain formation decision information or confirmation information;

[0401] Step 24: Send corresponding first information to each of the first base stations.

[0402] The information processing method provided in the embodiment of the present invention receives the primary service function chain orchestration information sent by at least one first base station; determines whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements based on the primary service function chain orchestration information and the second information, and obtains a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of AI tasks; according to the determination result, obtains the first information corresponding to each of the first base stations; the first information includes: the final service function chain formation decision information or confirmation information; sends the corresponding first information to each of the first base stations; can support the first base station and the cloud computing center device to collaborate to determine the AI ​​service function chain orchestration based on a global perspective (second information), thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0403] Wherein, obtaining the first information corresponding to each of the first base stations respectively according to the determination result includes: (1) when the determination result indicates that the computing resources of each base station on the service function chain can meet all current primary service function chain requirements, using the first information containing confirmation information as the first information corresponding to each of the first base stations respectively; (2) when the determination result indicates that the computing resources of each base station on the service function chain cannot meet all current primary service function chain requirements, updating at least one of the primary service function chain orchestration information to determine the first base station for which the primary service function chain is to be changed and the first base station for which the primary service function chain remains unchanged; using the first information containing target decision information as the first information corresponding to the first base station for which the primary service function chain is to be changed; using the first information containing confirmation information as the first information corresponding to the first base station for which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station for which the primary service function chain is to be changed.

[0404] In this way, the first information corresponding to each of the first base stations can be accurately obtained.

[0405] Furthermore, the information processing method also includes: sending the second information to each base station on the service function chain respectively; wherein the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0406] In this way, each base station can accurately obtain the second information.

[0407] The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the beginning position of a MAC protocol data unit in the MAC layer transmission block.

[0408] In this way, the second information can be accurately transmitted.

[0409] Furthermore, the information processing method also includes: receiving local status information of the base station sent respectively by each base station on the service function chain; wherein the local status information is related to the execution of the AI ​​task; and obtaining the second information based on the local status information.

[0410] In this way, the second information can be accurately obtained.

[0411] Among them, the local status information includes at least one of the following: information about the AI ​​model stored in the base station; information about the remaining computing resources in the base station; information about the storage resources in the base station; information about the physical location of the base station; information about the connection status of the base station with other base stations.

[0412] This allows for clear local state information.

[0413] In the embodiment of the present invention, the local state information is inserted into a MAC layer transmission block as a MAC control information element and is transmitted through a transmission channel.

[0414] In this way, the local status information can be accurately transmitted.

[0415] Furthermore, the information processing method also includes: obtaining initial parameter information of at least one AI model; performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each AI model; and sending the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0416] In this way, the final parameter information can be accurately obtained, and it can support avoiding the problems of completing the training of all AI models in the cloud computing center due to link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources in the cloud computing center, and large AI task processing delays. This can further support the improvement of orchestration effects and support better solutions to the problem that the information processing solutions for the orchestration of AI service function chains in the existing technology lack a global perspective, resulting in poor orchestration effects.

[0417] The obtaining of initial parameter information of at least one AI model includes at least one of the following: (1) receiving initial parameter information of the AI ​​model sent by at least one base station; (2) receiving an AI subtask sent by the first base station, training a corresponding AI model according to the AI ​​subtask, and obtaining the initial parameter information of the AI ​​model.

[0418] This allows you to obtain the initial parameter information of the AI ​​model in multiple ways.

[0419] In an embodiment of the present invention, the training of the corresponding AI model according to the AI ​​subtask includes: using transfer learning to train the corresponding AI model according to the AI ​​subtask.

[0420] This can improve training efficiency and reduce training costs.

[0421] The embodiment of the present invention also provides an information processing method, which is applied to a first base station, such as Figure 3 As shown, including:

[0422] Step 31: Decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0423] Step 32: Determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain initial parameter information of the AI ​​model; the training party is the first base station, a cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station;

[0424] Step 33: Receive final parameter information obtained by the cloud computing center device according to the initial parameter information.

[0425] The information processing method provided by the embodiment of the present invention decomposes the artificial intelligence AI task of the terminal into at least two different AI subtasks; determines the training party of the AI ​​model corresponding to each of the AI ​​subtasks, and uses the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station; receives the final parameter information obtained by the cloud computing center device according to the initial parameter information; can avoid the problems of link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources of the cloud computing center, and large AI task processing delay caused by the completion of the training of all AI models in the cloud computing center; thereby further supporting the improvement of the orchestration effect, and supporting better solving the problem that the information processing scheme for the orchestration of the AI ​​service function chain in the prior art lacks a global perspective and leads to poor orchestration effect.

[0426] Wherein, determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks includes: (1) when the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; (2) when the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determining whether the other base stations have the AI ​​model corresponding to the AI ​​subtask based on second information; if so, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device; wherein, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of the AI ​​task.

[0427] In this way, the training method of the AI ​​model corresponding to the AI ​​subtask can be accurately determined.

[0428] In an embodiment of the present invention, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0429] This makes the second information clear.

[0430] The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the beginning position of a MAC protocol data unit in the MAC layer transmission block.

[0431] In this way, the second information can be accurately transmitted.

[0432] The embodiment of the present invention also provides an information processing method, which is applied to a cloud computing center device, such as Figure 4 As shown, including:

[0433] Step 41: Obtaining initial parameter information of at least one AI model;

[0434] Step 42: performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0435] Step 43: Send final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0436] The information processing method provided in the embodiment of the present invention obtains the initial parameter information of at least one AI model; performs classification and aggregation operations on the initial parameter information to obtain the final parameter information of each AI model; and sends the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model. It can support avoiding the problems of link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources in the cloud computing center, and large AI task processing delays when completing the training of all AI models in the cloud computing center. It can further support the improvement of the orchestration effect and support better solving the problem that the information processing scheme for the orchestration of the AI ​​service function chain in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0437] The obtaining of initial parameter information of at least one AI model includes at least one of the following: (1) receiving initial parameter information of the AI ​​model sent by at least one base station; (2) receiving an AI subtask sent by a first base station, training a corresponding AI model according to the AI ​​subtask, and obtaining the initial parameter information of the AI ​​model.

[0438] In this way, the initial parameter information of the AI ​​model can be obtained in multiple ways. Among them, the first base station can be: a base station that decomposes the artificial intelligence AI task of the terminal into at least two different AI subtasks, but is not limited to this.

[0439] In an embodiment of the present invention, the training of the corresponding AI model according to the AI ​​subtask includes: using transfer learning to train the corresponding AI model according to the AI ​​subtask.

[0440] This can improve training efficiency and reduce training costs.

[0441] Furthermore, the information processing method also includes: sending second information to at least one base station respectively; wherein the second information is global status information corresponding to the cloud computing center device; and the global status information is related to the execution of AI tasks.

[0442] In this way, each base station can accurately obtain the second information.

[0443] Among them, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0444] This makes the second information clear.

[0445] In an embodiment of the present invention, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the starting position of a MAC protocol data unit in the MAC layer transmission block.

[0446] In this way, the second information can be accurately transmitted.

[0447] The information processing method provided in the embodiment of the present invention is illustrated below by taking an edge base station as an example.

[0448] In response to the above technical problems, the embodiments of the present invention provide an information processing method, which can be specifically implemented as a service method based on an edge federated learning architecture, including: a model training method combining federated learning and transfer learning for service orchestration, a hierarchical state feedback and orchestration collaboration mechanism, and a design of a state table data reporting control package, which can improve the AI ​​model training efficiency, AI service management and orchestration flexibility in 6G networks. Specifically, this solution mainly includes the following contents:

[0449] (1) In response to the training and location deployment issues of AI models, this solution proposes a model training method that combines federated learning and transfer learning for service orchestration: the edge base station (corresponding to the first base station mentioned above) first decomposes the UE task into different AI subtasks, and transmits the corresponding training data to the appropriate edge base station for model training according to the matching of the existing AI model and the subtask on the edge base station. Based on the idea of ​​federated learning, each edge base station uploads the local model training parameters (corresponding to the initial parameter information mentioned above) to the cloud center. The cloud center classifies AI models, aggregates similar AI models, and stores them in the model library (corresponding to the above classification and aggregation operations on the initial parameter information to obtain the final parameter information of each AI model), and sends the AI ​​model parameters to the corresponding edge base station (corresponding to the above sending of the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model) to complete the model parameter update. In order to reduce the training cost and benefit from the model sharing mechanism of the model library, the cloud center can use the transfer learning method to train the new model (corresponding to the above method of using transfer learning to train the corresponding AI model according to the AI ​​subtask).

[0450] (2) In response to the service function chain orchestration problem, this solution introduces a hierarchical state feedback and orchestration collaboration mechanism in the AI ​​service function chain orchestration of the edge computing network: the cloud center is responsible for the collection and distribution of the network global state resources (corresponding to the above-mentioned second information), as well as the confirmation and / or modification of the service function chain orchestration (corresponding to the above-mentioned determination of whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements based on the primary service function chain orchestration information and the second information, and obtaining a determination result; according to the determination result, obtaining the first information corresponding to each first base station). Among them, each edge base station reports the local state information to the cloud center, which is summarized by the cloud center into a global state table (corresponding to the above-mentioned global state information) and distributed to all edge base stations. The edge base station performs AI service function chain orchestration based on the global status table (corresponding to the above-mentioned orchestration of the service position of each AI subtask according to the second information, forming a primary service function chain, and obtaining the primary service function chain orchestration information), and uploads the orchestration decision to the cloud center for confirmation or modification (corresponding to the above-mentioned sending of the primary service function chain orchestration information to the cloud computing center device), and the cloud center then feeds back the confirmation and modification information to the corresponding edge base station (corresponding to the above-mentioned sending of the corresponding first information to each of the first base stations).

[0451] (3) In response to the lack of interaction between edge base stations for computing, storage, and model resources, this solution proposes a method for reporting MAC CE measurement data: the local state table of the edge base station is inserted into the MAC control information element (corresponding to the above-mentioned local state information being inserted into the MAC layer transmission block as a MAC control information element), which includes the computing resources, storage resources, model resources, location information, etc. of the edge base station. The local state table is regularly uploaded to the cloud center, which is integrated into a global state table and sent to all edge base stations so that the edge base stations can be fully aware of the resource status in the network and formulate reasonable AI service function chain orchestration.

[0452] The following are specific examples to illustrate the contents of this plan.

[0453] 1. Model training method combining federated learning and transfer learning for service orchestration;

[0454] Orchestrator base station (corresponding to the first base station mentioned above): For the UE (terminal), the edge base station connected to it will provide it with AI service function chain orchestration services. The edge base station is called the orchestrator base station of the UE.

[0455] Global status table: records all edge base stations in the network (i.e., all edge base stations under the cloud center) as well as the computing resources, storage resources, model resources of the cloud center, the location information of each edge base station, the interconnection status of the edge base stations, etc.; it can be carried in the MAC control cell and sent from the cloud center to each edge base station.

[0456] 1. The training method of the federated learning model for service orchestration is as follows:

[0457] The UE unloads the task data to the orchestrator base station; the orchestrator base station first decomposes the task into multiple AI subtasks, each of which corresponds to a different AI service; the orchestrator base station performs corresponding edge federated learning model training based on the global state table. Specifically:

[0458] According to the matching between the existing AI model on the edge base station and the required AI service, it is divided into the following three cases:

[0459] Case 1: If there is an AI model matching the required AI service in the scheduler base station, the AI ​​subtask data is left in the scheduler base station for model training; corresponding to the case where the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, it is determined that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station;

[0460] Case 2: There is no AI model matching the required AI service in the scheduler base station, but it can be found in the global state table that there is at least one edge base station with established connection, in which the AI ​​model matches the required AI service; then the scheduler edge base station transmits the AI ​​subtask to the base station (when multiple base stations are queried, one will be selected preferentially, such as the one with the shortest route) for model training; corresponding to the above-mentioned case where the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, according to the second information, it is determined whether the other base stations have the AI ​​model corresponding to the AI ​​subtask; if so, it is determined that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station;

[0461] Case three: There is no AI model matching the required AI service in the orchestrator edge base station (i.e., the orchestrator base station), and no edge base station matching the task and with a connection established can be found in the global status table. The orchestrator edge base station sends the subtask to the cloud center for training of the new model and stores it in the global model library; corresponding to the above-mentioned case where no AI model exists, the training party of the AI ​​model corresponding to the AI ​​subtask is determined to be the cloud computing center device.

[0462] Among them, each edge base station may serve as an orchestrator edge base station, and the AI ​​models trained in each edge base station may also be repeated (that is, multiple base stations may have trained the same AI model). After the local model training is completed, each edge base station uploads the model parameters (corresponding to the initial parameter information of the above-mentioned AI model) to the cloud center. After the cloud center classifies the AI ​​models, it completes the aggregation of similar AI models and sends the AI ​​model parameters to the corresponding edge base stations (corresponding to the above-mentioned classification and aggregation operations on the initial parameter information to obtain the final parameter information of each AI model; sending the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model) to complete the model parameter update.

[0463] Specific examples include Figure 5 As shown in the figure, the model training method based on edge federated learning includes the following operations:

[0464] (1) The orchestrator base station matches the AI ​​subtask with the AI ​​model and determines to keep the AI ​​subtask data in the orchestrator base station for model training and send the obtained initial model parameters to the cloud center, or sends the AI ​​subtask data to the cloud center and uses the cloud center to train a new model (transfer learning) and store it in the model library, or selects a connected edge base station for model training (sends the AI ​​subtask data to the selected edge base station) and sends the obtained initial model parameters to the cloud center.

[0465] (2) The cloud center performs model classification and model aggregation based on the initial model parameters to obtain the final model parameters; then the final model parameters are sent to each edge base station.

[0466] 2. The migration learning method of the cloud center is as follows:

[0467] First, classify the new training tasks (AI subtasks): for example, data analysis, natural language processing, image and video analysis, prediction and optimization, decision making, etc.

[0468] Secondly, select AI model resources of the same service type (as the service required by the UE) from the global model library as the original model for transfer learning. Transfer learning can be divided into the following three cases:

[0469] Case 1: The source domain (i.e., data collected by UE perception) and the target domain (i.e., data corresponding to the AI ​​model stored in the model library) belong to the same category of data, and the model or parameters are directly shared (i.e., the existing AI model is directly used);

[0470] Case 2: There is some similar data in the source domain and the target domain, and the allocation weight value of the original model is adjusted;

[0471] Case 3: The source domain and the target domain do not belong to the same type of data, but there are common cross-features. Through feature transformation, the features of the source domain and the target domain are mapped to the same space, so that the source domain data and the target domain data in the space have the same data distribution, and then traditional machine learning is performed.

[0472] 2. Hierarchical status feedback and orchestration collaboration mechanism;

[0473] Under the edge federated learning architecture, local AI model training and AI service function orchestration occur at the edge base station, and service function chain modification and model aggregation occur at the cloud center. This greatly reduces the amount of data transmitted in the network, avoids UE data leakage, and improves the computing resource utilization of the edge base station.

[0474] Specifically, Figure 6 As shown in the service flow chart based on state feedback and orchestration collaboration mechanism, the service implementation process based on the edge federated learning architecture includes:

[0475] Step 1: The cloud center sends a global status table to all edge base stations (including orchestrator base stations 1 and 2), which contains the following information: 1. Existing AI models in each edge base station and cloud center (corresponding to the information of AI models already stored in the above-mentioned base stations and cloud computing center equipment); 2. Remaining computing resources of each edge base station and cloud center (corresponding to the information of remaining computing resources in the above-mentioned base stations and cloud computing center equipment); 3. Remaining storage resources of each edge base station and cloud center (corresponding to the information of remaining storage resources in the above-mentioned base stations and cloud computing center equipment); 4. The physical location of each edge base station (corresponding to the information of the physical location of each base station); 5. The interconnection status of each edge base station (corresponding to the information of the interconnection status of each base station).

[0476] Step 2: UE1 sends an AI task to the scheduler base station 1 to trigger the primary service function chain orchestration; the scheduler base station 1 performs the primary service function chain orchestration and reports the primary service function chain to the cloud center (corresponding to the above-mentioned sending of the primary service function chain orchestration information to the cloud computing center device); UE2 sends an AI task to the scheduler base station 2 to trigger the primary service function chain orchestration; the scheduler base station 2 performs the primary service function chain orchestration and reports the primary service function chain to the cloud center (corresponding to the above-mentioned sending of the primary service function chain orchestration information to the cloud computing center device);

[0477] Specifically, the orchestrator base station 1 (or the orchestrator base station 2) can decompose the AI ​​task of UE1 (or UE2) into multiple different AI subtasks, and then arrange the service position of each AI subtask according to the global state table to form a primary service function chain (corresponding to the above-mentioned decomposition of the terminal's artificial intelligence AI task into at least two different AI subtasks; according to the second information, arrange the service position of each of the AI ​​subtasks to form a primary service function chain, and obtain the primary service function chain arrangement information). Among them, the primary service function chain has the characteristics of the minimum number of hops and meets all the AI ​​services required in the AI ​​task. The orchestrator base station 1 (or the orchestrator base station 2) uploads the primary service function chain arrangement information to the cloud center for confirmation or modification.

[0478] Step 3: Since the decisions of each primary service function chain are only made at the edge base station of the scheduler, there is a lack of a global perspective on the current decision, which may cause a certain edge service to appear on multiple primary service function chains, and the computing resources of the base station are insufficient to support the computing needs of all primary service chains. Therefore, the cloud center collects all primary service function chain information, and based on the latest global status table, determines whether the computing resources of each edge base station on the service function chain can meet the current needs of all primary service function chains (corresponding to the above-mentioned determination of whether the computing resources of each base station on the service function chain meet the current needs of all primary service function chains based on the primary service function chain arrangement information and the second information). If the conditions are met, the cloud center confirms all primary service function chains (corresponding to the above-mentioned case where the computing resources of each base station on the determination result indicating that the service function chain can meet the current requirements of all primary service function chains, the first information containing the confirmation information is used as the first information corresponding to each of the first base stations), and the UE's AI service is executed according to the service function chain formulated by its scheduler base station; if the conditions are not met, the cloud center modifies the service function chain and selects other edge base station nodes to form a chain (corresponding to the above-mentioned case where the computing resources of each base station on the determination result indicating that the service function chain cannot meet the current requirements of all primary service function chains, at least one of the primary service functions is modified). The energy chain arrangement information is updated to determine the first base station for changing the primary service function chain and the first base station for maintaining the primary service function chain unchanged), and the final service function chain formation decision (which can be the final service function chain) is sent to the arranger base station whose primary service function chain needs to be modified (corresponding to the first information that will contain the target decision information as the first information corresponding to the first base station for changing the primary service function chain), and the arranger base station whose primary service function chain has not been modified (i.e., no modification is required) sends a confirmation message (corresponding to the first information that will contain the confirmation information as the first information corresponding to the first base station for maintaining the primary service function chain unchanged). Among them, the cloud center can use game theory algorithms to modify the decision, that is, when the combined strategies of all service function chains reach Nash equilibrium, it is the best strategy. In other words, on this strategy combination, any single change of any service function chain node cannot make the system optimization target better.

[0479] Specifically, orchestrator base station 1 can be regarded as the orchestrator base station whose primary service function chain has not been modified, and orchestrator base station 2 can be regarded as the orchestrator base station whose primary service function chain needs to be modified. Correspondingly, the cloud center confirms the service function chain to orchestrator base station 1 and modifies the service function chain to orchestrator base station 2.

[0480] Step 4: The computing task of the UE (UE1 or UE2) is processed in the service function chain and the processing result is fed back to the UE (corresponding to the above-mentioned use of the final service function chain to execute the AI ​​task and obtain the processing result; sending the processing result to the terminal). Due to the mobility of the UE, if the UE switches the edge base station to which it is connected at this moment, there are two situations: one is that the edge base station to which the UE is connected (that is, the base station after switching) is on the service function chain, then all the calculation results on the service function chain are directly transmitted to the base station and fed back to the UE; the other is that the edge base station to which the UE is connected is not on the service function chain, then each base station on the service function chain broadcasts the UE location information, and transmits the calculation results to the target base station through routing forwarding, and finally feeds back to the UE.

[0481] Specifically, it can be assumed that the organizer base station 1 is the edge base station connected to UE1 and is on the service function chain, then the organizer base station 1 can directly feed back the calculation results to the UE; assuming that the organizer base station 2 is the edge base station connected to UE2 and is not on the service function chain, then the organizer base station 2 can broadcast and feed back the calculation results to the UE.

[0482] Step 5: All edge base stations on the service function chain (including orchestrator base station 1 and orchestrator base station 2) feed back the local status table to the cloud center. The local status table may include the following information: 1. The existing AI model of the local edge base station (corresponding to the information of the AI ​​model stored in the first base station); 2. The remaining computing resources of the local edge base station (corresponding to the information of the remaining computing resources in the first base station); 3. The storage resources of the local edge base station (corresponding to the information of the storage resources in the first base station); 4. The physical location of the local edge base station (corresponding to the information of the physical location of the first base station); 4. The connection status of the local edge base station with other base stations (corresponding to the information of the connection status of the first base station with other base stations).

[0483] 3. Design of status table data reporting control package;

[0484] Before arranging the AI ​​service function chain, each edge base station needs to upload the local state table to the cloud center to be integrated into the global state table. Figure 7As shown, the local status table can be inserted into the MAC layer transmission block as a MAC control information element and transmitted through the transmission channel (corresponding to the above-mentioned local status information being inserted into the MAC layer transmission block as a MAC control information element and transmitted through the transmission channel). Among them, the MAC PDU of the MAC layer transmission block can be provided with multiple MAC SDUs (service data units) and multiple H fields, and the H field may include an R field (reserved field), an F field (8 or 16-bit length field), an LCID field (can be used to carry a logical channel identifier (6 bits)) and an SDU length field (can be used to indicate the SDU length (8 or 16 bits)). For details, please refer to Figure 7 .

[0485] Specifically, the local status table may include the following information: 1. The existing AI model of the local edge base station (i.e., the local existing AI model); 2. The remaining computing resources of the local edge base station (i.e., the local remaining computing resources); 3. The storage resources of the local edge base station (i.e., the local remaining storage resources); 4. The physical location of the local edge base station (i.e., the location of the local edge base station); 5. The connection status of the local edge base station and other base stations (i.e., the interconnection status of the local edge base station).

[0486] It is assumed that the existing local AI model is represented by 8 bits, for example, 0000 0000-0000 0011 can represent image recognition model, speech recognition model, content recommendation model, online decision model, etc.; the local remaining computing resources are represented by 4 bits, and the total local computing resources are evenly divided into 16 levels. For example, an edge base station has 160GHz CPU resources, and 0000-1111 respectively represent the remaining CPU resources are 0-10GHz, 10-20GHz, and so on; the local storage computing resources are also represented by 4 bits, and the total local storage resources are evenly divided into 16 levels. For example, an edge base station has 3200G memory resources, and 0000-1111 respectively represent the remaining CPU resources are 0-200G, 200-400G, and so on; the edge base station interconnection status field can record the IDs of other edge base stations connected to the edge base station. The location of the local edge base station can adopt a unified coding rule: city code (2 characters) + network code (1 character) + manufacturer code (1 character) + indoor or private network code (optional: 1 character) + base station serial number (3 characters) + cell serial number (1 character), for example:

[0487] City code: BJ, indicating the first city;

[0488] Network code: G, indicating GSM900;

[0489] Equipment manufacturer code: H, indicating manufacturer H;

[0490] Indoor distribution code: M, indicating indoor station;

[0491] Base station number: 239;

[0492] Community number: 1.

[0493] Subsequently, after receiving the local status tables of each edge base station, the cloud center can integrate the status table information into a global status table and broadcast it downlink to all edge base stations. In the downlink transmission, the global status table can be used as a MAC control cell at the beginning of the MAC PDU (belonging to the MAC layer transmission block), such as Figure 8 As shown, this placement of the global state table is beneficial to the low-latency operation of the terminal.

[0494] Among them, the global state table contains the following information: 1. Existing AI models in each edge base station and cloud center (including existing AI models of all edge base stations); 2. Remaining computing resources of each edge base station and cloud center (including remaining computing resources of all edge base stations); 3. Remaining storage resources of each edge base station and cloud center (including remaining storage resources of all edge base stations); 4. Physical location of each edge base station and cloud center (including the location of all edge base stations); 5. The interconnection between each edge base station and cloud center (including the interconnection of all edge base stations). Specifically, the structure of the global state table is similar to that of the local state table. For related introduction, please refer to the relevant content of the local state table, which will not be repeated here.

[0495] It should be noted that the relevant contents of the above parts can refer to each other, and the repeated contents will not be repeated.

[0496] As described above, the solution provided by the embodiment of the present invention mainly includes the following three parts:

[0497] 1. Model training method combining federated learning and transfer learning for service orchestration: The edge base station first decomposes the UE task into different AI subtasks, and transfers the corresponding training data to the appropriate location for model training according to the matching of the existing AI model and the subtask on the edge base station. Each edge base station uploads the model parameters (corresponding to the above initial parameter information) to the cloud center. The cloud center classifies AI models, aggregates similar AI models, stores them in the model library, and sends the AI ​​model parameters (corresponding to the above final parameter information) to the corresponding edge base station to complete the model parameter update. In this scheme, based on the classification of subtasks, the cloud center uses transfer learning to compare the data attributes of the source domain and the target domain, and can adjust the parameters (such as adjusting the above allocation weight value) to more efficiently implement new model training.

[0498] 2. Hierarchical state feedback and orchestration collaboration mechanism: Step 1: The cloud center sends a global state table to all edge base stations; Step 2: AI tasks are decomposed to form a primary service function chain, and the primary service function chain orchestration information is uploaded to the cloud center for confirmation or modification. Step 3: The cloud center collects all primary service function chain information, and confirms and modifies the service function chain based on the latest global state table. Step 4: The UE's computing task is processed in the service function chain and the processing results are fed back to the UE. Step 5: All edge base stations feed back the local state table to the cloud center.

[0499] 3. Design of status table data reporting control packet: The local status table is inserted into the MAC layer transmission block as a MAC control information element and transmitted through the transmission channel. In the uplink transmission, the local status table information may include: the existing AI model (information) of the local edge base station, the remaining computing resources (information) of the local edge base station, the remaining storage resources (information) of the local edge base station, the physical location (information) of the local edge base station, and the connection status (information) between the local edge base station and other base stations; in the downlink transmission, the global status table may include the following information: the existing AI model (information) in each edge base station and the cloud center, the remaining computing resources (information) of each edge base station and the cloud center, the remaining storage resources (information) of each edge base station and the cloud center, the physical location (information) of each edge base station, and the mutual connection status (information) of each edge base station.

[0500] In summary, the solution provided by the present invention has the following advantages:

[0501] (1) This solution proposes a model training method that combines federated learning and transfer learning for service orchestration: By using the distributed learning architecture of federated learning, the transmission of model parameters replaces the transmission of original data (data used as model input), alleviating the problem of limited computing power of edge base stations, while greatly reducing data traffic in the network and ensuring data security. In addition, the use of transfer learning in the cloud center can more efficiently implement new model training.

[0502] (2) This scheme introduces a hierarchical state feedback and orchestration collaboration mechanism in the service function chain orchestration of the edge computing network: through the design and sharing of the global state table of the cloud center, each edge base station can perceive the global state information of the network, and use the game theory algorithm to achieve the optimal orchestration scheme of the service function chain.

[0503] (3) This solution designs a state table data reporting control package: the edge base stations’ own computing resource information, storage resource information, model resource information, location information, interconnection information, and other information are added to the interaction between edge base stations, so that through the sharing of the global state table, each edge base station in the network can make comprehensive considerations for the orchestration of the AI ​​service function chain.

[0504] The embodiment of the present application also provides an information processing device, which is applied to a first base station, such as Fig. 9 As shown, including:

[0505] A first processing module 91, configured to decompose an artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0506] A first sending module 92, configured to send primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks;

[0507] The first receiving module 93 is used to receive the first information sent by the cloud computing center device; the first information includes: the final service function chain formation decision information or confirmation information;

[0508] The first determination module 94 is used to determine a final service function chain according to the first information.

[0509] The information processing device provided by the embodiment of the present invention decomposes the artificial intelligence AI task of the terminal into at least two different AI sub-tasks; sends primary service function chain orchestration information to the cloud computing center device according to the at least two different AI sub-tasks; receives the first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information; determines the final service function chain according to the first information; can support the first base station and the cloud computing center device to collaborate to determine the AI ​​service function chain orchestration based on a global perspective (second information), thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0510] Among them, sending primary service function chain orchestration information to the cloud computing center device according to the at least two different AI subtasks includes: arranging the service position of each of the AI ​​subtasks according to the second information to form a primary service function chain, and obtaining the primary service function chain orchestration information; wherein the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of AI tasks; and sending the primary service function chain orchestration information to the cloud computing center device.

[0511] In an embodiment of the present invention, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0512] The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the beginning position of a MAC protocol data unit in the MAC layer transmission block.

[0513] Furthermore, the information processing device also includes: a second sending module, used to send local status information of the first base station to the cloud computing center device; wherein the local status information is related to the execution of AI tasks.

[0514] Among them, the local status information includes at least one of the following: information about the AI ​​model stored in the first base station; information about the remaining computing resources in the first base station; information about the storage resources in the first base station; information about the physical location of the first base station; information about the connection status of the first base station with other base stations.

[0515] In the embodiment of the present invention, the local state information is inserted into a MAC layer transmission block as a MAC control information element and is transmitted through a transmission channel.

[0516] Furthermore, the information processing device also includes: a second processing module, used to utilize the final service function chain to execute the AI ​​task and obtain a processing result; and a third sending module, used to send the processing result to the terminal.

[0517] In an embodiment of the present invention, the information processing device further includes: a third processing module, which is used to determine the training party of the AI ​​model corresponding to each of the AI ​​subtasks before using the final service function chain to execute the AI ​​task and obtain the processing result, and use the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station; a second receiving module, which is used to receive the final parameter information obtained by the cloud computing center device according to the initial parameter information; using the final service function chain to execute the AI ​​task and obtain the processing result, including: according to the final service function chain and the final parameter information, using the AI ​​model to execute the AI ​​task and obtain the processing result.

[0518] In an embodiment of the present invention, determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks includes: (1) when the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; (2) when the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determining whether the other base stations have the AI ​​model corresponding to the AI ​​subtask based on the second information; if so, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

[0519] Among them, the implementation embodiments of the information processing method on the first base station side mentioned above are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0520] The embodiment of the present invention also provides an information processing device, which is applied to cloud computing center equipment, such as Fig.10 As shown, including:

[0521] The third receiving module 101 is used to receive primary service function chain arrangement information sent by at least one first base station;

[0522] The second determination module 102 is used to determine whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements according to the primary service function chain arrangement information and the second information, and obtain a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of the AI ​​task;

[0523] A first acquisition module 103 is used to acquire first information corresponding to each of the first base stations according to the determination result; the first information includes: final service function chain formation decision information or confirmation information;

[0524] The fourth sending module 104 is used to send corresponding first information to each of the first base stations.

[0525] The information processing device provided in the embodiment of the present invention receives primary service function chain orchestration information sent by at least one first base station; determines whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements based on the primary service function chain orchestration information and the second information, and obtains a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of AI tasks; according to the determination result, obtains the first information corresponding to each of the first base stations; the first information includes: final service function chain formation decision information or confirmation information; sends the corresponding first information to each of the first base stations; can support the first base station and the cloud computing center device to collaborate to determine the AI ​​service function chain orchestration based on a global perspective (second information), thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0526] Wherein, obtaining the first information corresponding to each of the first base stations respectively according to the determination result includes: (1) when the determination result indicates that the computing resources of each base station on the service function chain can meet all current primary service function chain requirements, using the first information containing confirmation information as the first information corresponding to each of the first base stations respectively; (2) when the determination result indicates that the computing resources of each base station on the service function chain cannot meet all current primary service function chain requirements, updating at least one of the primary service function chain orchestration information to determine the first base station for which the primary service function chain is to be changed and the first base station for which the primary service function chain remains unchanged; using the first information containing target decision information as the first information corresponding to the first base station for which the primary service function chain is to be changed; using the first information containing confirmation information as the first information corresponding to the first base station for which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station for which the primary service function chain is to be changed.

[0527] Furthermore, the information processing device also includes: a fifth sending module, used to send the second information to each base station on the service function chain respectively; wherein the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0528] The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the beginning position of a MAC protocol data unit in the MAC layer transmission block.

[0529] Furthermore, the information processing device also includes: a fourth receiving module, used to receive local status information of the base station sent respectively by each base station on the service function chain; wherein the local status information is related to the execution of the AI ​​task; and a third processing module, used to obtain the second information based on the local status information.

[0530] Among them, the local status information includes at least one of the following: information about the AI ​​model stored in the base station; information about the remaining computing resources in the base station; information about the storage resources in the base station; information about the physical location of the base station; information about the connection status of the base station with other base stations.

[0531] In the embodiment of the present invention, the local state information is inserted into a MAC layer transmission block as a MAC control information element and is transmitted through a transmission channel.

[0532] Furthermore, the information processing device also includes: a second acquisition module, used to obtain initial parameter information of at least one AI model; a fourth processing module, used to perform classification and aggregation operations on the initial parameter information to obtain final parameter information of each AI model; and a sixth sending module, used to send the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0533] The obtaining of initial parameter information of at least one AI model includes at least one of the following: (1) receiving initial parameter information of the AI ​​model sent by at least one base station; (2) receiving an AI subtask sent by the first base station, training a corresponding AI model according to the AI ​​subtask, and obtaining the initial parameter information of the AI ​​model.

[0534] In an embodiment of the present invention, the training of the corresponding AI model according to the AI ​​subtask includes: using transfer learning to train the corresponding AI model according to the AI ​​subtask.

[0535] Among them, the implementation embodiments of the information processing method on the device side of the cloud computing center are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0536] The embodiment of the present invention further provides an information processing device, which is applied to a first base station, such as Fig.11 As shown, including:

[0537] A fifth processing module 111, configured to decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0538] The sixth processing module 112 is used to determine the training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station;

[0539] The fifth receiving module 113 is used to receive final parameter information obtained by the cloud computing center device according to the initial parameter information.

[0540] The information processing device provided by the embodiment of the present invention decomposes the artificial intelligence (AI) task of the terminal into at least two different AI subtasks; determines the training party of the AI ​​model corresponding to each of the AI ​​subtasks, and uses the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station; receives the final parameter information obtained by the cloud computing center device according to the initial parameter information; can avoid the problems of link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources of the cloud computing center, and large AI task processing delay caused by the completion of the training of all AI models in the cloud computing center; thereby further supporting the improvement of the orchestration effect, and supporting a better solution to the problem that the information processing scheme for the orchestration of the AI ​​service function chain in the prior art lacks a global perspective and leads to poor orchestration effect.

[0541] Wherein, determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks includes: (1) when the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; (2) when the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determining whether the other base stations have the AI ​​model corresponding to the AI ​​subtask based on second information; if so, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device; wherein, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of the AI ​​task.

[0542] In an embodiment of the present invention, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0543] In an embodiment of the present invention, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the starting position of a MAC protocol data unit in the MAC layer transmission block.

[0544] Among them, the implementation embodiments of the information processing method on the first base station side mentioned above are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0545] The embodiment of the present invention also provides an information processing device, which is applied to cloud computing center equipment, such as Fig.12 As shown, including:

[0546] A third acquisition module 121 is used to acquire initial parameter information of at least one AI model;

[0547] A seventh processing module 122 is used to perform classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0548] The seventh sending module 123 is used to send the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

[0549] The information processing device provided by the embodiment of the present invention obtains initial parameter information of at least one AI model; performs classification and aggregation operations on the initial parameter information to obtain final parameter information of each AI model; and sends the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model. This can support avoiding the problems of link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources in the cloud computing center, and large AI task processing delays when completing the training of all AI models in the cloud computing center. This can further support improving the orchestration effect and support better solving the problem that the information processing scheme for the orchestration of the AI ​​service function chain in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0550] The obtaining of initial parameter information of at least one AI model includes at least one of the following: (1) receiving initial parameter information of the AI ​​model sent by at least one base station; (2) receiving an AI subtask sent by a first base station, training a corresponding AI model according to the AI ​​subtask, and obtaining the initial parameter information of the AI ​​model.

[0551] In an embodiment of the present invention, the training of the corresponding AI model according to the AI ​​subtask includes: using transfer learning to train the corresponding AI model according to the AI ​​subtask.

[0552] Furthermore, the information processing device also includes: an eighth sending module, used to send second information to at least one base station respectively; wherein, the second information is global status information corresponding to the cloud computing center device; and the global status information is related to the execution of AI tasks.

[0553] Among them, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0554] In an embodiment of the present invention, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the starting position of a MAC protocol data unit in the MAC layer transmission block.

[0555] Among them, the implementation embodiments of the information processing method on the device side of the cloud computing center are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0556] The embodiment of the present invention further provides an information processing device, wherein the information processing device is a first base station, such as Fig.13 As shown, it includes: a processor 131 and a transceiver 132;

[0557] The processor 131 is used to decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0558] According to the at least two different AI subtasks, primary service function chain arrangement information is sent to the cloud computing center device through the transceiver 132;

[0559] Receiving, through the transceiver 132, first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information;

[0560] According to the first information, a final service function chain is determined.

[0561] The information processing device provided by the embodiment of the present invention decomposes the artificial intelligence AI task of the terminal into at least two different AI sub-tasks; sends primary service function chain orchestration information to the cloud computing center device according to the at least two different AI sub-tasks; receives the first information sent by the cloud computing center device; the first information includes: final service function chain formation decision information or confirmation information; determines the final service function chain according to the first information; can support the first base station and the cloud computing center device to collaborate to determine the AI ​​service function chain orchestration based on a global perspective (second information), thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0562] Among them, according to the at least two different AI subtasks, sending the primary service function chain orchestration information to the cloud computing center device through the transceiver, includes: according to the second information, arranging the service position of each of the AI ​​subtasks to form a primary service function chain, and obtaining the primary service function chain orchestration information; wherein, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of AI tasks; and sending the primary service function chain orchestration information to the cloud computing center device through the transceiver.

[0563] In an embodiment of the present invention, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0564] The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the beginning position of a MAC protocol data unit in the MAC layer transmission block.

[0565] Furthermore, the processor is also used to: send local status information of the first base station to the cloud computing center device through the transceiver; wherein the local status information is related to the execution of the AI ​​task.

[0566] Among them, the local status information includes at least one of the following: information about the AI ​​model stored in the first base station; information about the remaining computing resources in the first base station; information about the storage resources in the first base station; information about the physical location of the first base station; information about the connection status of the first base station with other base stations.

[0567] In the embodiment of the present invention, the local state information is inserted into a MAC layer transmission block as a MAC control information element and is transmitted through a transmission channel.

[0568] Furthermore, the processor is also used to: utilize the final service function chain to execute the AI ​​task and obtain a processing result; and send the processing result to the terminal through the transceiver.

[0569] In an embodiment of the present invention, the processor is also used to: determine the training party of the AI ​​model corresponding to each of the AI ​​subtasks before using the final service function chain to execute the AI ​​task and obtain the processing result, and use the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station; through the transceiver, receive the final parameter information obtained by the cloud computing center device according to the initial parameter information; using the final service function chain to execute the AI ​​task and obtain the processing result, including: according to the final service function chain and the final parameter information, using the AI ​​model to execute the AI ​​task and obtain the processing result.

[0570] In an embodiment of the present invention, determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks includes: (1) when the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; (2) when the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determining whether the other base stations have the AI ​​model corresponding to the AI ​​subtask based on the second information; if so, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

[0571] Among them, the implementation embodiments of the information processing method on the first base station side mentioned above are all applicable to the embodiments of the information processing device and can achieve the same technical effect.

[0572] The embodiment of the present invention further provides an information processing device, wherein the information processing device is a cloud computing center device, such as Fig.14As shown, it includes: a processor 141 and a transceiver 142;

[0573] The processor 141 is configured to receive primary service function chain arrangement information sent by at least one first base station through the transceiver 142;

[0574] According to the primary service function chain arrangement information and the second information, determining whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements, and obtaining a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of the AI ​​task;

[0575] According to the determination result, first information corresponding to each of the first base stations is obtained; the first information includes: final service function chain formation decision information or confirmation information;

[0576] The corresponding first information is sent to each of the first base stations through the transceiver 142.

[0577] The information processing device provided in the embodiment of the present invention receives the primary service function chain orchestration information sent by at least one first base station; determines whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements based on the primary service function chain orchestration information and the second information, and obtains a determination result; the second information is the global status information corresponding to the cloud computing center device; the global status information is related to the execution of AI tasks; according to the determination result, obtains the first information corresponding to each of the first base stations; the first information includes: the final service function chain formation decision information or confirmation information; sends the corresponding first information to each of the first base stations; can support the first base station and the cloud computing center device to collaborate to determine the AI ​​service function chain orchestration based on a global perspective (second information), thereby improving the orchestration effect, and well solves the problem that the information processing scheme for AI service function chain orchestration in the prior art lacks a global perspective, resulting in poor orchestration effect.

[0578] Wherein, obtaining the first information corresponding to each of the first base stations respectively according to the determination result includes: (1) when the determination result indicates that the computing resources of each base station on the service function chain can meet all current primary service function chain requirements, using the first information containing confirmation information as the first information corresponding to each of the first base stations respectively; (2) when the determination result indicates that the computing resources of each base station on the service function chain cannot meet all current primary service function chain requirements, updating at least one of the primary service function chain orchestration information to determine the first base station for which the primary service function chain is to be changed and the first base station for which the primary service function chain remains unchanged; using the first information containing target decision information as the first information corresponding to the first base station for which the primary service function chain is to be changed; using the first information containing confirmation information as the first information corresponding to the first base station for which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station for which the primary service function chain is to be changed.

[0579] Furthermore, the processor is also used to: send the second information to each base station on the service function chain through the transceiver; wherein the second information includes at least one of the following: information on AI models stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is connected to the cloud computing center equipment respectively.

[0580] The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the beginning position of a MAC protocol data unit in the MAC layer transmission block.

[0581] Furthermore, the processor is also used to: receive, through the transceiver, local status information of the base station sent respectively by each base station on the service function chain; wherein the local status information is related to the execution of the AI ​​task; and obtain the second information based on the local status information.

[0582] Among them, the local status information includes at least one of the following: information about the AI ​​model stored in the base station; information about the remaining computing resources in the base station; information about the storage resources in the base station; information about the physical location of the base station; information about the connection status of the base station with other base stations.

[0583] In the embodiment of the present invention, the local state information is inserted into a MAC layer transmission block as a MAC control information element and is transmitted through a transmission channel.

[0584] Furthermore, the processor is also used to: obtain initial parameter information of at least one AI model; perform classification and aggregation operations on the initial parameter information to obtain final parameter information of each AI model; and send the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model through the transceiver.

[0585] Among them, the obtaining of initial parameter information of at least one AI model includes at least one of the following: (1) receiving initial parameter information of the AI ​​model sent by at least one base station through the transceiver; (2) receiving an AI subtask sent by the first base station through the transceiver, training the corresponding AI model according to the AI ​​subtask, and obtaining the initial parameter information of the AI ​​model.

[0586] In an embodiment of the present invention, the training of the corresponding AI model according to the AI ​​subtask includes: using transfer learning to train the corresponding AI model according to the AI ​​subtask.

[0587] Among them, the implementation embodiments of the information processing method on the cloud computing center device side are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0588] The embodiment of the present invention further provides an information processing device, wherein the information processing device is a first base station, such as Fig.15 As shown, it includes: a processor 151 and a transceiver 152;

[0589] The processor 151 is used to decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks;

[0590] Determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain initial parameter information of the AI ​​model; the training party is the first base station, a cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station;

[0591] The final parameter information obtained by the cloud computing center device according to the initial parameter information is received through the transceiver 152.

[0592] The information processing device provided by the embodiment of the present invention decomposes the artificial intelligence (AI) task of the terminal into at least two different AI subtasks; determines the training party of the AI ​​model corresponding to each of the AI ​​subtasks, and uses the training party to obtain the initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device connected to the first base station, or other base stations under the cloud computing center device except the first base station; receives the final parameter information obtained by the cloud computing center device according to the initial parameter information; can avoid the problems of link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources of the cloud computing center, and large AI task processing delay caused by the training of all AI models in the cloud computing center; thereby further supporting the improvement of the orchestration effect, and supporting a better solution to the problem that the information processing scheme for the orchestration of the AI ​​service function chain in the prior art lacks a global perspective and leads to poor orchestration effect.

[0593] Wherein, determining the training party of the AI ​​model corresponding to each of the AI ​​subtasks includes: (1) when the AI ​​model corresponding to the AI ​​subtask exists locally in the first base station, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; (2) when the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determining whether the other base stations have the AI ​​model corresponding to the AI ​​subtask based on second information; if so, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determining that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device; wherein, the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; the global status information is related to the execution of the AI ​​task.

[0594] In an embodiment of the present invention, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0595] In an embodiment of the present invention, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the starting position of a MAC protocol data unit in the MAC layer transmission block.

[0596] Among them, the implementation embodiments of the information processing method on the first base station side mentioned above are all applicable to the embodiments of the information processing device and can achieve the same technical effect.

[0597] The embodiment of the present invention further provides an information processing device, wherein the information processing device is a cloud computing center device, such as Fig.16 As shown, it includes: a processor 161 and a transceiver 162;

[0598] The processor 161 is used to obtain initial parameter information of at least one AI model;

[0599] Performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models;

[0600] The final parameter information of the AI ​​model is sent to the base station corresponding to the AI ​​model through the transceiver 162.

[0601] The information processing device provided by the embodiment of the present invention obtains the initial parameter information of at least one AI model; performs classification and aggregation operations on the initial parameter information to obtain final parameter information of each AI model; and sends the final parameter information of the AI ​​model to the base station corresponding to the AI ​​model. It can support avoiding the problems of link congestion and poor data security caused by uploading training data, occupying a large amount of computing resources in the cloud computing center, and large AI task processing delay when completing the training of all AI models in the cloud computing center; thereby further supporting the improvement of orchestration effects and supporting better solutions to the problem that the information processing solutions for the orchestration of AI service function chains in the prior art lack a global perspective, resulting in poor orchestration effects.

[0602] Among them, the obtaining of initial parameter information of at least one AI model includes at least one of the following: (1) receiving initial parameter information of the AI ​​model sent by at least one base station through the transceiver; (2) receiving an AI subtask sent by a first base station through the transceiver, training a corresponding AI model according to the AI ​​subtask, and obtaining the initial parameter information of the AI ​​model.

[0603] In an embodiment of the present invention, the training of the corresponding AI model according to the AI ​​subtask includes: using transfer learning to train the corresponding AI model according to the AI ​​subtask.

[0604] Furthermore, the processor is also used to: send second information to at least one base station respectively through the transceiver; wherein the second information is global status information corresponding to the cloud computing center device; and the global status information is related to the execution of AI tasks.

[0605] Among them, the second information includes at least one of the following: information on AI models already stored in each base station and cloud computing center equipment; information on remaining computing resources in each base station and cloud computing center equipment; information on remaining storage resources in each base station and cloud computing center equipment; information on the physical location of each base station; information on the mutual connection status of each base station; wherein each base station is respectively connected to the cloud computing center equipment.

[0606] In an embodiment of the present invention, the second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; wherein the second information is located at the starting position of a MAC protocol data unit in the MAC layer transmission block.

[0607] Among them, the implementation embodiments of the information processing method on the cloud computing center device side are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0608] An embodiment of the present invention also provides an information processing device, including a memory, a processor, and a program stored in the memory and executable on the processor; when the processor executes the program, the information processing method on the first base station side or the cloud computing center device side is implemented.

[0609] Among them, the implementation embodiments of the information processing method on the first base station side or the cloud computing center device side are all applicable to the embodiments of the information processing device and can achieve the same technical effects.

[0610] An embodiment of the present invention further provides a readable storage medium on which a program is stored. When the program is executed by a processor, the steps in the information processing method on the first base station side or the cloud computing center device side are implemented.

[0611] Among them, the implementation embodiments of the information processing method on the first base station side or the cloud computing center device side are all applicable to the embodiments of the readable storage medium and can achieve the same technical effect.

[0612] It should be noted that many functional components described in this specification are referred to as modules in order to more particularly emphasize the independence of their implementation methods.

[0613] In the embodiment of the present invention, module can be implemented with software so that it can be executed by various types of processors. For example, an executable code module of an identification can include one or more physical or logical blocks of computer instructions, for example, it can be constructed as an object, process or function. Nevertheless, the executable code of the identified module does not need to be physically located together, but can include different instructions stored in different positions, and when these instructions are logically combined together, it constitutes a module and realizes the specified purpose of the module.

[0614] In fact, executable code module can be a single instruction or many instructions, and can even be distributed on a plurality of different code segments, distributed among different programs, and distributed across a plurality of memory devices. Similarly, operating data can be identified in the module, and can be implemented and organized in the data structure of any appropriate type according to any appropriate form. The operating data can be collected as a single data set, or can be distributed in different locations (including on different storage devices), and can only be present on a system or network as an electronic signal at least in part.

[0615] When a module can be implemented by software, considering the level of existing hardware technology, a person skilled in the art can build a corresponding hardware circuit to implement the corresponding function of the module that can be implemented by software without considering the cost. The hardware circuit includes a conventional very large scale integration (VLSI) circuit or gate array and existing semiconductors such as logic chips, transistors, or other discrete components. The module can also be implemented by a programmable hardware device, such as a field programmable gate array, a programmable array logic, a programmable logic device, etc.

[0616] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary personnel in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An information processing method, applied to a first base station, characterized in that: include: Decompose the terminal's artificial intelligence (AI) task into at least two different AI subtasks; Sending primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks; Receiving first information sent by the cloud computing center device; The first information includes: final service function chain formation decision information or confirmation information; According to the first information, a final service function chain is determined.

2. The information processing method according to claim 1, characterized in that: The sending primary service function chain arrangement information to the cloud computing center device according to the at least two different AI subtasks includes: Arrange the service position of each of the AI ​​subtasks according to the second information to form a primary service function chain, and obtain primary service function chain arrangement information; wherein the second information is the global status information corresponding to the cloud computing center device to which the first base station belongs; and the global status information is related to the execution of the AI ​​task; The primary service function chain orchestration information is sent to the cloud computing center device.

3. The information processing method according to claim 2, characterized in that: The second information includes at least one of the following: Information about AI models stored in each base station and cloud computing center equipment; Information about remaining computing resources in each base station and cloud computing center equipment; Information about remaining storage resources in each base station and cloud computing center equipment; Information about the physical location of each base station; Information about the interconnection status of each base station; Wherein, each base station is connected to the cloud computing center equipment respectively.

4. The information processing method according to claim 2, characterized in that: The second information is inserted into a MAC layer transmission block as a media access control MAC control information element and transmitted through a transmission channel; The second information is located at the beginning of the MAC protocol data unit in the MAC layer transmission block.

5. The information processing method according to claim 1, characterized in that: Also includes: Sending local status information of the first base station to the cloud computing center device; The local status information is related to executing the AI ​​task.

6. The information processing method according to claim 5, characterized in that: The local status information includes at least one of the following: Information about the AI ​​model stored by the first base station; Information about remaining computing resources in the first base station; Information about resources stored in the first base station; information about the physical location of the first base station; Information about the connection between the first base station and other base stations.

7. The information processing method according to claim 5, characterized in that: The local state information is inserted into the MAC layer transmission block as a MAC control information element and transmitted through the transmission channel.

8. The information processing method according to claim 2, characterized in that: Also includes: Utilizing the final service function chain, executing the AI ​​task and obtaining a processing result; The processing result is sent to the terminal.

9. The information processing method according to claim 8, characterized in that: Before using the final service function chain to execute the AI ​​task and obtain the processing result, the process further includes: Determine a training party of the AI ​​model corresponding to each of the AI ​​subtasks, and use the training party to obtain initial parameter information of the AI ​​model; the training party is the first base station, the cloud computing center device, or other base stations under the cloud computing center device except the first base station; Receiving final parameter information obtained by the cloud computing center device according to the initial parameter information; The using the final service function chain to execute the AI ​​task and obtain the processing result includes: According to the final service function chain and the final parameter information, the AI ​​model is used to execute the AI ​​task and obtain the processing result.

10. The information processing method according to claim 9, characterized in that: The method of determining the training method of the AI ​​model corresponding to each of the AI ​​subtasks includes: In a case where the first base station locally has an AI model corresponding to the AI ​​subtask, determining that a training party of the AI ​​model corresponding to the AI ​​subtask is the first base station; When the AI ​​model corresponding to the AI ​​subtask does not exist locally in the first base station, determine whether the AI ​​model corresponding to the AI ​​subtask exists in other base stations based on the second information; if so, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the other base station; if not, determine that the training party of the AI ​​model corresponding to the AI ​​subtask is the cloud computing center device.

11. An information processing method, applied to cloud computing center equipment, characterized in that: include: Receiving primary service function chain arrangement information sent by at least one first base station; Determine whether the computing resources of each base station on the service function chain meet all current requirements of the primary service function chain according to the primary service function chain arrangement information and the second information, and obtain a determination result; the second information is the global status information corresponding to the cloud computing center device; The global state information is related to the execution of the AI ​​task; According to the determination result, obtaining first information respectively corresponding to each of the first base stations; The first information includes: final service function chain formation decision information or confirmation information; Send corresponding first information to each of the first base stations.

12. The information processing method according to claim 11, characterized in that: The acquiring, according to the determination result, first information respectively corresponding to each of the first base stations includes: In a case where the determination result indicates that the computing resources of each base station on the service function chain can meet the current requirements of all primary service function chains, the first information including the confirmation information is used as the first information corresponding to each of the first base stations; When the determination result indicates that the computing resources of each base station on the service function chain cannot meet the current needs of all primary service function chains, at least one of the primary service function chain orchestration information is updated to determine the first base station in which the primary service function chain is changed and the first base station in which the primary service function chain remains unchanged; the first information containing the target decision information is used as the first information corresponding to the first base station in which the primary service function chain is changed; the first information containing the confirmation information is used as the first information corresponding to the first base station in which the primary service function chain remains unchanged; wherein the target decision information includes the final service function chain formation decision information corresponding to the first base station in which the primary service function chain is changed.

13. The information processing method according to claim 11, characterized in that: Also includes: Sending the second information to each base station on the service function chain respectively; The second information includes at least one of the following: Information about AI models stored in each base station and cloud computing center equipment; Information about remaining computing resources in each base station and cloud computing center equipment; Information about remaining storage resources in each base station and cloud computing center equipment; Information about the physical location of each base station; Information about the interconnection status of each base station; Wherein, each base station is connected to the cloud computing center equipment respectively.

14. The information processing method according to claim 11, characterized in that: Also includes: Receiving local state information of the base station respectively sent by each base station on the service function chain; wherein the local state information is related to the execution of the AI ​​task; The second information is obtained according to the local state information.

15. The information processing method according to claim 14, characterized in that: The local status information includes at least one of the following: Information about the AI ​​model stored in the base station; Information about remaining computing resources in the base station; Information about resources stored in the base station; Information about the physical location of the base station; Information about the connection between the base station and other base stations.

16. The information processing method according to claim 11, characterized in that: Also includes: Obtaining initial parameter information of at least one AI model; Performing classification and aggregation operations on the initial parameter information to obtain final parameter information of each of the AI ​​models; Send final parameter information of the AI ​​model to the base station corresponding to the AI ​​model.

17. The information processing method according to claim 16, characterized in that: The obtaining of initial parameter information of at least one AI model includes at least one of the following: Receiving initial parameter information of the AI ​​model sent by at least one base station; Receive the AI ​​subtask sent by the first base station, train the corresponding AI model according to the AI ​​subtask, and obtain initial parameter information of the AI ​​model.

18. The information processing method according to claim 17, characterized in that: The step of training a corresponding AI model according to the AI ​​subtask includes: By using transfer learning, the corresponding AI model is trained according to the AI ​​subtask.

19. An information processing device, applied to a first base station, characterized in that: include: A first processing module, configured to decompose an artificial intelligence (AI) task of the terminal into at least two different AI subtasks; A first sending module, configured to send primary service function chain arrangement information to a cloud computing center device according to the at least two different AI subtasks; A first receiving module, used to receive first information sent by the cloud computing center device; The first information includes: final service function chain formation decision information or confirmation information; The first determining module is used to determine a final service function chain according to the first information.

20. An information processing device, applied to cloud computing center equipment, characterized in that: include: A third receiving module, configured to receive primary service function chain arrangement information sent by at least one first base station; A second determination module is used to determine whether the computing resources of each base station on the service function chain meet all current primary service function chain requirements according to the primary service function chain arrangement information and the second information, and obtain a determination result; the second information is the global status information corresponding to the cloud computing center device; The global state information is related to the execution of the AI ​​task; A first acquisition module, configured to acquire first information respectively corresponding to each of the first base stations according to the determination result; The first information includes: final service function chain formation decision information or confirmation information; The fourth sending module is used to send the corresponding first information to each of the first base stations.

21. An information processing device, the information processing device being a first base station, characterized in that: include: processor and transceiver; The processor is used to decompose the artificial intelligence AI task of the terminal into at least two different AI subtasks; According to the at least two different AI subtasks, primary service function chain arrangement information is sent to a cloud computing center device through the transceiver; Receiving, through the transceiver, first information sent by the cloud computing center device; The first information includes: final service function chain formation decision information or confirmation information; According to the first information, a final service function chain is determined.

22. An information processing device, the information processing device being a cloud computing center device, characterized in that: include: processor and transceiver; The processor is configured to receive primary service function chain arrangement information sent by at least one first base station through the transceiver; Determine whether the computing resources of each base station on the service function chain meet all current requirements of the primary service function chain according to the primary service function chain arrangement information and the second information, and obtain a determination result; the second information is the global status information corresponding to the cloud computing center device; The global state information is related to the execution of the AI ​​task; According to the determination result, obtaining first information respectively corresponding to each of the first base stations; The first information includes: final service function chain formation decision information or confirmation information; The corresponding first information is sent to each of the first base stations through the transceiver.

23. An information processing device, comprising a memory, a processor, and a program stored in the memory and executable on the processor; characterized in that: When the processor executes the program, the information processing method according to any one of claims 1 to 18 is implemented.

24. A readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the information processing method according to any one of claims 1 to 18 are implemented.

Images