Communication method, communication device, communication system, storage medium, and program product

Abstract

Provided in the embodiments of the present disclosure are a communication method, a communication device, a communication system, a storage medium, and a program product. The method is executed by means of a first node, and comprises: sending a first message to at least one terminal and / or at least one second node, wherein the first message is used by the terminal and / or the second node to perform federated learning on the basis of a first model, a first node is deployed in a core network or an access network, and the second node is deployed in the access network. The technical solution provided in the embodiments of the present disclosure facilitates the use of a federated learning method in a communication network, and implements federated learning by using network resources of the communication network. Moreover, a hierarchical federated learning architecture is implemented by using a first node, a second node and a terminal in the communication network, such that not only can a training delay be shortened, but also network-side transmission overheads can be reduced.

Description

Communication methods, communication equipment, communication systems, storage media and software products Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a communication method, communication device, communication system, storage medium, and program product. Background Technology

[0002] Federated learning (FL) is a distributed machine learning approach. In this approach, multiple participants can build a collaborative machine learning model without sharing data, thereby addressing key issues such as data privacy, data security, data access permissions, and access to heterogeneous data.

[0003] Summary of the Invention

[0004] This disclosure provides a communication method, communication device, communication system, storage medium, and program product.

[0005] According to a first aspect of the present disclosure, a communication method is provided, wherein the method is performed by a first node, the method comprising: sending a first message to at least one terminal and / or at least one second node; the first message being used by the terminal and / or the second node to perform federated learning based on a first model; the first node being deployed in a core network or an access network, and the second node being deployed in the access network.

[0006] According to a second aspect of the present disclosure, a communication method is provided, wherein the method is executed by a second node, the method comprising: receiving a first message sent by a first node; the first message being used by the second node to perform federated learning based on a first model; the first node being deployed in a core network or an access network, and the second node being deployed in the access network; sending the first message to at least one terminal; the first message being further used by the terminal to perform federated learning based on the first model.

[0007] According to a third aspect of the present disclosure, a communication method is provided, wherein the method is executed by a terminal, the method comprising: receiving a first message sent by a first node or a second node; the first message being used by the terminal to perform federated learning based on a first model; the first node being deployed in a core network or an access network, and the second node being deployed in the access network.

[0008] According to a fourth aspect of the present disclosure, a communication method is provided, wherein the method is performed by a communication system, the method comprising: a first node sending a first message to at least one terminal and / or at least one second node; the first message being used by the terminal and / or the second node to perform federated learning based on a first model; the first node being deployed in a core network or an access network, and the second node being deployed in the access network.

[0009] According to a fifth aspect of the present disclosure, a communication device is provided, wherein the communication device performs the communication method provided by the first aspect, the second aspect, or the third aspect.

[0010] According to a sixth aspect of the present disclosure, a communication system is provided, wherein the communication system includes a terminal, a first node, and a second node, the first node being configured to implement the communication method provided in the first aspect, the second node being configured to implement the communication method provided in the second aspect, and the terminal being configured to implement the communication method provided in the third aspect.

[0011] According to a seventh aspect of the present disclosure, a storage medium is provided, wherein the storage medium stores instructions that, when executed on a communication device, cause the communication device to perform the communication method provided by the first aspect, the second aspect, or the third aspect.

[0012] According to an eighth aspect of the present disclosure, a program product is provided that, when executed by a communication device, causes the communication device to perform the communication method provided by the first aspect, the second aspect, or the third aspect.

[0013] The technical solution provided in this disclosure is beneficial for applying federated learning methods in communication networks, utilizing the network resources of the communication network to achieve federated learning. Furthermore, by utilizing the first node, second node, and terminal in the communication network to implement a hierarchical federated learning architecture, it can not only reduce training latency but also reduce network-side transmission overhead.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the embodiments of this disclosure. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the embodiments of the invention.

[0016] Figure 1A is a schematic diagram of the architecture of a communication system according to an exemplary embodiment;

[0017] Figure 1B is a schematic diagram of a split architecture of an access network device according to an exemplary embodiment;

[0018] Figure 1C is a schematic diagram of a split architecture for an access network device according to an exemplary embodiment;

[0019] Figure 1D is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment;

[0020] Figure 1E is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment;

[0021] Figure 1F is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment;

[0022] Figure 1G is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment;

[0023] Figure 1H is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment;

[0024] Figure 1I is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment;

[0025] Figure 1J is a schematic diagram of a protocol stack structure in a communication system according to an exemplary embodiment;

[0026] Figure 1K is a schematic diagram illustrating a protocol stack structure of a service-oriented interface according to an exemplary embodiment;

[0027] Figure 1L is a schematic diagram of a protocol stack structure in a communication system according to an exemplary embodiment;

[0028] Figure 1M is a schematic diagram of a protocol stack structure in a communication system according to an exemplary embodiment.

[0029] Figure 2A is an interactive schematic diagram of a communication method according to an exemplary embodiment;

[0030] Figure 2B is a schematic diagram of an interaction of a communication method according to an exemplary embodiment;

[0031] Figure 3 is an interactive schematic diagram of a communication method according to an exemplary embodiment;

[0032] Figure 4 is a flowchart illustrating a hierarchical federated learning method according to an exemplary embodiment;

[0033] Figure 5A is a schematic diagram of the structure of a network device according to an exemplary embodiment;

[0034] Figure 5B is a schematic diagram of the structure of a network device according to an exemplary embodiment;

[0035] Figure 5C is a schematic diagram of the structure of a terminal according to an exemplary embodiment;

[0036] Figure 6A is a schematic diagram of the structure of a communication device according to an exemplary embodiment;

[0037] Figure 6B is a schematic diagram of the structure of a chip according to an exemplary embodiment. Detailed Implementation

[0038] This disclosure provides a communication method, communication device, communication system, storage medium, and program product.

[0039] In a first aspect, embodiments of this disclosure provide a communication method, wherein the method is executed by a first node, and the method includes: sending a first message to at least one terminal and / or at least one second node; the first message is used by the terminal and / or the second node to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network.

[0040] The above embodiments utilize a first node deployed in the core network or access network to send a first message to at least one terminal or at least one second node, requesting the terminal and / or the second node deployed in the access network to perform federated learning based on the first model. This facilitates the application of federated learning methods in communication networks, leveraging network resources for federated learning. Furthermore, utilizing the first node, second node, and terminal within the communication network to implement a hierarchical federated learning architecture not only reduces training latency but also lowers network-side transmission overhead.

[0041] In conjunction with some embodiments of the first aspect, in some embodiments, sending a first message to at least one second node includes one of the following: sending a first control plane signaling to at least one second node, the first control plane signaling including the first message; the first node is deployed in a core network; sending the first signaling to at least one second node through a first interface, the first signaling including the first message; the first node is deployed in an access network, the first interface being a communication interface between access network devices; sending the first message to at least one second node through a service-oriented interface; the first node is deployed in a core network or an access network, and the access network supports a service-oriented interface.

[0042] In the above embodiments, when the first node sends a first message to at least one second node, different implementation methods for transmitting the first message between the first node and the second node are provided for different deployments of the first node in the access network and different deployments of the second node in the access network, thereby realizing the application of the federated learning method in the communication network.

[0043] In conjunction with some embodiments of the first aspect, in some embodiments, the first control plane signaling is transmitted based on a first protocol layer, which is a control plane-based protocol layer and is related to artificial intelligence (AI).

[0044] The above embodiments introduce a first protocol layer related to AI to realize control plane signaling transmission between the core network and the access network. This enables the transmission of the first message using control plane signaling between the core network and the access network, even when the first node is deployed in the core network.

[0045] In conjunction with some embodiments of the first aspect, in some embodiments, sending a first signaling message to at least one second node via a first interface includes one of the following: sending the first signaling message to at least one second node via an Xn interface; the first node and the second node are deployed in different access network devices; sending the first signaling message to at least one second node via an F1 interface; the first node is deployed in a Central Unit (CU) and the second node is deployed in a Distributed Unit (DU); sending the first signaling message to at least one second node via the F1-C interface of the F1 control plane; the first node is deployed in the CU Control Plane (CU-CP) and the second node is deployed in the DU; sending the first signaling message to at least one second node via an E1 interface; the first node is deployed in the CU-CP and the second node is deployed in the CU User Plane (CU-UP); sending the first signaling message to at least one second node via the F1-U interface of the F1 user plane; the first node is deployed in the CU-UP and the second node is deployed in the DU.

[0046] In the above embodiments, when the first node sends a first message to at least one second node, the first message transmission between the first node and the second node is realized by utilizing different interfaces between access network devices, depending on the different deployments of the first node in the access network and the different deployments of the second node in the access network, thereby realizing the application of the federated learning method in the communication network.

[0047] In conjunction with some embodiments of the first aspect, in some embodiments, sending a first message to at least one terminal includes one of the following: sending a second control plane signaling to at least one terminal, the second control plane signaling including the first message; a first node being deployed in a core network; sending user plane signaling to at least one terminal, the user plane signaling including the first message; a first node being deployed in a core network; sending a second signaling to at least one terminal through a second interface, the second signaling including the first message; a first node being deployed in an access network, the second interface being a communication interface between access network equipment and a terminal.

[0048] In the above embodiments, when the first node sends a first message to at least one terminal, different implementation methods for transmitting the first message between the first node and the terminal are provided for different deployments of the first node in the core network or the access network, thereby realizing the application of the federated learning method in the communication network.

[0049] In conjunction with some embodiments of the first aspect, in some embodiments, the second control plane signaling is transmitted based on the first protocol layer; the first protocol layer is a control plane-based protocol layer and is associated with AI; and / or, the user plane signaling is transmitted based on the second protocol layer; the second protocol layer is a user plane-based protocol layer and is associated with AI.

[0050] The above embodiments introduce a first protocol layer and / or a second protocol layer related to AI, utilizing the first protocol layer to implement control plane signaling transmission between the core network and the terminal. This enables the transmission of a first message using control plane signaling between the core network and the terminal when the first node is deployed in the core network. And / or, the second protocol layer is used to implement user plane signaling transmission between the core network and the terminal. This enables the transmission of a first message using user plane signaling between the core network and the terminal when the first node is deployed in the core network.

[0051] In conjunction with some embodiments of the first aspect, in some embodiments, the first message includes at least one of the following: first indication information for indicating a first model; second indication information for indicating a task identifier of a first task; the first task being a federated learning task based on the first model; and third indication information for indicating the training rounds of the first model in the first task.

[0052] In the above embodiments, the first message may include at least one of a first instruction, a second instruction, and a third instruction, wherein the instruction is information required by the terminal and / or the second node to perform federated learning based on the first model. The first node enables it to utilize the network resources of the terminal and / or the second node to perform the federated learning task by sending at least one of the aforementioned instruction to the terminal and / or the second node.

[0053] In conjunction with some embodiments of the first aspect, in some embodiments, the first message sent by the first node to at least one second node further includes fourth indication information for indicating a subtask performed by at least one terminal associated with the second node during a training round, wherein the at least one terminal associated with the second node accesses the network through the second node; or, the first message sent by the first node to at least one terminal further includes fifth indication information for indicating a subtask performed by the terminal during a training round.

[0054] In the above embodiments, when the first node sends a first message to at least one second node, the first message further includes fourth indication information, so that at least one second node can determine the terminal participating in federated learning within a training round based on the first message. When the first node sends a first message to at least one terminal, the first message further includes fifth indication information, so that the fifth indication information can be used to distinguish the subtasks executed by different terminals during subsequent federated learning.

[0055] In conjunction with some embodiments of the first aspect, in some embodiments, the first message is used by the terminal and / or the second node to perform federated learning based on the first model, including at least one of the following: the first message is used by at least one terminal to perform a training operation based on the first model; the first message is used by the second node to perform a first aggregation operation based on the first model parameters sent by at least one next-level device to obtain second model parameters; the second model parameters are used at least by the terminal to update the first model to be trained, and / or, the first node determines a third model; the third model is the learning result of federated learning based on the first model.

[0056] In the above embodiments, the first node sends a first message to enable at least one terminal to perform a training operation based on the first model, and at least one second node performs a first aggregation operation based on the first model parameters sent by the next-level device, thereby utilizing the network resources of the terminals and the second node in the communication network to perform the corresponding processing in federated learning. Furthermore, the second model parameters obtained by the second node performing the first aggregation operation can be used by the terminal to update the first model to be trained, thereby realizing multi-round federated learning in the communication network; and / or, the second model parameters can be used by the first node to determine a third model to obtain the learning result of federated learning based on the first model.

[0057] In conjunction with some embodiments of the first aspect, in some embodiments, the first model parameter is used to indicate the model parameters of a trained first model; the next-level device of the second node is a terminal; or, the first model parameter is used to indicate the aggregation result of the first aggregation operation; the aggregation result is determined by the model parameters of at least one trained first model; the next-level device of the second node is another second node.

[0058] In the above embodiments, depending on the different deployment scenarios of the second node in the federated learning architecture, when the next-level device of the second node is a terminal, the second node performs a first aggregation operation based on the model parameters of the trained first model sent by the terminal to obtain the second model parameters. When the next-level device of the second node is another second node, the second node performs a first aggregation operation based on the operation result of the first aggregation operation sent by the other second node to obtain the second model parameters.

[0059] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes: receiving first information sent by at least one second node, the first information including second model parameters; performing a second aggregation operation based on at least one second model parameter to obtain model parameters of a third model.

[0060] In the above embodiments, the first node can receive first information containing second model parameters sent by at least one second node, so as to aggregate the at least one second model parameter to obtain a third model parameter and complete the federated learning task.

[0061] In conjunction with some embodiments of the first aspect, in some embodiments, the first message includes: time information, which is used by the terminal and / or the second node to determine the time for executing the first task; the first task is a federated learning task based on the first model.

[0062] In the above embodiments, the first node may send time information to at least one terminal and / or at least one second node, so that the terminal and / or the second node can determine the time to execute the first task based on the time information, thereby scheduling network resources according to the time to complete the first task.

[0063] In conjunction with some embodiments of the first aspect, in some embodiments, the time information includes at least one of the following: first time information, used to indicate the deadline for the terminal to complete the training operation within a training round of the first task; second time information, used to indicate the deadline for the second node to complete the first aggregation operation within a training round of the first task.

[0064] In the above embodiments, the time information may include at least one of first time information and second time information, thereby using the first time information and second time information to indicate to the terminal and / or the second node the deadline for executing the first task, so that the terminal and / or the second node complete the first task before the corresponding deadline, thereby improving the time efficiency of the terminal and the second node in executing the first task.

[0065] In conjunction with some embodiments of the first aspect, in some embodiments, the first node includes at least one of the following: a Network Data Analytics Function (NWDAF); and a Data Collection Coordination Function (DCCF).

[0066] In the above embodiments, since NWDAF and DCCF themselves have data management capabilities, they can collect data from other network functions. Therefore, this embodiment reuses NWDAF and DCCF to perform the functions of the first node, so that the first node can obtain federated learning-related information from the second node and / or the terminal.

[0067] Secondly, embodiments of this disclosure provide a communication method, wherein the method is executed by a second node, the method comprising: receiving a first message sent by a first node; the first message being used by the second node to perform federated learning based on a first model; the first node being deployed in a core network or an access network, and the second node being deployed in the access network; sending the first message to at least one terminal; the first message being further used by the terminal to perform federated learning based on the first model.

[0068] The above embodiments utilize a second node deployed in the access network to receive a first message sent by a first node deployed in the access network or core network, and then send the first message to a terminal, enabling the second node and / or the terminal to perform federated learning based on the first model. Thus, on the one hand, by applying the federated learning method to a communication network, federated learning is achieved using the network resources of the communication network. On the other hand, by utilizing the first node, second node, and terminal in the communication network to implement a hierarchical federated learning architecture, not only can training latency be reduced, but network-side transmission overhead can also be reduced.

[0069] In conjunction with some embodiments of the second aspect, in some embodiments, receiving a first message sent by a first node includes one of the following: receiving a first control plane signaling sent by a first node, the first control plane signaling including a first message; the first node is deployed in a core network; receiving the first signaling sent by the first node through a first interface, the first signaling including a first message; the first node is deployed in an access network, the first interface being a communication interface between access network devices; receiving the first message sent by the first node through a service-oriented interface; the first node is deployed in an access network, and the access network supports service-oriented interfaces.

[0070] In conjunction with some embodiments of the second aspect, in some embodiments, the first control plane signaling is transmitted based on a first protocol layer, which is a control plane-based protocol layer and is associated with AI.

[0071] In conjunction with some embodiments of the second aspect, in some embodiments, receiving the first signaling sent by the first node through the first interface includes one of the following: receiving the first signaling sent by the first node through the Xn interface; the first node and the second node are deployed in different access network devices; receiving the first signaling sent by the first node through the F1 interface; the first node is deployed in the CU and the second node is deployed in the DU; receiving the first signaling sent by the first node through the F1-C interface; the first node is deployed in the CU-CP and the second node is deployed in the DU; receiving the first signaling sent by the first node through the E1 interface; the first node is deployed in the CU-CP and the second node is deployed in the CU-UP; receiving the first signaling sent by the first node through the user plane F1-U interface of F1; the first node is deployed in the CU-UP and the second node is deployed in the DU.

[0072] In conjunction with some embodiments of the second aspect, in some embodiments, the first message includes at least one of the following: a first indication information for indicating a first model; a second indication information for indicating a task identifier of a first task; the first task being a federated learning task based on the first model; a third indication information for indicating the training round of the first model in the first task; and a fourth indication information for indicating a subtask executed by at least one terminal associated with the second node within the training round, wherein the at least one terminal associated with the second node accesses the network through the second node.

[0073] In conjunction with some embodiments of the second aspect, in some embodiments, sending a first message to at least one terminal includes one of the following: sending a first message to at least one terminal; the first message further includes fifth indication information, the fifth indication information being used to indicate a sub-task performed by the terminal within the training round; the next-level device of the second node is a terminal; sending a first message to another second node; the first message being used by the other second node to send a first message to at least one terminal; the first message further includes fourth indication information, being used to indicate a sub-task performed by at least one terminal associated with the other second node within the training round, the at least one terminal associated with the other second node accessing the network through the other second node; the next-level device of the second node is another second node.

[0074] In the above embodiments, depending on the different deployment scenarios of the second node in the federated learning architecture, when the next-level device of the second node is a terminal, the second node can directly send a first message to at least one terminal. The first message sent by the second node to the terminal also includes fifth indication information, so that the fifth indication information can be used to distinguish the subtasks executed by different terminals during subsequent federated learning. When the next-level device of the second node is another second node, the second node can send a first message to the other second node to trigger the second node to send a first message to the terminal. The first message sent by the second node to the other second node also includes fourth indication information, so that the other second node can determine the terminal participating in federated learning within a training round based on the fourth indication information. In this way, the first message can be used to request both the other second node and the terminal to participate in federated learning.

[0075] In conjunction with some embodiments of the second aspect, in some embodiments, the first message is used by the second node to perform federated learning based on the first model, including at least one of the following: the first message is used by the second node to trigger at least one terminal to perform a training operation based on the first model; the first message is used by the second node to perform a first aggregation operation based on the first model parameters sent by at least one next-level device to obtain second model parameters; the second model parameters are used at least by the terminal to update the first model to be trained, and / or, the first node determines a third model; the third model is the learning result of federated learning based on the first model.

[0076] In conjunction with some embodiments of the second aspect, in some embodiments, the first model parameter is used to indicate the model parameters of a trained first model; the next-level device of the second node is a terminal; or, the first model parameter is used to indicate the aggregation result of the first aggregation operation; the aggregation result is determined by the model parameters of at least one trained first model; the next-level device of the second node is another second node.

[0077] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes: receiving second information sent by at least one next-level device, the second information including first model parameters; performing a first aggregation operation based on at least one first model parameter to obtain second model parameters.

[0078] In the above embodiments, the second node can receive first model parameters sent by at least one next-level device, and perform a first aggregation operation based on at least one first model parameter to obtain second model parameters. Since the first model parameters are sent by different next-level devices, the aggregated second model parameters utilize more diverse training data.

[0079] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes: sending first information to a first node, the first information including second model parameters; the first information is used by the first node to perform a second aggregation operation based on the second model parameters.

[0080] In the above embodiments, after obtaining the second model parameters, the second node can send the second model parameters to the first node by sending the first information, so that the first node can perform the second aggregation operation based on the second model parameters to obtain the third model, thereby making the third model obtained by the first node have higher generalization ability.

[0081] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes: sending third information to the terminal, the third information including second model parameters; the third information is used by the terminal to update the first model to be trained based on the second model parameters.

[0082] In the above embodiments, after obtaining the second model parameters, the second node can send third information to the terminal to transmit the second model parameters, enabling the terminal to update the first model to be trained based on the second model parameters. Since the second model parameters are obtained by the second node by aggregating the model parameters of the trained first model sent by at least one terminal, it enables at least one terminal to collaboratively complete the learning of the first model without sharing local data.

[0083] In conjunction with some embodiments of the second aspect, in some embodiments, the first message sent by the second node to at least one terminal is transmitted based on a third protocol layer; the second node is deployed in the DU; the third protocol layer is at least used by the DU to perform security protection and reordering of the first message.

[0084] The above embodiments introduce a third protocol layer, which enables the transmission of the first message between the DU and the terminal when the second node is deployed on the DU. This provides security protection and reordering for the first message sent from the DU to the terminal, thereby ensuring the security and privacy of information transmission between the DU and the terminal.

[0085] In conjunction with some embodiments of the second aspect, in some embodiments, the first message includes: time information, which is used by the terminal and / or the second node to determine the time for executing the first task; the first task is a federated learning task based on the first model.

[0086] In conjunction with some embodiments of the second aspect, in some embodiments, the time information includes at least one of the following: first time information, used to indicate the deadline for the terminal to complete the training operation within a training round of the first task; second time information, used to indicate the deadline for the second node to complete the first aggregation operation within a training round of the first task.

[0087] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes: sending fourth information, the fourth information being included in a system message; the fourth information including at least one of the following: sixth indication information, for indicating one or more first cells provided by the access network equipment deployed by the second node; the first cells being cells that support federated learning; seventh indication information, for indicating one or more first frequency points provided by the access network equipment deployed by the second node, the first frequency points being frequency points that support federated learning.

[0088] In the above embodiments, the second node may use system messages to send fourth information to at least one terminal, so as to indicate one or more cells that support federated learning provided by the second node, and / or one or more frequency points that support federated learning provided by the second node, so that all terminals entering the coverage area of ​​the cells provided by the second node can obtain the above information.

[0089] In conjunction with some embodiments of the second aspect, in some embodiments, the fourth information is used by the terminal to determine the priority corresponding to the first cell and / or the first frequency point; the priority corresponding to the first cell and / or the first frequency point is used to assist the terminal in cell reselection.

[0090] In the above embodiments, the second node uses system messages to send fourth information to at least one terminal, so that the terminal can determine the priority corresponding to the first cell and / or the first frequency point based on the fourth information. Thus, if the terminal has a need to participate in federated learning, the terminal can prioritize the first cell and / or the first frequency point during cell reselection.

[0091] Thirdly, embodiments of this disclosure provide a communication method, wherein the method is executed by a terminal, and the method includes: receiving a first message sent by a first node or a second node; the first message is used by the terminal to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network.

[0092] In the above embodiments, the terminal can receive a first message sent by a first node deployed in the core network or access network, or a second node deployed in the access network, to participate in the federated learning of the first model. Thus, on the one hand, by applying the federated learning method to the communication network, federated learning is achieved using the network resources of the communication network. On the other hand, by utilizing the first node, the second node, and the terminal in the communication network to implement a hierarchical federated learning architecture, not only can training latency be reduced, but network-side transmission overhead can also be reduced.

[0093] In conjunction with some embodiments of the third aspect, in some embodiments, receiving a first message sent by a first node includes one of the following: receiving a second control plane signaling sent by the first node, the second control plane signaling including the first message; the first node is deployed in a core network; receiving user plane signaling sent by the first node, the user plane signaling including the first message; the first node is deployed in a core network; receiving a second signaling sent by the first node through a second interface, the second signaling including the first message; the first node is deployed in an access network, and the second interface is a communication interface between access network equipment and a terminal.

[0094] In conjunction with some embodiments of the third aspect, in some embodiments, the second control plane signaling is transmitted based on the first protocol layer, the first protocol layer being a control plane protocol layer and related to AI; and / or, the user plane signaling is transmitted based on the second protocol layer, the second protocol layer being a user plane protocol layer and related to AI.

[0095] In conjunction with some embodiments of the third aspect, in some embodiments, the first message sent by the second node is transmitted based on the third protocol layer; the second node is deployed in the DU; the third protocol layer is at least used by the DU to perform security protection and reordering of the first message.

[0096] In conjunction with some embodiments of the third aspect, in some embodiments, the first message includes at least one of the following: a first indication information for indicating a first model; a second indication information for indicating a task identifier of a first task; the first task being a federated learning task based on the first model; a third indication information for indicating the training round of the first model in the first task; and a fifth indication information for indicating a subtask executed by the terminal within the training round.

[0097] In conjunction with some embodiments of the third aspect, in some embodiments, the method further includes: performing a training operation based on a first model to obtain first model parameters; the first model parameters being used to indicate the model parameters of the trained first model; sending second information to a second node, the second information including the first model parameters; the second information being used by the second node to perform a first aggregation operation based on the first model parameters.

[0098] In some embodiments of the third aspect, the method further includes: receiving third information sent by the second node; the third information includes second model parameters, which are obtained by the second node performing a first aggregation operation based on at least one first model parameter; the third information is used by the terminal to update the first model to be trained based on the second model parameters.

[0099] In conjunction with some embodiments of the third aspect, in some embodiments, the first message includes: time information, which is used by the terminal to determine the time for executing the first task; the first task is a federated learning task based on the first model.

[0100] In conjunction with some embodiments of the third aspect, in some embodiments, time information: first time information, used to indicate the deadline for the terminal to complete the training operation within a training round of the first task.

[0101] In conjunction with some embodiments of the third aspect, in some embodiments, the method further includes: receiving fourth information sent by the second node, the fourth information being contained in a system message, the fourth information including: sixth indication information for indicating one or more first cells provided by the access network equipment deployed by the second node; the first cells being cells supporting federated learning; and seventh indication information for indicating one or more first frequency points provided by the access network equipment deployed by the second node, the first frequency points being frequency points supporting federated learning.

[0102] In conjunction with some embodiments of the third aspect, in some embodiments, the fourth information is used by the terminal to determine the priority corresponding to the first cell and / or the first frequency point; the priority corresponding to the first cell and / or the first frequency point is used to assist the terminal in cell reselection.

[0103] Fourthly, embodiments of this disclosure provide a communication method, wherein the method is executed by a communication system, the method comprising: a first node sending a first message to at least one terminal and / or at least one second node; the first message being used by the terminal and / or the second node to perform federated learning based on a first model; the first node being deployed in a core network or an access network, and the second node being deployed in the access network.

[0104] Fifthly, embodiments of this disclosure provide a communication device, wherein the communication device performs the communication method provided in the first, second, or third aspect.

[0105] In a sixth aspect, embodiments of this disclosure provide a communication system, wherein the communication system includes a terminal, a first node, and a second node, the first node being configured to implement the communication method described in the optional implementation of the first aspect, the second node being configured to implement the communication method described in the optional implementation of the second aspect, and the terminal being configured to implement the communication method described in the optional implementation of the third aspect.

[0106] In a seventh aspect, embodiments of this disclosure provide a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform the communication method described in the optional implementations of the first, second, or third aspects.

[0107] Eighthly, embodiments of this disclosure provide a program product that, when executed by a communication device, causes the communication device to perform the communication method described in an optional implementation of the first, second, or third aspect.

[0108] In a ninth aspect, embodiments of this disclosure provide a computer program that, when run on a computer, causes the computer to perform the communication method described in an optional implementation of the first, second, or third aspect.

[0109] It is understood that the aforementioned communication devices, communication systems, storage media, program products, and computer programs are all used to execute the methods provided in the embodiments of this disclosure. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0110] This disclosure provides a communication method, a communication device, a communication system, a storage medium, and a program product. In some embodiments, the terms "communication method" and "information processing method," "information transmission method," etc., can be used interchangeably, as can the terms "communication system" and "information processing system."

[0111] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0112] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0113] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0114] In this embodiment of the disclosure, unless otherwise stated, elements expressed in the singular form, such as "a," "an," "the," "the," "the," "the," "the," "the," "this," etc., can mean "one and only one," or "one or more," "at least one," etc. For example, when using articles such as "a," "an," "the," etc. in translation, the noun following the article can be understood as either a singular expression or a plural expression.

[0115] In the embodiments disclosed herein, "multiple" refers to two or more.

[0116] In some embodiments, the terms “at least one of”, “one or more”, “a plurality of”, “multiple”, etc., may be used interchangeably.

[0117] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "A in one case, B in another", etc., may include the following technical solutions depending on the situation: in some embodiments, A (A is executed regardless of B); in some embodiments, B (B is executed regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). The same applies when there are more branches such as A, B, C, etc.

[0118] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execution of A regardless of B); in some embodiments, B (execution of B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, C, etc.

[0119] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not impose restrictions on the position, order, priority, quantity, or content of the descriptive objects. The description of the descriptive objects is found in the claims or the context of the embodiments, and the use of prefixes should not constitute unnecessary restrictions. For example, if the descriptive object is a "field," the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is a "level," the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the number of descriptive objects is not limited by ordinal numbers and can be one or more. For example, in "first device," the number of "devices" can be one or more. Furthermore, the objects modified by different prefixes can be the same or different. For example, if the object being described is "device", then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. As another example, if the object being described is "information", then "first message" and "first information" can be the same information or different information, and their content can be the same or different.

[0120] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0121] In some embodiments, terms such as “…”, “determine…”, “in the case of…”, “when…”, “when…”, “if…”, etc. can be used interchangeably.

[0122] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0123] In some embodiments, devices, etc., can be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. Terms such as “device”, “equipment”, “circuit”, “network element”, “node”, “function”, “unit”, “section”, “system”, “network”, “chip”, “chip system”, “entity”, and “subject” can be used interchangeably.

[0124] In some embodiments, "network" can be interpreted as devices included in a network (e.g., access network devices, core network devices, etc.).

[0125] In some embodiments, the terms "access network device (AN device)," "radio access network device (RAN device)," "base station (BS)," "radio base station," "fixed station," "node," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "antenna panel," "antenna array," "cell," "macro cell," "small cell," "femto cell," "pico cell," "sector," "cell group," "serving cell," "carrier," "component carrier," and "bandwidth part (BWP)" can be used interchangeably.

[0126] In some embodiments, the terms "terminal", "terminal device", "user equipment (UE)", "user terminal", "mobile station (MS)", "mobile terminal (MT)", "subscriber station", "mobile unit", "subscriber unit", "wireless unit", "remote unit", "mobile device", "wireless device", "wireless communication device", "remote device", "mobile subscriber station", "access terminal", "mobile terminal", "wireless terminal", "remote terminal", "handset", "user agent", "mobile client", and "client" can be used interchangeably.

[0127] In some embodiments, access network devices, core network devices, or network devices can be replaced by terminals. For example, embodiments of this disclosure can also be applied to structures where communication between access network devices, core network devices, or network devices and terminals is replaced by communication between multiple terminals (e.g., device-to-device (D2D), vehicle-to-everything (V2X), etc.). In this case, the structure can also be configured such that the terminal has all or part of the functions of the access network device. Furthermore, terms such as "uplink" and "downlink" can be replaced with terms corresponding to communication between terminals (e.g., "sidelink"). For example, uplink channel, downlink channel, etc., can be replaced with sidelink channel, and uplink link, downlink, etc., can be replaced with sidelink link.

[0128] In some embodiments, the terminal may be replaced by an access network device, a core network device, or a network device. In this case, the access network device, core network device, or network device may also be configured to have all or some of the functions of the terminal.

[0129] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0130] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0131] Furthermore, each element, each row, or each column in the table of this disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0132] Figure 1A is a schematic diagram of the architecture of a communication system according to an exemplary embodiment.

[0133] As shown in Figure 1A, the communication system 100 includes: a terminal 101 and a network device 102.

[0134] In some embodiments, terminal 101 includes, but is not limited to, at least one of the following: mobile phone, wearable device, Internet of Things device, car with communication function, smart car, tablet computer, computer with wireless transceiver function, virtual reality (VR) terminal device, augmented reality (AR) terminal device, wireless terminal device in industrial control, wireless terminal device in self-driving, wireless terminal device in remote medical surgery, wireless terminal device in smart grid, wireless terminal device in transportation safety, wireless terminal device in smart city, and wireless terminal device in smart home.

[0135] In some embodiments, network device 102 may include access network device and / or core network device.

[0136] In some embodiments, the access network device may be a node or device that connects a terminal to a wireless network. The access network device may include, but is not limited to, at least one of the following in a 5G communication system: evolved Node B (eNB), next-generation eNB (ng-eNB), next-generation Node B (gNB), node B (NB), home node B (HNB), home evolved node B (HeNB), radio backhaul device, radio network controller (RNC), base station controller (BSC), base transceiver station (BTS), base band unit (BBU), mobile switching center, base station in a 6G communication system, open RAN, cloud RAN, base station in other communication systems, and access node in a Wi-Fi system.

[0137] In some embodiments, the technical solutions of this disclosure can be applied to the Open RAN architecture. In this case, the interfaces between or within access network devices involved in the embodiments of this disclosure can be transformed into internal interfaces of Open RAN. The processes and information interactions between these internal interfaces can be implemented by software or programs.

[0138] In some embodiments, the access network device may be composed of a central unit (CU) and a distributed unit (DU). The CU may also be called a control unit. The CU-DU structure can separate the protocol layer of the access network device. Some of the protocol layer functions are centrally controlled by the CU, while the remaining part or all of the protocol layer functions are distributed in the DU and centrally controlled by the CU. However, this is not the only possibility.

[0139] In some embodiments, the Radio Resource Control (RRC) layer, Service Data Adaptation Protocol (SDAP) layer, and Packet Data Convergence Protocol (PDCP) layer are deployed in the CU. The Radio Link Control (RLC) layer, Media Access Control (MAC) layer, and Physical Layer (PHY) layer are deployed in the DU.

[0140] In some embodiments, an access network device may consist of one CU and one or more DUs. The CU and DU can be connected via an F1 interface. In one embodiment, as shown in FIG1B, FIG1B is a schematic diagram of a discrete architecture of an access network device according to an exemplary embodiment. An access network device gNB may consist of one gNB-CU and multiple gNB-DUs. The gNB-CUs can be connected via an Xn-C interface, the gNB-CU and gNB-DU can be connected via an F1 interface, and the gNB can be connected to the 5G Core Network (5GC) via an NG interface.

[0141] In some embodiments, a CU can be connected to one or more DUs simultaneously.

[0142] In some embodiments, a DU supports one or more cells.

[0143] In some embodiments, the CU may include a control plane (CP) and a user plane (UP).

[0144] In some embodiments, the CP (CU-CP) and UP (CU-UP) of the CU can be on different physical devices. Alternatively, the CU-CP and CU-UP can be on the same physical device.

[0145] In some embodiments, a CU may include a CU-CP and one or more CU-UPs. The CU-CP and CU-UP are connected via an E1 interface; the CU-CP and DU are connected via an F1-C interface; and the CU-UP and DU are connected via an F1-U interface. In one embodiment, as shown in FIG1C, FIG1C is a schematic diagram of a split architecture of an access network device according to an exemplary embodiment. In this embodiment, the gNB-CU-CP and gNB-CU-UP are connected via an E1 interface, the gNB-CU-CP and gNB-DU are connected via an F1-C interface, and the gNB-CU-UP and gNB-DU are connected via an F1-U interface.

[0146] In some embodiments, a core network device can be a single device, including one or more network elements, or it can be multiple devices or a group of devices, each including one or more network elements. Network elements can be virtual or physical. The core network may include, for example, at least one of the Evolved Packet Core (EPC), 5GC, and Next Generation Core (NGC).

[0147] It is understood that the communication system described in this disclosure is for the purpose of more clearly illustrating the technical solutions of this disclosure, and does not constitute a limitation on the technical solutions provided in this disclosure. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided in this disclosure are also applicable to similar technical problems.

[0148] The following embodiments of this disclosure can be applied to the communication system 100 shown in FIG1A, or to some of the main bodies, but are not limited thereto. The main bodies shown in FIG1A are illustrative. The communication system may include all or some of the main bodies in FIG1A, or it may include other main bodies outside of FIG1A. The number and form of each main body are arbitrary. The connection relationship between the main bodies is illustrative. The main bodies may not be connected or may be connected. The connection can be in any way, it can be a direct connection or an indirect connection, it can be a wired connection or a wireless connection.

[0149] The embodiments disclosed herein can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G new radio (NR), 6th generation mobile communication system (6G), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New radio access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20, Ultra-Wideband (UWB), Bluetooth (a registered trademark), Public Land Mobile Network (PLMN) networks, Device-to-Device (D2D) systems, Machine-to-Machine (M2M) systems, Internet of Things (IoT) systems, Vehicle-to-Everything (V2X) systems, systems utilizing other communication methods, and next-generation systems built upon them, etc. Furthermore, multiple systems can be combined (e.g., a combination of LTE or LTE-A with 5G).

[0150] The following is an explanation and interpretation of the terminology used in this disclosure.

[0151] In some embodiments, it is assumed that the federated learning algorithm is the Federated Averaging Algorithm (FedAvg). There are K federated learning members. Federated learning typically runs for multiple rounds (total number of rounds H (H≥1)). Taking round t as an example, the following steps are performed:

[0152] 1. Member Selection: Select K members from the K federated learning members to begin training on the local data, while the other members wait for the next federated round. Where K... t This refers to the members who train the model in round t.

[0153] 2. Configuration: The central server notifies the selected members to train the model on local data according to a preset mode (e.g., stochastic gradient descent (SGD) or mini-batch gradient descent). The configuration information can be notified separately for each round, or it can be configured only once in a training task of federated learning.

[0154] 3. Training: The selected member k obtains the current global model θ. t It then selects local data for training based on a preset mode. For example, when using SGD, member k selects a local training dataset for training, obtaining the updated local model θ. t,k .

[0155] 4. Reporting: Each selected member k sends its local model θ to the central server. t,k The central server aggregates the received local models to obtain a new global model. This global model is the same global model used in the next federated learning round.

[0156] In some embodiments, the related technologies propose a hierarchical federated learning (HLL) method. In this method, mobile users (MUs) are grouped according to their location. In each round of federated learning, each MU in a group sends its updated local model to the corresponding small cell base station (SBS) for aggregation. The SBS then averages the local models obtained from multiple MUs and sends them back to the corresponding MU for training. After every H (H≥1) training rounds, each SBS can send the aggregated model to a macro base station (MBS) for aggregation to obtain a global model, which can then be used by each MU for further training. Compared to traditional federated learning (where all aggregation is performed on a central server), hierarchical federated learning offers advantages such as lower latency and reduced network-side transmission overhead.

[0157] There is currently no framework for applying hierarchical federated learning to mobile communication networks.

[0158] This disclosure provides a communication method, communication device, communication system, storage medium, and program product to apply hierarchical federated learning in mobile communication networks, thereby reducing training latency and network-side transmission overhead while implementing federated learning.

[0159] In some embodiments, the hierarchical federated learning architecture may include two logical nodes, namely a first node and a second node; wherein, the first node may be used to select federated learning members, distribute federated learning tasks and global models to be trained to the second node and the terminal, and aggregate the local models and branch models obtained from the second node and the terminal into a global model.

[0160] Figure 1D is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment, and Figure 1E is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment. The deployment of the second node within the hierarchical federated learning architecture can be divided into single-layer deployment (as shown in Figure 1D) and multi-layer deployment (as shown in Figure 1E). In Figure 1D, the second node can be used to send the global model to be trained and / or the aggregated branch model to the terminal, aggregate the local model obtained from the terminal into a branch model, and send the branch model to the first node. In Figure 1E, the second node is used to send the global model and the aggregated branch model to the terminal and / or the next-level second node, aggregate the local model obtained from the terminal and / or the next-level second node into a branch model, and send the branch model to the previous-level second node or the first node.

[0161] In some embodiments, the first node and the second node may be deployed in a network device.

[0162] In some embodiments, the first node may be deployed in an access network device or a core network device.

[0163] In some embodiments, the second node can be deployed in an access network device. In one example, the second node can be deployed in a gNB.

[0164] In some embodiments, the first node and the second node may be deployed in multiple devices, or the first node and the second node may be deployed in one device.

[0165] In one embodiment, the first node can be deployed in a core network device, and the second node can be deployed in an access network device. In another embodiment, the first node and the second node can be deployed in different access network devices.

[0166] In one example, as shown in Figure 1F, which is a schematic diagram three illustrating a hierarchical federated learning architecture according to an exemplary embodiment, gNB1 and gNB2 can be second nodes, and the FL server can be a first node. Here, the first node can be a separate Network Function (NF), or the first node can be deployed in a gNB. In another example, as shown in Figure 1G, which is a schematic diagram four illustrating a hierarchical federated learning architecture according to an exemplary embodiment, the macro base station and small base station in Figure 1G can be second nodes, and the FL server can be a first node. Here, the first node can be a separate NF, or the first node can be deployed in a macro base station.

[0167] In some embodiments, where the access network device supports a CU-DU separation architecture, the first node can be deployed in the core network device or CU, and the second node can be deployed in the DU and / or CU.

[0168] In one embodiment, as shown in FIG1H, FIG1H is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment. In FIG1H, gNB-DU and gNB-CU can be second nodes, and the FL server can be a first node. Here, the first node can be a separate NF, or the first node can be deployed in gNB-CU.

[0169] In some embodiments, where the access network device supports a CU-DU separation architecture and the access network device supports a CP-UP separation architecture, the first node can be deployed in the core network device, CU-UP, or CU-CP, and the second node can be deployed in the DU, CU-UP, and / or CU-CP.

[0170] In one embodiment, as shown in FIG1I, FIG1I is a schematic diagram of a hierarchical federated learning architecture according to an exemplary embodiment. In FIG1I, gNB-DU and gNB-CU-UP can be second nodes, and the FL server can be a first node. Here, the first node can be a separate NF, or the first node can be deployed in gNB-CU-UP or gNB-CU-CP.

[0171] In some embodiments, the first node and the second node can be deployed in the same access network device. In one example, the first node and the second node can be deployed in a gNB, wherein the first node is deployed in gNB-CU-CP, and the second node is deployed in gNB-DU and gNB-CU-UP.

[0172] In some embodiments, when gNB-DU is the second node, gNB-DU needs to be able to send the aggregated branch model to the terminal, and aggregate the local model obtained from the terminal into a branch model. Therefore, gNB-DU needs a complete protocol stack for communicating with the terminal.

[0173] In one embodiment, in a 5G / NR system, the gNB-DU carries the RLC, MAC, and PHY layers, but not the PDCP layer. The PDCP layer has security functions (encryption, decryption, integrity protection), reordering, etc. Considering the large amount of data transmitted between the gNB-DU and the terminal, and the need for security functions to protect data privacy, when the gNB-DU acts as a second node, it needs to support the PDCP layer or a protocol layer with similar functions. This protocol layer can be a protocol layer designed for federated learning, or the gNB-DU can support the SBI protocol stack, or the gNB-DU can support a first protocol stack, which includes Hypertext Transfer Protocol (HTTP) / 3; Quick User Datagram Protocol (UDP) Internet Connections (QUIC); User Datagram Protocol (UDP); and Internet Protocol (IP) layers.

[0174] Figure 2A is an interactive schematic diagram of a communication method according to an exemplary embodiment. As shown in Figure 2A, this disclosure relates to a communication method for a communication system 100, the method comprising:

[0175] In this embodiment of the disclosure, the federated learning process is illustrated by taking the deployment of the first node on the core network device and the deployment of the second node on the access network device as an example.

[0176] In some embodiments, when the access network equipment is an integrated architecture (i.e., the CU and DU are not separated), the second node can be deployed in access network equipment such as base stations, small base stations, and / or macro base stations, as shown in Figures 1F and 1I.

[0177] In some embodiments, when the access network device has a CU-DU separation architecture, the second node can be deployed in the CU of the access network device and / or in the DU of the access network device, as shown in Figure 1H.

[0178] In some embodiments, when the access network device has a CU-DU separation and CP-UP separation architecture, the second node can be deployed in the CU-CP of the access network device, deployed in the CU-UP of the access network device, and / or deployed in the DU of the access network device, as shown in Figure 1I.

[0179] Step S2101: The core network device sends a first message to at least one access network device and / or at least one terminal.

[0180] In some embodiments, the core network device sends a first message to at least one access network device. In one embodiment, the core network device sends the first message to at least one access network device, which then sends the first message to the terminal.

[0181] In some embodiments, at least one access network device receives a first message sent by a core network device.

[0182] In some embodiments, the first message may be sent by the core network device to the access network device, or the first message may be forwarded by other network elements to the access network device. The core network device may send the first message to the Access and Mobility Management Function (AMF), and the AMF may forward the first message to the access network device.

[0183] In some embodiments, a core network device may send a first control plane signaling message to at least one access network device, the first control plane signaling message including a first message.

[0184] In some embodiments, the core network device may send a first control plane signaling to at least one access network device via the AMF.

[0185] In some embodiments, the first control plane signaling may be control plane signaling between core network equipment and access network equipment.

[0186] In some embodiments, the first control plane signaling is transmitted based on a first protocol layer, which is a control plane-based protocol layer and is related to AI.

[0187] In some embodiments, the name of the first protocol layer is not limited, for example, AI layer, AI transport layer, etc.

[0188] In some embodiments, the first protocol layer of the core network device is a protocol above the fourth protocol layer, which is related to HTTP. In one embodiment, the fourth protocol layer may be the HTTP / 2 layer.

[0189] In some embodiments, the first protocol layer of the access network device is an upper protocol of the NG AP layer.

[0190] In one embodiment, as shown in FIG1J, FIG1J is a schematic diagram of a protocol stack structure in a communication system according to an exemplary embodiment. When the core network device sends the first control plane signaling to the access network device through the AMF, the protocol layers of the core network device from top to bottom are: AI layer, HTTP / 2 layer, Transport Layer Security (TLS) layer, Transmission Control Protocol (TCP) layer, IP layer, data link layer (i.e., L2 layer), and physical layer (i.e., L1 layer). The protocol layers of the access network device from top to bottom are: AI layer, NG Application Protocol (NG AP) layer, Stream Control Transmission Protocol (SCTP) layer, IP layer, L2 layer, and L1 layer.

[0191] In some embodiments, where the access network device supports a service-based interface, the core network device can send a first message to at least one access network device through the service-based interface.

[0192] In some embodiments, the first node may be a core network function, and the second node may be an access network function.

[0193] In some embodiments, the service interface can be used for communication between core network functions and access network functions.

[0194] In one embodiment, as shown in FIG1K, FIG1K is a schematic diagram of a protocol stack structure of a service-oriented interface according to an exemplary embodiment. The protocol layers of the service-oriented interface of the core network device and the access network device, from top to bottom, are the Application layer, HTTP / 2 layer, TLS layer, TCP layer, IP layer, and L2 layer.

[0195] In some embodiments, the core network device sends a first message to at least one terminal. In one embodiment, the core network device sends the first message directly to at least one terminal, which can reduce the transmission overhead on the network side.

[0196] In some embodiments, at least one terminal receives a first message sent by a core network device.

[0197] In some embodiments, the first message may be sent by a core network device, or it may be forwarded by other network elements. In one example, the first message may be forwarded by the AMF and the access network device. In another example, the first message may be forwarded by the User Plane Function (UPF) and the access network device.

[0198] The core network equipment can send a second control plane signaling message to at least one terminal, the second control plane signaling message including the first message.

[0199] In some embodiments, the second control plane signaling may be control plane signaling between the core network equipment and the terminal.

[0200] In some embodiments, the second control plane signaling is transmitted based on the first protocol layer; the first protocol layer is a control plane-based protocol layer and is related to AI.

[0201] In some embodiments, the name of the first protocol layer is not limited, for example, AI layer, AI transport layer, etc.

[0202] In some embodiments, the first protocol layer of the core network device is a protocol above the fourth protocol layer, which is related to HTTP. In one embodiment, the fourth protocol layer may be the HTTP / 2 layer.

[0203] In some embodiments, the first protocol layer of the terminal is an upper protocol of the Non-Access Stratum (NAS) layer.

[0204] In one embodiment, as shown in FIG1L, FIG1L is a schematic diagram of a protocol stack structure in a communication system according to an exemplary embodiment. When the core network device sends second control plane signaling to the terminal through AMF and RAN, the protocol layers of the core network device from top to bottom are: AI layer, HTTP / 2 layer, TLS layer, TCP layer, IP layer, L2 layer, and L1 layer. The protocol layers of the terminal from top to bottom are: AI layer, NAS layer, RRC layer, PDCP layer, RLC layer, MAC layer, and L1 layer.

[0205] The core network equipment can send user plane signaling to at least one terminal, and the user plane signaling includes a first message.

[0206] In some embodiments, user plane signaling may be user plane signaling between core network equipment and a terminal.

[0207] In some embodiments, user plane signaling is transmitted based on a second protocol layer; the second protocol layer is a user plane-based protocol layer and is related to AI.

[0208] In some embodiments, the name of the second protocol layer is not limited, for example, AI layer, AI transport layer, etc.

[0209] In some embodiments, the second protocol layer of the core network device is a protocol above the TCP protocol.

[0210] In some embodiments, the second protocol layer of the terminal is a protocol above the TCP protocol.

[0211] In one embodiment, as shown in FIG1M, FIG1M is a schematic diagram of a protocol stack structure in a communication system according to an exemplary embodiment. When the core network device sends user plane signaling to the terminal through UPF and RAN, the protocol layers of the core network device from top to bottom are: AI layer, TCP layer, IP layer, L2 layer, and L1 layer. The protocol layers of the terminal from top to bottom are: AI layer, TCP layer, IP layer, PDCP layer, RLC layer, MAC layer, and L1 layer.

[0212] In some embodiments, the core network device sends a first message to at least one access network device and at least one terminal.

[0213] In some embodiments, at least one access network device and at least one terminal receive a first message sent by a core network device.

[0214] In some embodiments, before the core network device sends a first message to at least one access network device, the core network device determines at least one access network device and at least one terminal to participate in federated learning. In one embodiment, the core network device may first select federated learning members and then send the first message to the selected federated learning members.

[0215] In some embodiments, before the core network device sends a first message to at least one access network device, the core network device may receive a fifth message sent by at least one access network device, the fifth message being used to indicate at least one candidate access network device and / or at least one candidate terminal.

[0216] In some embodiments, the fifth information may be used by the core network device to determine at least one access network device and / or at least one terminal participating in federated learning.

[0217] In some embodiments, the access network devices participating in federated learning may be one or more of the candidate access network devices indicated by the fifth information. In some embodiments, the terminals participating in federated learning may be one or more of the candidate terminals indicated by the fifth information.

[0218] In some embodiments, upon receiving the sixth message, the core network device may send a first message to at least one access network device and / or at least one terminal. The sixth message is used to request federated learning services.

[0219] In one embodiment, the core network device can act as a provider of federated learning services. The core network device can receive sixth information sent by consumers of the federated learning services. In one embodiment, the consumer of the federated learning services can be an application function (AF) or an application service (AS), etc.

[0220] In some embodiments, the core network device can determine the first message based on the sixth information. In one embodiment, the core network device can determine the model to be trained in the federated learning task and the total number of training rounds of the model to be trained based on the sixth information.

[0221] In some embodiments, the first message is used by the terminal and / or access network device to perform federated learning based on the first model.

[0222] In some embodiments, the first message sent by the core network device to the terminal can be used by at least one terminal to perform a training operation based on a first model. In some embodiments, the first message sent by the core network device to the access network device can be used by the access network device to perform a first aggregation operation based on first model parameters sent by at least one next-level device to obtain second model parameters.

[0223] In some embodiments, the first message sent by the core network device to the access network device can also be used by the access network device to send a first message to at least one terminal; the first message sent by the access network device to at least one terminal can be used by at least one terminal to perform a training operation based on the first model.

[0224] In some embodiments, the first message may include at least one of the following:

[0225] The first instruction information is used to indicate the first model;

[0226] The second instruction information is used to indicate the task identifier of the first task; the first task is a federated learning task based on the first model.

[0227] The third instruction information is used to indicate the training round of the first model in the first task.

[0228] In some embodiments, the first indication information may be used by access network devices and / or terminals to determine the first model to be trained.

[0229] In some embodiments, the first indication information may be used to indicate one of the following: the meta information of the first model; the model structure of the first model; the model parameters of the first model; and the training configuration of the first model.

[0230] In some embodiments, the meta-information of the first model is information describing the model attributes of the first model.

[0231] In some embodiments, the metadata of the first model may include, but is not limited to, at least one of the following: the model version of the first model; the model identifier of the first model; the model size of the first model; and the supplier identifier of the first model.

[0232] In some embodiments, the model structure of the first model is used to indicate the network structure of the first model.

[0233] In some embodiments, the model parameters of the first model may include trainable parameters and non-trainable parameters. The values ​​of trainable parameters are learned during model training, such as weights. Non-trainable parameters are fixed and do not update during model training, such as the loss function.

[0234] In some embodiments, the training configuration of the first model is used by the access network device and / or terminal to determine the learning parameters for federated learning of the first model.

[0235] In some embodiments, the training configuration of the first model may include, but is not limited to, at least one of the following: training mode; aggregation algorithm.

[0236] In some embodiments, the task identifier of the first task can be any information used to identify the first task.

[0237] In some embodiments, task identifiers can be used by access network devices and / or terminals to distinguish different federated learning tasks.

[0238] In some embodiments, the third indication information may include: round information, used to indicate the total number of training rounds of the first model in the first task.

[0239] In one embodiment, when the total training rounds of the first model in the first task are 1, after the access network device receives the first model parameters sent by at least one next-level device, it can perform a first aggregation operation based on the at least one first model parameters to obtain the second model parameters, and send the second model parameters to the core network device so that the core network device can determine the model parameters of the third model based on the second model parameters, thereby obtaining the learning result of the first task.

[0240] In one embodiment, when the total number of training rounds for the first model in the first task is H, after the access network device receives the first model parameters sent by at least one next-level device, it can perform a first aggregation operation based on the at least one first model parameter to obtain second model parameters. The second model parameters are then sent to at least one next-level device (e.g., a terminal), allowing the terminal to update the first model to be trained based on the first model parameters and train the updated first model to resend the first model parameters to the access network device. This process is repeated until the terminal has trained the first model for H rounds. At this point, the access network device sends the second model parameters to the core network device so that the core network device can determine the model parameters of the third model based on the second model parameters, thereby obtaining the learning result of the first task. Here, H is a positive integer greater than 1.

[0241] In some embodiments, the third indication information may further include: a round identifier, used to indicate the number of training rounds of the first model in the first task.

[0242] In some embodiments, when a core network device sends a first message to at least one access network device, the first message further includes fourth indication information, which is used to indicate a subtask performed by at least one terminal associated with the access network device within a training round.

[0243] In some embodiments, the fourth indication information can be used by the access network device to determine which terminal among at least one terminal associated with it is participating in federated learning. The at least one terminal associated with the access network device accesses the network through the access network device.

[0244] In some embodiments, the fourth indication information may include: at least one first identifier; the first identifier is used to determine the terminal performing the subtask.

[0245] In one embodiment, the first identifier may be a terminal identifier. In another embodiment, the first identifier may be a tuple including a terminal identifier and a subtask identifier.

[0246] In some embodiments, when the core network device sends a first message to at least one terminal, the first message further includes fifth indication information for indicating the sub-tasks performed by the terminal within a training round.

[0247] In some embodiments, the fifth indication information may be a first identifier; the first identifier is used to determine the terminal performing the subtask.

[0248] In one embodiment, the first identifier may be a terminal identifier. In another embodiment, the first identifier may be a tuple including a terminal identifier and a subtask identifier.

[0249] In some embodiments, a first message sent by a core network device to at least one access network device may include at least one of the following: a first indication message; a second indication message; a third indication message; or a fourth indication message.

[0250] In some embodiments, the first message sent by the core network device to at least one terminal may include at least one of the following: a first indication message; a second indication message; a third indication message; or a fifth indication message.

[0251] In some embodiments, the first message may include time information, which is used to instruct the terminal and / or the second node to determine the time for performing the first task.

[0252] In some embodiments, the time information may include at least one of the following: first time information, used to indicate the deadline for the terminal to complete the training operation within a training round of the first task; and second time information, used to indicate the deadline for the second node to complete the first aggregation operation within a training round of the first task.

[0253] Step S2102: The access network device sends a first message to at least one terminal.

[0254] In some embodiments, at least one terminal receives a first message sent by an access network device.

[0255] In some embodiments, when the next-level device of the access network device is a terminal, the access network device sends a first message to at least one terminal. In one embodiment, when the access network device is deployed as a single layer within a hierarchical federated learning architecture, the access network device sends a first message to at least one terminal after receiving the first message sent by the core network device.

[0256] In some embodiments, at least one terminal is a terminal participating in federated learning among at least one terminal associated with an access network device. In one embodiment, when a core network device sends a first message to at least one access network device, each access network device that receives the first message sends the first message to at least one terminal associated with it that needs to participate in federated learning.

[0257] In some embodiments, the access network device can send a third signaling message to at least one terminal through a second interface, the third signaling message including a first message. The second interface is the interface between the access network device and the terminal. In one embodiment, the second interface can be a Uu interface. In another embodiment, the Uu interface is used to transmit user data and control information.

[0258] In some embodiments, when the second node is deployed on the DU of the access network device, the first message sent by the DU is transmitted based on the third protocol layer; the third protocol layer is at least used by the DU to perform security protection and reordering of the first message.

[0259] In some embodiments, the third protocol layer may be the PDCP layer; or the third protocol layer may be any protocol layer capable of performing security protection functions and reordering functions.

[0260] In some embodiments, the third protocol layer of the DU is an upper protocol layer of the RLC layer. In one embodiment, when the DU sends a first message to the terminal, the protocol layers of the DU from top to bottom are: the third protocol layer; the RLC layer; the MAC layer and the PHY layer.

[0261] In some embodiments, a first message sent by the access network device to at least one terminal is used to request at least one terminal to perform federated learning based on a first model.

[0262] In some embodiments, the first message sent by the access network device to at least one terminal may include at least one of the following: a first indication message; a second indication message; a third indication message; or a fifth indication message.

[0263] In some embodiments, the first message sent by the access network device to at least one terminal may include time information, which is used to instruct the terminal to determine the time to perform the first task.

[0264] In some embodiments, the time information may include at least one of the following: first time information, used to indicate the deadline for the terminal to complete the training operation within one training round of the first task.

[0265] In some embodiments, when the next-level device of the access network device is another access network device, the access network device sends a first message to the terminal through the other access network device. In one embodiment, when the access network device is deployed in a multi-layered federated learning architecture (e.g., two layers, i.e., the access network device can be divided into a first-level access network device and a second-level access network device, with the second-level access network device being the next-level device of the first-level access network device), after receiving the first message sent by the core network device, the first-level access network device sends a first message to the second-level access network device, so that the second-level access network device sends the first message to at least one terminal.

[0266] In some embodiments, where the access network device supports a service-oriented interface, the first-level access network device can send a first message to the second-level access network device through the service-oriented interface.

[0267] In some embodiments, a first-level access network device can send a first message to a second-level access network device through a first interface. The first interface can be a communication interface between access network devices.

[0268] In one embodiment, when the first-level access network device and the second-level access network device are deployed in different access network devices, the first-level access network device can send a first message to the second-level access network device through the Xn interface.

[0269] In some embodiments, a first message sent by a first-level access network device to a second-level access network device may include at least one of the following: a first indication information for indicating a first model; a second indication information for indicating a task identifier of a first task; the first task being a federated learning task based on the first model; a third indication information for indicating the training round of the first model in the first task; and a fourth indication information for indicating a sub-task performed by at least one terminal associated with the access network device within a training round.

[0270] In some embodiments, the first message may include time information, which is used to instruct the terminal and / or access network device to determine the time for performing the first task.

[0271] In some embodiments, the time information may include at least one of the following: first time information, used to indicate the deadline for the terminal to complete the training operation within a training round of the first task; and second time information, used to indicate the deadline for the access network device to complete the first aggregation operation within a training round of the first task.

[0272] Step S2103: The terminal performs training operations based on the first model.

[0273] In some embodiments, when the terminal receives the first message, the terminal can perform a training operation based on the first model to obtain the model parameters of the trained first model.

[0274] In some embodiments, the first message can be used by the terminal to perform training operations.

[0275] In some embodiments, the model parameters of the trained first model can be used by the access network device to determine the parameters of the second model.

[0276] In some embodiments, the second model parameters are determined based on the model parameters of at least one trained first model.

[0277] In some embodiments, when the terminal receives third information, the terminal can perform training operations based on the first model.

[0278] In some embodiments, upon receiving third information, the terminal updates the first model to be trained based on the second model parameters and performs training operations based on the first model. The third information may include the second model parameters.

[0279] In some embodiments, the third information can be used by the terminal to update the first model to be trained based on the second model parameters. In one embodiment, after receiving the third information sent by the access network device, the terminal can update the first model to be trained based on the second model parameters, and perform a training operation based on the updated first model to obtain the model parameters of the trained first model.

[0280] Step S2104: The terminal sends the second information to the access network device.

[0281] In some embodiments, the access network device receives second information sent by at least one terminal.

[0282] In some embodiments, when the next-level device of the access network device is a terminal, the access network device receives second information sent by at least one terminal.

[0283] In some embodiments, after the terminal completes the training operation of the first model, the terminal sends the second information to the access network device.

[0284] In some embodiments, the terminal can send second information to the access network device through a second interface. The second interface is the interface between the access network device and the terminal. In one embodiment, the second interface may be a Uu interface.

[0285] In some embodiments, the second information may be used by the access network device to perform a first aggregation operation.

[0286] In some embodiments, the second information may include the first model parameters.

[0287] In some embodiments, when the next-level device of the access network device is a terminal, the first model parameter is used to indicate the model parameters of the trained first model. In one embodiment, after the terminal completes the training operation of the first model, it can send the model parameters of the trained first model to the access network device, and the access network device can perform a first aggregation operation on the model parameters of the trained first model sent by at least one terminal.

[0288] In some embodiments, the second information may further include at least one of the following: second indication information for indicating the task identifier of the first task; the first task is a federated learning task based on the first model; fifth indication information for indicating the sub-task executed by the terminal within a training round; and round identifier for indicating the training round of the first model in the first task.

[0289] In some embodiments, the round identifier can be used by the access network device to determine whether to send the second model parameters to the terminal or the core network device. In one embodiment, after obtaining the second model parameters, the access network device can determine whether to continue training the first model based on the number of training rounds of the first model and the total number of training rounds of the first model. If the number of training rounds of the first model is less than the total number of training rounds, it is determined that training the first model needs to continue, and the access network device sends the second model parameters to the terminal. If the number of training rounds of the first model is equal to the total number of training rounds, it is determined that training the first model does not need to continue, and the access network device sends the second model parameters to the core network device.

[0290] Step S2105: The access network device performs a first aggregation operation based on at least one first model parameter.

[0291] In some embodiments, after the access network device receives second information sent by at least one next-level device, the access network device performs a first aggregation operation based on at least one first model parameter to obtain the second model parameter.

[0292] In some embodiments, the second information may include the first model parameters.

[0293] In some embodiments, when the next-level device of the access network device is a terminal, the first model parameter is used to indicate the model parameters of the trained first model.

[0294] In some embodiments, when the next-level device of the access network device is another access network device, the first model parameters are used to indicate the aggregation result of the first aggregation operation. The aggregation result is determined by the model parameters of at least one trained first model.

[0295] In one embodiment, when the access network devices are deployed in a multi-layered federated learning architecture (e.g., two layers, meaning the access network devices can be divided into first-level and second-level access network devices, with the second-level access network device being the next level down from the first-level access network device), after the terminal completes the training operation on the first model, it can send the model parameters of the trained first model to the second-level access network device. The second-level access network device can perform a first aggregation operation on the model parameters of the trained first model sent by at least one terminal to obtain an aggregation result. The second-level access network device can then send the aggregation result of the first aggregation operation to the first-level access network device, and the first-level access network device can perform a first aggregation operation on the aggregation result sent by at least one second-level access network device.

[0296] In some embodiments, the second model parameters can be used by the terminal to update the first model to be trained.

[0297] In some embodiments, if the number of training epochs of the first model is less than the total number of training epochs, the parameters of the second model can be used by the terminal to update the first model to be trained.

[0298] In some embodiments, if the number of training rounds of the first model is less than the total number of training rounds, the access network device may send third information to the terminal, the third information including the second model parameters.

[0299] In some embodiments, the third information can be used by the terminal to update the first model to be trained based on the second model parameters.

[0300] In some embodiments, the third information can also be used by the terminal to perform training operations based on the first model.

[0301] In one embodiment, when the access network device obtains the second model parameters, if the number of training rounds of the first model is less than the total number of training rounds, the access network device can send third information containing the second model parameters to the terminal to trigger the terminal to update the first model to be trained based on the second model parameters, and perform training operations based on the updated first model.

[0302] In some embodiments, the second model parameters can be used by the core network equipment to determine the third model.

[0303] In some embodiments, when the number of training epochs of the first model is equal to the total number of training epochs, the parameters of the second model can be used by the core network device to determine the third model.

[0304] Step S2106: The access network device sends the first information to the core network device.

[0305] In some embodiments, the core network device receives first information sent by at least one access network device.

[0306] In some embodiments, when the number of training rounds of the first model is equal to the total number of training rounds, the access network device sends the first information to the core network device.

[0307] In one embodiment, the access network device may send a first control plane signaling to the core network device, the first control plane signaling including first information.

[0308] In one embodiment, when the access network device supports a service-oriented interface, the access network device can send first information to the core network device through the service-oriented interface.

[0309] In some embodiments, the first information may be used by the core network device to perform a second aggregation operation.

[0310] In some embodiments, the first information may include second model parameters, which are used by the core network device to determine the third model.

[0311] In some embodiments, the third model is the learning result of federated learning based on the first model.

[0312] In some embodiments, the first information may further include: second indication information for indicating a task identifier of the first task; the first task is a federated learning task based on the first model; fourth indication information for indicating a sub-task performed by at least one terminal associated with the access network device within a training round; and round identifier for indicating the number of training rounds of the first model in the first task.

[0313] Step S2107: The core network device performs a second aggregation operation based on at least one second model parameter to obtain the model parameters of the third model.

[0314] In some embodiments, after the core network device receives first information sent by at least one access network device, the core network device performs a second aggregation operation based on at least one second model parameter to obtain the model parameters of the third model; wherein, the first information may include the second model parameters.

[0315] In some embodiments, the method may further include: the access network device may send fourth information to at least one terminal, the fourth information being included in a system message, the fourth information including at least one of the following: sixth indication information, for indicating one or more first cells provided by the access network device deployed by the access network device; the first cells being cells that support federated learning; seventh indication information, for indicating one or more first frequency points provided by the access network device deployed by the access network device, the first frequency points being frequency points that support federated learning.

[0316] In some embodiments, the fourth information is used by the terminal to determine the priority corresponding to the first cell and / or the first frequency point; the priority corresponding to the first cell and / or the first frequency point is used to assist the terminal in cell reselection.

[0317] In one embodiment, after receiving the fourth information sent by the access network device, the terminal can determine the priority corresponding to the first cell and / or the first frequency point based on the fourth information. If the terminal supports federated learning and has a need to participate in federated learning, the terminal can increase the priority corresponding to the first cell and / or the first frequency point. If the terminal does not support federated learning or has no need to participate in federated learning, the terminal can decrease the priority corresponding to the first cell and / or the first frequency point, so that subsequent cell reselection can be performed based on the priorities of each candidate cell and / or each candidate frequency point.

[0318] The communication method involved in the embodiments of this disclosure may include at least one of steps S2101 to S2107. For example, step S2101 may be implemented as a standalone embodiment. For example, steps S2101 and S2102 may be implemented as standalone embodiments. For example, step S2101 combined with steps S2103 to S2107 may be implemented as a standalone embodiment.

[0319] In some embodiments, steps S2102 to S2107 are optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0320] In some embodiments, steps S2103 to S2107 are optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0321] In some embodiments, step S2102 is optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0322] Figure 2B is a schematic diagram of an interaction of a communication method according to an exemplary embodiment. As shown in Figure 2B, this disclosure relates to a communication method for a communication system 100, the method comprising:

[0323] In this embodiment of the disclosure, the federated learning process is illustrated using the example of both the first node and the second node being deployed on an access network device.

[0324] In some embodiments, when the access network equipment is an integrated architecture (i.e., the CU and DU are not separated), the first node can be deployed in access network equipment such as a base station and / or a macro base station, and the second node can be deployed in access network equipment such as a base station, a small base station, and / or a macro base station. The first node and the second node are deployed in different access network equipment, as shown in Figures 1F and 1I.

[0325] In some embodiments, when the access network device has a CU-DU separation architecture, the first node can be deployed in the CU of the access network device. The second node can be deployed in the CU of the access network device, and / or, in the DU of the access network device, as shown in Figure 1H.

[0326] In some embodiments, when the access network device has a CU-DU separate and CP-UP separate architecture, the first node can be deployed in the CU-CP of the access network device, and / or, in the CU-UP of the access network device. The second node can be deployed in the CU-CP of the access network device, in the CU-UP of the access network device, and / or, in the DU of the access network device, as shown in Figure 1I.

[0327] In some embodiments, the first node and the second node can be deployed in different access network devices. Alternatively, the first node and the second node can be deployed in the same access network device, for example, in the CU and DU of the same access network device.

[0328] Step S2201: The first access network device sends a first message to at least one second access network device and / or at least one terminal.

[0329] In some embodiments, the first node is deployed in the first access network device, and the second node is deployed in the second access network device.

[0330] In some embodiments, the first access network device sends a first message to at least one second access network device.

[0331] In some embodiments, at least one second access network device receives a first message sent by the first access network device.

[0332] In some embodiments, a first access network device may send a first signaling message to at least one second access network device through a first interface, the first signaling message including a first message.

[0333] In some embodiments, the first interface may be a communication interface between access network devices.

[0334] In some embodiments, the first access network device may send a first signaling message to at least one second access network device via the Xn interface.

[0335] In some embodiments, the Xn interface is used to connect different access network devices. Different access network devices can communicate with each other through the Xn interface.

[0336] In one embodiment, when the first access network device is a macro base station and the second access network device is a small base station, the macro base station can send a first signaling message to at least one small base station through the Xn interface.

[0337] In one embodiment, when the first access network device is a first macro base station and the second access network device is a second macro base station, the first macro base station can send a first signaling to at least one second macro base station through the Xn interface.

[0338] In some embodiments, the Xn interface may include a user plane Xn-U interface and a control plane Xn-C interface. When the access network device supports CU-DU separation, the Xn-U interface may be a user plane interface between different CUs, and the Xn-C interface may be a control plane interface between different CUs.

[0339] In some embodiments, when the first access network device is a first CU and the second access network device is a second CU, the first CU can send a first signaling to at least one second CU through the Xn-C interface.

[0340] In some embodiments, when the first access network device is a CU and the second access network device is a DU, the CU can send a first signaling to at least one DU through the F1 interface.

[0341] In some embodiments, the F1 interface is used to connect the CU and DU. The CU and DU can communicate with each other via the F1 interface.

[0342] In some embodiments, the F1 interface may include a user plane F1-U interface and a control plane F1-C interface. When the access network device supports CU-DU separation and CP-UP separation, the F1-U interface may be a communication interface between the CU-UP and DU, and the F1-C interface may be a communication interface between the CU-CP and DU.

[0343] In some embodiments, when the first access network device is a CU-CP and the second access network device is a DU, the CU-CP can send a first signaling message to at least one DU via the F1-C interface.

[0344] In some embodiments, when the first access network device is a CU-UP and the second access network device is a DU, the CU-UP can send a first signaling message to at least one DU via the F1-U interface.

[0345] In some embodiments, when the first access network device is a CU-CP and the second access network device is a CU-UP, the CU-CP can send the first signaling to at least one CU-UP through the E1 interface.

[0346] In some embodiments, the E1 interface is a point-to-point interface between CU-CP and CU-UP, used for information exchange between the user plane and control plane of the access network device.

[0347] In some embodiments, where the first access network device and the second access network device support a service-based interface, the first access network device can send a first message to at least one second access network device through the service-based interface.

[0348] In some embodiments, both the first access network device and the second access network device may have access network functions.

[0349] In some embodiments, the service-oriented interface can also be used for communication between access network functions.

[0350] In some embodiments, the first access network device sends a first message to at least one terminal.

[0351] In some embodiments, at least one terminal receives a first message sent by a first access network device.

[0352] In some embodiments, the first access network device can send a first message to at least one terminal through a second interface. The second interface is the interface between the access network device and the terminal. In one embodiment, the second interface can be a Uu interface. In another embodiment, the Uu interface is used to transmit user data and control information.

[0353] In some embodiments, the first access network device sends a first message to at least one second access network device and at least one terminal.

[0354] In some embodiments, at least one second access network device and at least one terminal receive a first message sent by the first access network device.

[0355] In some embodiments, before sending a first message to at least one second access network device, the first access network device determines at least one second access network device and at least one terminal participating in federated learning. In one embodiment, the first access network device may first select federated learning members and then send the first message to the selected federated learning members.

[0356] In some embodiments, before the first access network device sends a first message to at least one second access network device, the first access network device may receive fifth information sent by at least one second access network device, the fifth information being used to indicate at least one candidate second access network device and / or at least one candidate terminal.

[0357] In some embodiments, the fifth information may be used by the first access network device to determine at least one second access network device and / or at least one terminal participating in federated learning.

[0358] In some embodiments, upon receiving the sixth message, the first access network device may send a first message to at least one second access network device and / or at least one terminal. The sixth message is used to request federated learning services.

[0359] In one embodiment, the first access network device can act as a provider of the federated learning service. The first access network device can receive sixth information sent by a consumer of the federated learning service. In one embodiment, the consumer of the federated learning service can be an application function (AF) or an application service (AS), etc.

[0360] In some embodiments, the first access network device may determine the first message based on the sixth information. In one embodiment, the first access network device may determine the model to be trained in the federated learning task and the total number of training rounds of the model to be trained based on the sixth information.

[0361] In some embodiments, the first message is used by the terminal and / or the second access network device to perform federated learning based on the first model.

[0362] In some embodiments, the first message sent by the first access network device to the terminal is used by at least one terminal to perform a training operation based on the first model.

[0363] In some embodiments, the first message sent by the first access network device to the second access network device can be used by the second access network device to perform a first aggregation operation based on the first model parameters sent by at least one next-level device to obtain the second model parameters.

[0364] In some embodiments, the first message sent by the first access network device to the second access network device can also be used by the second access network device to send a first message to at least one terminal; the first message sent by the second access network device to at least one terminal can be used by at least one terminal to perform training operations based on the first model.

[0365] In some embodiments, the first message may include at least one of the following:

[0366] The first instruction information is used to indicate the first model;

[0367] The second instruction information is used to indicate the task identifier of the first task; the first task is a federated learning task based on the first model.

[0368] The third instruction information is used to indicate the training round of the first model in the first task.

[0369] In some embodiments, the first indication information may be used by the second access network device and / or terminal to determine the first model to be trained.

[0370] In some embodiments, the first indication information may be used to indicate one of the following: the meta information of the first model; the model structure of the first model; the model parameters of the first model; and the training configuration of the first model.

[0371] In some embodiments, the meta-information of the first model is information describing the model attributes of the first model.

[0372] In some embodiments, the metadata of the first model may include, but is not limited to, at least one of the following: the model version of the first model; the model identifier of the first model; the model size of the first model; and the supplier identifier of the first model.

[0373] In some embodiments, the model structure of the first model is used to indicate the network structure of the first model.

[0374] In some embodiments, the model parameters of the first model may include trainable parameters and non-trainable parameters. The values ​​of trainable parameters are learned during model training, such as weights. Non-trainable parameters are fixed and do not update during model training, such as the loss function.

[0375] In some embodiments, the training configuration of the first model is used by the second access network device and / or terminal to determine the learning parameters for federated learning of the first model.

[0376] In some embodiments, the training configuration of the first model may include, but is not limited to, at least one of the following: training mode; aggregation algorithm.

[0377] In some embodiments, the task identifier of the first task can be any information used to identify the first task.

[0378] In some embodiments, task identifiers can be used by second access network devices and / or terminals to distinguish different federated learning tasks.

[0379] In some embodiments, the third indication information may include: round information, used to indicate the total number of training rounds of the first model in the first task.

[0380] In one embodiment, when the total training rounds of the first model in the first task are 1, after the second access network device receives the first model parameters sent by at least one next-level device, it can perform a first aggregation operation based on at least one first model parameter to obtain the second model parameter, and send the second model parameter to the first access network device so that the first access network device can determine the model parameters of the third model based on the second model parameter, thereby obtaining the learning result of the first task.

[0381] In one embodiment, when the total number of training rounds for the first model in the first task is H, after receiving the first model parameters sent by at least one next-level device, the second access network device can perform a first aggregation operation based on the at least one first model parameter to obtain second model parameters. The second model parameters are then sent to at least one next-level device (e.g., a terminal), allowing the terminal to update the first model to be trained based on the first model parameters and train the updated first model to resend the first model parameters to the second access network device. This process is repeated until the terminal has trained the first model for H rounds. At this point, the second access network device sends the second model parameters to the first access network device so that the first access network device can determine the model parameters of the third model based on the second model parameters, thereby obtaining the learning result of the first task. Here, H is a positive integer greater than 1.

[0382] In some embodiments, the third indication information may further include: a round identifier, used to indicate the number of training rounds of the first model in the first task.

[0383] In some embodiments, when a first access network device sends a first message to at least one second access network device, the first message further includes fourth indication information, which is used to indicate a subtask performed by at least one terminal associated with the second access network device within a training round.

[0384] In some embodiments, the fourth indication information can be used by the second access network device to determine which terminals among at least one terminal associated with it are participating in federated learning. The at least one terminal associated with the second access network device accesses the network through the second access network device.

[0385] In some embodiments, the fourth indication information may include: at least one first identifier; the first identifier is used to determine the terminal performing the subtask.

[0386] In one embodiment, the first identifier may be a terminal identifier. In another embodiment, the first identifier may be a tuple including a terminal identifier and a subtask identifier.

[0387] In some embodiments, when the first access network device sends a first message to at least one terminal, the first message further includes fifth indication information for indicating a sub-task performed by the terminal within a training round.

[0388] In some embodiments, the fifth indication information may be a first identifier; the first identifier is used to determine the terminal performing the subtask.

[0389] In one embodiment, the first identifier may be a terminal identifier. In another embodiment, the first identifier may be a tuple including a terminal identifier and a subtask identifier.

[0390] In some embodiments, a first message sent by a first access network device to at least one second access network device may include at least one of the following: a first indication message; a second indication message; a third indication message; or a fourth indication message.

[0391] In some embodiments, a first message sent by a first access network device to at least one terminal may include at least one of the following: a first indication message; a second indication message; a third indication message; or a fifth indication message.

[0392] In some embodiments, the first message may include time information, which is used to instruct the terminal and / or the second node to determine the time for performing the first task.

[0393] In some embodiments, the time information may include at least one of the following: first time information, used to indicate the deadline for the terminal to complete the training operation within a training round of the first task; and second time information, used to indicate the deadline for the second node to complete the first aggregation operation within a training round of the first task.

[0394] Step S2202: The second access network device sends a first message to at least one terminal.

[0395] In some embodiments, other optional implementations of step S2202 can be found in the optional implementations of step S2102 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0396] Step S2203: The terminal performs training operations based on the first model.

[0397] In some embodiments, other optional implementations of step S2203 can be found in the optional implementations of step S2103 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0398] Step S2204: The terminal sends the second information to the second access network device.

[0399] In some embodiments, other optional implementations of step S2204 can be found in the optional implementations of step S2104 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0400] Step S2205: The second access network device performs a first aggregation operation based on at least one first model parameter.

[0401] In some embodiments, other optional implementations of step S2205 can be found in the optional implementations of step S2105 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0402] Step S2206: The second access network device sends the first information to the first access network device.

[0403] In some embodiments, the second access network device may send first information to the first access network device through the first interface.

[0404] In some embodiments, the first interface may be a communication interface between access network devices.

[0405] In some embodiments, the second access network device may send first information to the first access network device via the Xn interface.

[0406] In some embodiments, where the first access network device and the second access network device support service-based interfaces, the second access network device can send first information to the first access network device through the service-based interface.

[0407] In some embodiments, the first information may be used to trigger the first access network device to perform a second aggregation operation.

[0408] In some embodiments, the first information may include second model parameters, which are used by the first access network device to determine the third model.

[0409] In some embodiments, the third model is the learning result of federated learning based on the first model.

[0410] In some embodiments, the first information may further include: second indication information for indicating a task identifier of the first task; the first task is a federated learning task based on the first model; fourth indication information for indicating a sub-task performed by at least one terminal associated with the second access network device within a training round; and round identifier for indicating the number of training rounds of the first model in the first task.

[0411] Step S2207: The first access network device performs a second aggregation operation based on at least one second model parameter to obtain the model parameters of the third model.

[0412] In some embodiments, other optional implementations of step S2207 can be found in the optional implementations of step S2107 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0413] The communication method involved in the embodiments of this disclosure may include at least one of steps S2201 to S2207. For example, step S2201 may be implemented as a standalone embodiment. For example, steps S2201 and S2202 may be implemented as standalone embodiments. For example, step S2201 combined with steps S2203 to S2207 may be implemented as a standalone embodiment.

[0414] In some embodiments, steps S2202 to S2207 are optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0415] In some embodiments, steps S2203 to S2207 are optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0416] In some embodiments, step S2202 is optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0417] Figure 3 is an interactive schematic diagram of a communication method according to an exemplary embodiment. As shown in Figure 3, this disclosure relates to a communication method for a communication system 100, the method including one of the following steps:

[0418] Step S3101: The first node sends a first message to at least one second node and / or at least one terminal.

[0419] In some embodiments, at least one second node and / or at least one terminal receives a first message sent by the first node.

[0420] In some embodiments, the first message is used by the terminal and / or the second node to perform federated learning based on the first model.

[0421] In some embodiments, the first node is deployed in the core network or the access network, and the second node is deployed in the access network.

[0422] In some embodiments, the first node sends a first message to at least one second node, including one of the following:

[0423] The first node sends a first control plane signaling to at least one second node, the first control plane signaling including a first message; the first node is deployed in the core network;

[0424] The first node sends a first signaling message to at least one second node through a first interface. The first signaling message includes a first message. The first node is deployed in the access network, and the first interface is a communication interface between access network devices.

[0425] The first node sends a first message to at least one second node through a service-oriented interface; the first node is deployed in the core network or the access network, and the access network supports the service-oriented interface.

[0426] In some embodiments, the first control plane signaling is transmitted based on a first protocol layer, which is a control plane-based protocol layer and is related to artificial intelligence (AI).

[0427] In some embodiments, the first node sends a first signaling message to at least one second node through a first interface, including one of the following:

[0428] The first node sends a first signaling message to at least one second node via the Xn interface; the first node and the second node are deployed in different access network devices;

[0429] The first node sends a first signaling message to at least one second node via the F1 interface; the first node is deployed in a centralized unit (CU), and the second node is deployed in a distributed unit (DU);

[0430] The first node sends a first signaling message to at least one second node through the F1-C interface of the control plane of F1; the first node is deployed in the control plane CU-CP of CU, and the second node is deployed in DU;

[0431] The first node sends a first signaling message to at least one second node via the E1 interface; the first node is deployed in CU-CP, and the second node is deployed in the user plane CU-UP of the CU;

[0432] The first node sends a first signaling message to at least one second node through the F1 user plane F1-U interface; the first node is deployed in CU-UP and the second node is deployed in DU.

[0433] In some embodiments, other optional implementations of step S3101 can be found in the optional implementations of step S2101 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0434] In some embodiments, the first node sends a first message to at least one terminal, including one of the following:

[0435] The first node sends a second control plane signaling to at least one terminal, the second control plane signaling including a first message; the first node is deployed in the core network;

[0436] The first node sends user plane signaling to at least one terminal, and the user plane signaling includes a first message; the first node is deployed in the core network.

[0437] The first node sends a second signaling message to at least one terminal through the second interface. The second signaling message includes the first message. The first node is deployed in the access network, and the second interface is the communication interface between the access network equipment and the terminal.

[0438] In some embodiments, the second control plane signaling is transmitted based on the first protocol layer; the first protocol layer is a control plane-based protocol layer and is related to AI.

[0439] In some embodiments, user plane signaling is transmitted based on a second protocol layer; the second protocol layer is a user plane-based protocol layer and is related to AI.

[0440] In some embodiments, other optional implementations of step S3101 can be found in the optional implementations of step S2101 in FIG2A and step S2201 in FIG2B, as well as other related parts in the embodiments involved in FIG2A and FIG2B, which will not be repeated here.

[0441] To better understand the embodiments of this disclosure, the following exemplary embodiments will be used to further illustrate this disclosure.

[0442] In some embodiments, the hierarchical federated learning architecture has two logical nodes: a central federated learning node and a branch federated learning node. It is understood that the central federated learning node is the first node in the above embodiments, and the branch federated learning node is the second node in the above embodiments. There are two ways to specifically implement the central federated learning node and the branch federated learning node:

[0443] Implementation Method 1: Implemented by defining Network Functions (NFs). These network functions can be deployed in the core network, access network, CU, DU, CU-CP, CU-UP, Master Node (MN), Secondary Node (SN), etc. The different deployment configurations of the central federated learning node and branch federated learning nodes can be seen in Figures 1F to 1I. This method has three sub-methods:

[0444] Sub-method 1a: Define two network functions. The first network function corresponds to the aforementioned central federated learning node, and the second network function corresponds to the aforementioned branch federated learning node.

[0445] Sub-method 1b: Define a network element, such as a third network function, corresponding to the aforementioned central federated learning node and branch federated learning node.

[0446] Sub-method 1c: Reuse existing network functions, such as network NWDAF and DCCF, which support the functions of the aforementioned central federated learning node and branch federated learning node.

[0447] Implementation Method 2: Implemented by defining specific signaling between network nodes / elements, for example:

[0448] The signaling between the NF corresponding to the FL server and the access network device is defined in the signaling between the central federated learning node and the branch federated learning node. This signaling can be transmitted through the control plane signaling between the access network device and the core network, the user plane signaling between the access network device and the core network, the SBI protocol stack transmission, or the first protocol stack transmission.

[0449] Sub-method 2a: When the central federated learning node is located in the core network, information transmission between the central federated learning node and the branch federated learning nodes is achieved through control plane signaling between the access network equipment and the central federated learning node (located in the core network). This control plane signaling is forwarded through AMF. It is assumed that the control plane signaling is transmitted at a new protocol layer (AI layer in Figure 1J).

[0450] Sub-method 2b: When the central federated learning node is located in the core network, information transmission between the central federated learning node and the branch federated learning nodes is achieved through user plane signaling between the base station and the central federated learning node (located in the core network).

[0451] Sub-mode 2c: When the central federated learning node is located in the core network, information transmission between the central federated learning node and the terminal is achieved through control plane signaling between the central federated learning node (located in the core network) and the terminal. As shown in Figure 1L.

[0452] Sub-mode 2d: When the central federated learning node is located in the core network, information transmission between the central federated learning node and the terminal is achieved through user plane signaling between the central federated learning node (located in the core network) and the terminal. As shown in Figure 1M, this user plane signaling is forwarded via UPF. It is assumed that the terminal and the central federated learning node use the TCP / IP protocol.

[0453] Sub-method 2e: Define the signaling between the central federated learning node and the branch federated learning nodes, as well as the signaling between the branch federated learning nodes, in the Xn signaling.

[0454] Sub-method 2f: Define the signaling between the central federated learning node and the branch federated learning nodes, as well as the signaling between the branch federated learning nodes, in the F1 signaling.

[0455] In some embodiments, there are two modes regarding whether branch federated learning nodes send the aggregated branch model to the terminal in hierarchical federated learning: mode 1 and mode 2.

[0456] In Mode 1 (i.e., when the total number of rounds H=1), the terminal only performs local training based on the global model. The branch federated learning nodes are only responsible for aggregating the local model obtained by the terminal and / or the branch models obtained by other branch federated learning nodes and uploading them to the next higher-level branch federated learning node or the central federated learning node for aggregation. The branch federated learning nodes do not send the branch models to the terminal for retraining.

[0457] In Mode 2 (i.e., when the total number of rounds H>1), the branch federated learning node sends the branch model to the terminal for retraining. After each training round H, the branch model is uploaded to the next higher-level branch federated learning node or the central federated learning node for aggregation.

[0458] In some embodiments of hierarchical federated learning, the central federated learning node sends the global model to the terminals in two ways. Method A involves the central federated learning node sending the global model to each terminal individually. Method B involves the central federated learning node sending the global model to branch federated learning nodes, which then send the global model to the next level branch federated learning nodes or the terminals. The advantage of Method B is reduced network-side transmission overhead.

[0459] In some embodiments, in hierarchical federated learning, the task of local training for each terminal requires the following identifier:

[0460] Task identifiers for federated learning: used to identify / distinguish different federated learning tasks;

[0461] Round identifier: Used to identify / distinguish different rounds within the same federated learning task. For Mode 2 above, a round identifier can be used to indicate multiple rounds of training within a branch federated learning node, or a round identifier plus a sub-round identifier can be used.

[0462] Subtask identifier: Used to identify / distinguish different terminals performing tasks within the same round. This subtask identifier can be a UE identifier or a unique identifier within a tuple (task identifier, round identifier in federated learning).

[0463] In some embodiments, when the central federated learning node sends the global model to each terminal individually, the central federated learning node also indicates the task identifier and round identifier of the federated learning to the terminal. When the subtask identifier is not the terminal identifier, the central federated learning node indicates the subtask identifier to the terminal. In addition, the central federated learning node can indicate the model's metadata, model structure, model parameters, and training configuration information to the terminal, such as the training mode (SGD or mini-batch gradient descent).

[0464] When the central federated learning node sends the global model to the terminal through branch federated learning nodes, the central federated learning node sends the global model to the branch federated learning nodes and indicates to them the task identifier, round identifier, and a list of subtask identifiers related to the branch federated learning nodes participating in this round, or a list of tuples (terminal identifier, subtask identifier). When the subtask identifier is not the terminal identifier, the list of the aforementioned tuples is used. The central federated learning node can indicate the model's metadata, model structure, model parameters, and training configuration information to the branch federated learning nodes, such as the training mode (SGD or mini-batch gradient descent) and the aggregation algorithm (e.g., FedAvg).

[0465] The information sent by a branch federated learning node when sending the global model to the next level branch federated learning node is similar to the information sent by the central federated learning node when sending the global model to a branch federated learning node.

[0466] When a branch federated learning node sends the global model to the terminal, it also indicates the task identifier and round identifier of the federated learning process. If the subtask identifier is not the terminal identifier, the branch federated learning node indicates the subtask identifier to the terminal. The branch federated learning node can also indicate the model's metadata, model structure, model parameters, and training configuration information to the terminal, such as the training mode (SGD or mini-batch gradient descent).

[0467] The central federated node can instruct the terminal, the central federated learning node to the branch federated learning node, the branch federated learning node to the next level branch federated learning node, and the branch federated learning node to the terminal to indicate that a task in a specific round needs to be completed before a certain time. In this way, the branch federated learning node and the central federated learning node can complete the aggregation before the corresponding time.

[0468] Once the terminal receives the global model or a branch model, it performs local training and sends the updated local model to the corresponding branch federated learning node or the central federated learning node. The terminal also indicates the task identifier, round identifier, and subtask identifier of the federated learning to the corresponding branch federated learning node or the central federated learning node.

[0469] Branch federated learning nodes aggregate local models obtained from the terminal and / or next-level branch federated learning nodes into branch models, and send the aggregated branch models to the previous-level branch federated learning node or the central federated learning node. Branch federated learning nodes indicate the task identifier and round identifier of the federated learning, as well as a list of subtask identifiers for completing that round.

[0470] For Mode 2 above, the information sent by the branch federated learning node to the terminal and / or the next level branch federated learning node after the aggregation of the branch model is similar to the information sent when sending the global model.

[0471] In one embodiment, as shown in FIG4, FIG4 is a flowchart illustrating a hierarchical federated learning method according to an exemplary embodiment. The hierarchical federated learning architecture deploys two layers of branch federated learning nodes; and the hierarchical federated learning method uses method B to send the global model, with a total number of rounds H=1. The hierarchical federated learning method includes:

[0472] 1. The central federated learning node sends the global model and related information to the branch federated learning node 2.

[0473] 2. Branch Federated Learning Node 2 sends the global model and related information to Branch Federated Learning Node 1.

[0474] 3. Branch Federated Learning Node 1 sends the global model and related information to the terminal.

[0475] 4. The terminal performs local training.

[0476] 5. The terminal sends the updated local model and related information to branch federated learning node 1.

[0477] 6. Branch Federated Learning Node 1 will aggregate the local models sent from each terminal into a branch model.

[0478] 7. Branch Federated Learning Node 1 sends the branch model and related information to Branch Federated Learning Node 2.

[0479] 8. Branch Federated Learning Node 2 will aggregate the branch models sent from the relevant branch federated learning nodes and the local models sent from the terminal into a branch model.

[0480] 9. Branch Federated Learning Node 2 sends the branch model and related information to the central Federated Learning Node.

[0481] 10. The central federated learning node will aggregate the branch models sent from the relevant branch federated learning nodes and the local models sent from the terminals into a global model.

[0482] In some embodiments, a cell may indicate in its system broadcast information that it can serve as a branch federated learning node or indicate one or more frequency points that support branch federated learning. If a UE supports federated learning and is interested in a federated learning task (e.g., is participating in a federated learning task or will participate in a federated learning task), the UE may increase the priority of frequency points or cells that support branch federated learning during cell reselection.

[0483] This disclosure also provides apparatus for implementing any of the above methods. For example, an apparatus is provided that includes units or modules for implementing the steps performed by the terminal in any of the above methods. Alternatively, another apparatus is provided that includes units or modules for implementing the steps performed by a network device (e.g., an access network device, or a core network device) in any of the above methods.

[0484] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC). The functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD). Taking a field-programmable gate array (FPGA) as an example, it can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0485] In this embodiment, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a type of microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), a Deep Learning Processing Unit (DPU), etc.

[0486] Figure 5A is a schematic diagram of the structure of a network device according to an exemplary embodiment. As shown in Figure 5A, the network device 5100 includes: a first transceiver module 5101, configured to send a first message to at least one terminal and / or at least one second node; the first message is used to request the terminal and / or the second node to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network. Optionally, the first transceiver module 5101 is used to instruct at least one of the communication steps (e.g., steps S2101, S2106, S2201, S2205, but not limited thereto) performed by the first node in any of the above communication methods, which will not be described in detail here. Optionally, the network device 5100 may also include a first processing module, which is used to execute at least one of the information processing-related steps (e.g., steps S2107, S22206, but not limited thereto) performed by the first node in any of the above communication methods, which will not be described in detail here.

[0487] Figure 5B is a schematic diagram of the structure of a network device according to an exemplary embodiment. As shown in Figure 5B, the network device 5200 includes: a second transceiver module 5201, configured to receive a first message sent by a first node; send the first message to at least one terminal; the first message is further used to request the terminal to perform federated learning based on a first model; the first message is used to request a second node to perform federated learning based on the first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network. Optionally, the second transceiver module 5201 is used to instruct the second node to perform at least one of the communication steps such as sending and / or receiving in any of the above communication methods (e.g., steps S2101, S2102, S2104, S2106, S2203, S2205, but not limited thereto), which will not be elaborated here. Optionally, the network device 5200 may further include a second processing module, which is used to execute at least one of the information processing-related steps (such as step S2105, step S22204, but not limited thereto) executed by the second node in any of the above communication methods, which will not be described in detail here.

[0488] Figure 5C is a schematic diagram of a terminal structure according to an exemplary embodiment. As shown in Figure 5C, the terminal 5300 includes: a third transceiver module 5301, configured to receive a first message sent by a first node or a second node; the first message is used to request the terminal to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network. Optionally, the third transceiver module 5301 is used to instruct the terminal to perform at least one of the communication steps (e.g., steps S2102, S2104, S2201, S2203, but not limited thereto) in any of the above communication methods, which will not be elaborated here. Optionally, the terminal 5300 may also include a third processing module, which is used to perform at least one of the information processing-related steps (e.g., steps S2103, S22203, but not limited thereto) in any of the above communication methods, which will not be elaborated here.

[0489] Figure 6A is a schematic diagram illustrating the structure of a communication device 6100 according to an exemplary embodiment. The communication device 6100 can be a network device (e.g., an access network device or a core network device), a terminal (e.g., a user equipment), a chip, chip system, or processor that supports the network device in implementing any of the above methods, or a chip, chip system, or processor that supports the terminal in implementing any of the above communication methods. The communication device 6100 can be used to implement the communication methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0490] As shown in Figure 6A, the communication device 6100 includes one or more processors 6101. The processor 6101 can be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit (CPU). The baseband processor can be used to process communication protocols and communication data, while the CPU can be used to control communication devices (e.g., base stations, baseband chips, terminal devices, terminal device chips, DUs or CUs, etc.), execute programs, and process program data. The processor 6101 is used to invoke instructions to cause the communication device 6100 to execute any of the above communication methods.

[0491] In some embodiments, the communication device 6100 further includes one or more memories 6102 for storing instructions. Optionally, all or part of the memories 6102 may also be located outside the communication device 6100.

[0492] In some embodiments, the communication device 6100 further includes one or more transceivers 6103. When the communication device 6100 includes one or more transceivers 6103, the communication steps such as sending and receiving in the above method are performed by the transceivers 6103, and other steps are performed by the processor 6101.

[0493] In some embodiments, a transceiver may include a receiver and a transmitter, which may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, transceiver circuit, etc., may be used interchangeably; the terms transmitter, transmitting unit, transmitter, transmitting circuit, etc., may be used interchangeably; and the terms receiver, receiving unit, receiver, receiving circuit, etc., may be used interchangeably.

[0494] Optionally, the communication device 6100 further includes one or more interface circuits 6104 connected to the memory 6102. The interface circuits 6104 can be used to receive signals from the memory 6102 or other devices, and can be used to send signals to the memory 6102 or other devices. For example, the interface circuits 6104 can read instructions stored in the memory 6102 and send the instructions to the processor 6101.

[0495] The communication device 6100 described in the above embodiments may be a network device or a terminal, but the scope of the communication device 6100 described in this disclosure is not limited thereto, and the structure of the communication device 6100 may not be limited by FIG. 6A. The communication device may be a standalone device or a part of a larger device. For example, the communication device may be: (1) a standalone integrated circuit IC, or chip, or chip system or subsystem; (2) a collection of one or more ICs, optionally, the IC collection may also include storage components for storing data and programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, smart terminal device, cellular phone, wireless device, handheld device, mobile unit, vehicle device, network device, cloud device, artificial intelligence device, etc.; (6) others, etc.

[0496] Figure 6B is a schematic diagram of a chip 6200 according to an exemplary embodiment. For cases where the communication device 6100 can be a chip or a chip system, the schematic diagram of the chip 6200 shown in Figure 6B can be referenced, but is not limited thereto.

[0497] Chip 6200 includes one or more processors 6201, which are used to invoke instructions to cause chip 6200 to execute any of the above communication methods.

[0498] In some embodiments, chip 6200 further includes one or more interface circuits 6202 connected to memory 6203. Interface circuits 6202 can be used to receive signals from memory 6203 or other devices, and can also be used to send signals to memory 6203 or other devices. For example, interface circuit 6202 can read instructions stored in memory 6203 and send those instructions to processor 6201. Optionally, terms such as interface circuit, interface, transceiver pin, and transceiver can be used interchangeably.

[0499] In some embodiments, chip 6200 further includes one or more memories 6203 for storing instructions. Optionally, all or part of the memories 6203 may be located outside of chip 6200.

[0500] This disclosure also provides a storage medium storing instructions that, when executed on a communication device 6100, cause the communication device 6100 to perform any of the methods described above. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but it can also be a storage medium readable by other devices. Optionally, the storage medium can be a non-transitory storage medium, but it can also be a temporary storage medium.

[0501] This disclosure also provides a program product, which, when executed by a communication device 6100, causes the communication device 6100 to perform any of the above communication methods. Optionally, the program product is a computer program product.

[0502] This disclosure also provides a computer program that, when run on a computer, causes the computer to perform any of the above communication methods.

[0503] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0504] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A communication method, wherein, Executed by the first node, the method includes: Send a first message to at least one terminal and / or at least one second node; the first message is used by the terminal and / or the second node to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network.

2. The method according to claim 1, wherein, Sending the first message to at least one second node includes one of the following: Send a first control plane signaling message to the at least one second node, the first control plane signaling message including the first message; the first node is deployed in the core network; A first signaling message is sent to the at least one second node through a first interface, the first signaling message including the first message; the first node is deployed in the access network, and the first interface is a communication interface between access network devices; A first message is sent to the at least one second node through a service-oriented interface; the first node is deployed in the core network or the access network, and the access network supports service-oriented interfaces.

3. The method according to claim 2, wherein, The first control plane signaling is transmitted based on the first protocol layer, which is a control plane-based protocol layer and is related to artificial intelligence (AI).

4. The method according to claim 2, wherein, Sending the first signaling to the at least one second node through the first interface includes one of the following: The first signaling is sent to the at least one second node via the Xn interface; the first node and the second node are deployed in different access network devices; The first signaling is sent to the at least one second node via the F1 interface; the first node is deployed in a centralized unit (CU), and the second node is deployed in a distributed unit (DU); The first signaling is sent to the at least one second node via the F1-C interface of the control plane of F1; the first node is deployed in the control plane CU-CP of CU, and the second node is deployed in DU; The first signaling is sent to the at least one second node via the E1 interface; the first node is deployed in CU-CP, and the second node is deployed in the user plane CU-UP of the CU; The first signaling is sent to the at least one second node via the F1 user plane F1-U interface; the first node is deployed in CU-UP and the second node is deployed in DU.

5. The method according to claim 1, wherein, Sending the first message to at least one terminal includes one of the following: Send a second control plane signaling message to the at least one terminal, the second control plane signaling message including the first message; the first node is deployed in the core network; Send user plane signaling to the at least one terminal, the user plane signaling including the first message; the first node is deployed in the core network; A second signaling message is sent to the at least one terminal through a second interface, the second signaling message including the first message; the first node is deployed in the access network, and the second interface is a communication interface between the access network device and the terminal.

6. The method according to claim 5, wherein, The second control plane signaling is transmitted based on the first protocol layer; the first protocol layer is a control plane-based protocol layer and is associated with AI; and / or, the user plane signaling is transmitted based on the second protocol layer; the second protocol layer is a user plane-based protocol layer and is associated with AI.

7. The method according to any one of claims 1 to 6, wherein, The first message includes at least one of the following: First indication information is used to indicate the first model; The second indication information is used to indicate the task identifier of the first task; the first task is a federated learning task based on the first model. The third indication information is used to indicate the training round of the first model in the first task.

8. The method according to claim 7, wherein, The first message sent by the first node to the at least one second node further includes fourth indication information, used to indicate a subtask performed by at least one terminal associated with the second node during the training round, wherein the at least one terminal associated with the second node accesses the network through the second node. or, The first message sent by the first node to the at least one terminal further includes fifth indication information, which is used to indicate the sub-tasks performed by the terminal within the training round.

9. The method according to any one of claims 1 to 8, wherein, The first message is used by the terminal and / or the second node to perform federated learning based on the first model, including at least one of the following: The first message is used by at least one of the terminals to perform a training operation based on the first model; The first message is used by the second node to perform a first aggregation operation based on the first model parameters sent by at least one next-level device. The first node performs federated learning to obtain second model parameters; the second model parameters are used at least to update the first model to be trained on the terminal, and / or the first node determines a third model; the third model is the learning result of federated learning based on the first model.

10. The method according to claim 9, wherein, The first model parameter is used to indicate the model parameters of the trained first model; the next-level device of the second node is the terminal; Alternatively, the first model parameters are used to indicate the aggregation result of the first aggregation operation; the aggregation result is determined by the model parameters of at least one trained first model; the next-level device of the second node is another second node.

11. The method according to claim 9 or 10, wherein, The method further includes: Receive first information sent by the at least one second node, the first information including the second model parameters; A second aggregation operation is performed based on at least one second model parameter to obtain the model parameters of the third model.

12. The method according to any one of claims 9 to 11, wherein, The first message includes: time information, which is used by the terminal and / or the second node to determine the time for executing the first task; the first task is a federated learning task based on the first model.

13. The method according to claim 12, wherein, The time information includes at least one of the following: First-time information is used to indicate the deadline for the terminal to complete the training operation within one training round of the first task; The second time information is used to indicate the deadline for the second node to complete the first aggregation operation within one training round of the first task.

14. A communication method, wherein, Executed by the second node, the method includes: The system receives a first message sent by a first node; the first message is used by the second node to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network. Send a first message to at least one terminal; the first message is also used by the terminal to perform federated learning based on the first model.

15. The method according to claim 14, wherein, The first message sent by the first node is received, including one of the following: The first control plane signaling sent by the first node is received, and the first control plane signaling includes the first message; the first node is deployed in the core network. The system receives a first signaling message sent by the first node through a first interface, the first signaling message including the first message. The first node is deployed in the access network, and the first interface is a communication interface between access network devices; Receive the first message sent by the first node through the service interface; The first node is deployed in the access network or core network, and the access network supports service-oriented interfaces.

16. The method according to claim 15, wherein, The first control plane signaling is transmitted based on the first protocol layer, which is a control plane-based protocol layer and is associated with AI.

17. The method according to claim 16, wherein, The step of receiving the first signaling sent by the first node through the first interface includes one of the following: The first signaling sent by the first node is received through the Xn interface; the first node and the second node are deployed in different access network devices; The first signaling sent by the first node is received via the F1 interface; the first node is deployed in the CU, and the second node is deployed in the DU; The first signaling sent by the first node is received via the F1-C interface; the first node is deployed in CU-CP, and the second node is deployed in DU; The first signaling sent by the first node is received via the E1 interface; the first node is deployed in CU-CP, and the second node is deployed in CU-UP. The first signaling sent by the first node is received through the user plane F1-U interface of F1; the first node is deployed in CU-UP and the second node is deployed in DU.

18. The method according to any one of claims 14 to 17, wherein, The first message includes at least one of the following: First indication information is used to indicate the first model; The second indication information is used to indicate the task identifier of the first task; the first task is a federated learning task based on the first model. The third indication information is used to indicate the training round of the first model in the first task.

19. The method according to claim 18, wherein, Sending the first message to at least one terminal includes one of the following: The first message is sent to at least one terminal; the first message also includes fifth indication information, which is used to indicate the sub-tasks performed by the terminal in the training round; the next level device of the second node is the terminal; The first message is sent to another second node; the first message is used by the other second node to send a first message to the at least one terminal; the first message also includes fourth indication information for indicating the other second node during the training round. The subtask is executed by at least one terminal associated with the second node, and the at least one terminal associated with the other second node accesses the network through the other second node; the next level device of the second node is the other second node.

20. The method according to any one of claims 14 to 19, wherein, The first message is used by the second node to perform federated learning based on the first model, and includes at least one of the following: The first message is used by the second node to trigger at least one of the terminals to perform a training operation based on the first model; The first message is used by the second node to perform a first aggregation operation based on the first model parameters sent by at least one next-level device to obtain second model parameters; the second model parameters are used at least by the terminal to update the first model to be trained, and / or, the first node determines a third model; the third model is the learning result of federated learning based on the first model.

21. The method according to claim 20, wherein, The first model parameter is used to indicate the model parameters of the trained first model; the next-level device of the second node is the terminal; or, the first model parameter is used to indicate the aggregation result of the first aggregation operation; the aggregation result is determined by the model parameters of at least one trained first model; the next-level device of the second node is another second node.

22. The method according to claim 20, wherein, The method further includes: Receive second information sent by at least one next-level device, the second information including the first model parameters; The first aggregation operation is performed based on at least one first model parameter to obtain the second model parameter.

23. The method according to claim 22, wherein, The method further includes: Send first information to the first node, the first information including the second model parameters; the first information is used by the first node to perform a second aggregation operation based on the second model parameters.

24. The method according to claim 22, wherein, The method further includes: The terminal sends third information, which includes the second model parameters; the third information is used by the terminal to update the first model to be trained based on the second model parameters.

25. The method according to any one of claims 20 to 24, wherein, The first message sent by the second node to at least one terminal is transmitted based on the third protocol layer; the second node is deployed in the DU; The third protocol layer is at least used by the DU to perform security protection and reordering of the first message.

26. The method according to any one of claims 20 to 25, wherein, The first message includes: time information, which is used by the terminal and / or the second node to determine the time for executing the first task; the first task is a federated learning task based on the first model.

27. The method according to claim 26, wherein, The time information includes at least one of the following: First-time information is used to indicate the deadline for the terminal to complete the training operation within one training round of the first task; The second time information is used to indicate the deadline for the second node to complete the first aggregation operation within one training round of the first task.

28. The method according to any one of claims 14 to 27, wherein, The method further includes: Send a fourth message, which is contained in a system message; the fourth message includes at least one of the following: The sixth instruction information is used to indicate one or more first cells provided by the access network equipment deployed by the second node; the first cell is a cell that supports federated learning; The seventh indication information is used to indicate one or more first frequency points provided by the access network equipment deployed by the second node, wherein the first frequency point is a frequency point that supports federated learning.

29. The method according to claim 28, wherein, The fourth information is used by the terminal to determine the priority corresponding to the first cell and / or the first frequency point; the priority corresponding to the first cell and / or the first frequency point is used to assist the terminal in cell reselection.

30. A communication method, wherein, The method, executed by a terminal, includes: The terminal receives a first message sent by a first node or a second node; the first message is used for the terminal to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network.

31. The method according to claim 30, wherein, The first message sent by the first node is received, including one of the following: The first node receives a second control plane signaling sent by the first node, the second control plane signaling including the first message; the first node is deployed in the core network. The system receives user plane signaling sent by the first node, the user plane signaling including the first message; the first node is deployed in the core network. The second interface receives a second signaling message sent by the first node, the second signaling message including the first message; the first node is deployed in the access network, and the second interface is a communication interface between the access network device and the terminal.

32. The method according to claim 31, wherein, The second control plane signaling is transmitted based on the first protocol layer, which is a control plane protocol layer and is related to AI; and / or, the user plane signaling is transmitted based on the second protocol layer, which is a user plane protocol layer and is related to AI.

33. The method according to claim 30, wherein, The first message sent by the second node is transmitted based on the third protocol layer; the second node is deployed in the DU; The third protocol layer is at least used by the DU to perform security protection and reordering of the first message.

34. The method according to any one of claims 30 to 33, wherein, The first message includes at least one of the following: First indication information is used to indicate the first model; The second indication information is used to indicate the task identifier of the first task; the first task is a federated learning task based on the first model. The third indication information is used to indicate the training round of the first model in the first task; The fifth instruction information is used to indicate the sub-tasks performed by the terminal within the training round.

35. The method according to any one of claims 30 to 34, wherein, The method further includes: Training operations are performed based on the first model to obtain the first model parameters; the first model parameters are used to indicate the model parameters of the trained first model. Send a second message to the second node, the second message including the first model parameters; the second message is used by the second node to perform a first aggregation operation based on the first model parameters.

36. The method according to claim 35, wherein, The method further includes: The terminal receives third information sent by the second node; the third information includes second model parameters, which are obtained by the second node performing a first aggregation operation based on at least one first model parameter; the third information is used by the terminal to update the first model to be trained based on the second model parameters.

37. The method according to claim 35 or 36, wherein, The first message includes: time information, which is used by the terminal to determine the time to execute the first task; the first task is a federated learning task based on the first model.

38. The method according to claim 37, wherein, The time information: First time information, used to indicate the deadline for the terminal to complete the training operation within one training round of the first task.

39. The method according to any one of claims 30 to 38, wherein, The method further includes: Receive fourth information sent by the second node, the fourth information being contained in a system message, the fourth information including: The sixth instruction information is used to indicate one or more first cells provided by the access network equipment deployed by the second node; the first cell is a cell that supports federated learning; The seventh indication information is used to indicate one or more first frequency points provided by the access network equipment deployed by the second node, wherein the first frequency point is a frequency point that supports federated learning.

40. The method according to claim 39, wherein, The fourth information is used by the terminal to determine the priority corresponding to the first cell and / or the first frequency point; the priority corresponding to the first cell and / or the first frequency point is used to assist the terminal in cell reselection.

41. A communication method, wherein, Performed by a communication system, the method includes: A first node sends a first message to at least one terminal and / or at least one second node; the first message is used by the terminal and / or the second node to perform federated learning based on a first model; the first node is deployed in a core network or an access network, and the second node is deployed in the access network.

42. A communication device, wherein, The communication device is used to perform the communication method according to any one of claims 1 to 13, 14 to 29, and 30 to 40.

43. A communication system, wherein, The communication system includes a first terminal node and a second terminal node; the first terminal node is configured to implement the communication method of any one of claims 1 to 13, the second terminal node is configured to implement the communication method of any one of claims 14 to 29, and the terminal is configured to implement the communication method of any one of claims 30 to 40.

44. A storage medium, wherein, The storage medium stores instructions that, when executed on a communication device, cause the communication device to perform the communication method according to any one of claims 1 to 13, 14 to 29, and 30 to 40.

45. A computer program product comprising a computer program that, when executed by a processor, implements the communication method according to any one of claims 1 to 13, 14 to 29, and 30 to 40.

Images