Control equipment, first terminal equipment, second terminal equipment, third terminal equipment, terminal equipment, methods, storage media, computer program products and equipment used in wireless communication systems

Abstract

The present invention relates to electronic equipment, methods, and storage media used in wireless communication systems. The present invention relates to a control device used in a wireless communication system, which includes a processing circuit configured to receive a plurality of models associated with a plurality of terminal devices in the wireless communication system, wherein the plurality of models include at least one aggregated model, each of which is generated by a corresponding terminal device by aggregating its local model and each local model from one or more other terminal devices, and each local model is obtained by a corresponding terminal device by training on the local data of that terminal device.

Description

[Technical Field]

[0001] (Cross-reference of related applications) This application claims priority over Chinese Patent Application 202310436139.X, filed on April 21, 2023, with the title of the invention being "Electronic device, method and storage medium used in wireless communication systems," and the entirety of the contents of said application is incorporated herein by reference.

[0002] This disclosure generally relates to wireless communication systems, and more specifically to federated learning in wireless communication systems. [Background technology]

[0003] With the advancements in wireless communication and artificial intelligence (AI) / machine learning (ML), the integration of AI and ML into wireless communication systems is being considered. For example, the introduction of AI / ML technology can bring about significant improvements in wireless communication systems in terms of performance, energy efficiency, security, and privacy.

[0004] Traditional AI / ML solutions typically require collecting user data and performing aggregated training on that data. However, such traditional AI / ML solutions can have drawbacks in terms of user privacy and data volume. A new machine learning framework called Federated Learning (FL) has emerged to overcome these drawbacks. In a federated learning system, client devices train using local data based on an initial model determined by the FL server, and upload the resulting local model to the FL server. The FL server aggregates the local models received from each client device and sends the aggregated model to each client device. Each client device then trains again using local data based on the received aggregated model, uploads the resulting new local model to the FL server for a new round of aggregation and training. Federated learning eliminates the need for users to upload their original local data to the server, thus improving user privacy and reducing data transmission volume. Therefore, federated learning is a desirable AI / ML solution for wireless communication systems. [Overview of the project] [Means for solving the problem]

[0005] This disclosure proposes solutions for federated learning in wireless communication systems, and specifically provides control equipment, terminal equipment, methods, and storage media used in wireless communication systems.

[0006] One aspect of the present disclosure relates to a control device used in a wireless communication system, comprising a processing circuit configured to receive a plurality of models associated with a plurality of terminal devices in the wireless communication system, wherein the plurality of models comprises at least one aggregated model, each of which is generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by a corresponding terminal device by training on the local data of that terminal device.

[0007] Another aspect of the present disclosure relates to a first terminal device used in a wireless communication system, and includes a processing circuit configured to upload a first local model, which is obtained by training on local data, to a control device of the wireless communication system, and to transmit the first local model to a second terminal device of the wireless communication system via sidelink communication for generation of an aggregated model by the second terminal device.

[0008] Another aspect of the present disclosure relates to a second terminal device used in a wireless communication system, and includes a processing circuit configured to receive, via sidelink communication, at least one first local model obtained by training on local data by one of the first terminal devices, each of which is obtained by training on local data by one of the first terminal devices, aggregate the second local model obtained by training on local data and the at least one first local model to generate an aggregated model, and upload the aggregated model to a control device of the wireless communication system.

[0009] Another aspect of the present disclosure relates to a third terminal device used in a wireless communication system, the third terminal device constituting a terminal device group with one or more other terminal devices, the third terminal device includes a processing circuit configured to receive, via sidelink communication, at least one local model obtained by each of the at least one other terminal devices in the terminal device group having been trained on its local data by one of the other terminal devices, the third local model obtained by being trained on its local data and at least one of the received local models to generate an aggregated model, upload the aggregated model to the control device, and transmit the third local model to each of the other terminal devices in the terminal device group via sidelink communication for the generation of an aggregated model by each of the other terminal devices.

[0010] Another aspect of the present disclosure relates to a control device-side method used in a wireless communication system, which includes receiving a plurality of models associated with a plurality of terminal devices in the wireless communication system, the plurality of models including at least one aggregated model, each of the at least one aggregated model being generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model being obtained by training on local data of a corresponding terminal device.

[0011] Another aspect of the present disclosure relates to a method for a first terminal device used in a wireless communication system, which includes uploading a first local model, which has been trained on local data, to a control device of the wireless communication system, and transmitting the first local model to a second terminal device of the wireless communication system via sidelink communication for generation of an aggregated model by the second terminal device.

[0012] Another aspect of the present disclosure relates to a method for a second terminal device used in a wireless communication system, which includes receiving, via sidelink communication, at least one first local model obtained by training on local data by one of the first terminal devices of the wireless communication system, each obtained by training on local data by one of the first terminal devices; aggregating the second local model obtained by training on local data and the at least one first local model to generate an aggregated model; and uploading the aggregated model to a control device of the wireless communication system.

[0013] Another aspect of the present disclosure relates to a method for a third terminal device used in a wireless communication system, wherein the third terminal device constitutes a terminal device group with one or more other terminal devices, and the method includes receiving, via sidelink communication, at least one local model obtained by each of the at least one other terminal devices in the terminal device group having been trained on its local data by one of the other terminal devices; aggregating the third local model obtained by training on the local data with at least one of the received local models to generate an aggregated model, uploading the aggregated model to the control device; and transmitting the third local model to each of the other terminal devices in the terminal device group via sidelink communication for the generation of an aggregated model by each of the other terminal devices.

[0014] Another aspect of the present disclosure relates to a non-temporary, computer-readable storage medium that stores executable instructions, when executed, causing the method described in the above aspect to be implemented. Another aspect of this disclosure relates to a computer program product that, when executed, includes executable instructions that enable the method described in the above aspects.

[0015] Another aspect of this disclosure relates to a device, which includes a processor and a memory device storing executable instructions that, when executed, enable the aforementioned method.

[0016] The above summary is provided to provide a basic understanding of each aspect of the subject matter described herein, and to summarize several exemplary embodiments. Therefore, the above features are merely examples and should not be construed as limiting the scope or spirit of the subject matter described herein. Other features, aspects, and advantages of the subject matter described herein will become apparent from the embodiments for carrying out the invention described in conjunction with the following drawings. [Brief explanation of the drawing]

[0017] A better understanding of the contents of this disclosure can be obtained by considering the following specific descriptions of embodiments in conjunction with the drawings. In all drawings, identical or similar components are indicated by identical or similar reference numerals. Each drawing, together with the following specific descriptions, is included herein and forms part of the specification, intended to illustrate and describe embodiments of this disclosure and to interpret the principles and merits of this disclosure. In the drawings,

[0018] [Figure 1] Figure 1 schematically shows a conventional associative learning network structure. [Figure 2] Figure 2 schematically illustrates the problem of performance degradation in conventional federative learning due to unreliable communication. [Figure 3] Figure 3 schematically shows a first embodiment according to the present disclosure. [Figure 4A] Figure 4A schematically illustrates an exemplary information exchange in the first implementation form of the first embodiment of the present disclosure. [Figure 4B] Figure 4B schematically illustrates an exemplary information exchange in the first implementation form of the first embodiment of the present disclosure. [Figure 5A] Figure 5A schematically illustrates an example of information exchange in a second implementation of the first embodiment of the present disclosure. [Figure 5B]Figure 5B schematically illustrates an exemplary information exchange in a second implementation form of the first embodiment according to this disclosure. [Figure 6] Figure 6 schematically shows a second embodiment according to the present disclosure. [Figure 7A] Figure 7A schematically illustrates the information exchange of the first implementation form of the second embodiment of the present disclosure. [Figure 7B] Figure 7B schematically illustrates the information exchange of the first implementation form of the second embodiment of the present disclosure. [Figure 7C] Figure 7C schematically illustrates the information exchange of the first implementation form of the second embodiment of the present disclosure. [Figure 8A] Figure 8A schematically illustrates an exemplary information exchange in a second implementation form of the second embodiment of the present disclosure. [Figure 8B] Figure 8B schematically illustrates an exemplary information exchange in a second implementation form of the second embodiment of the present disclosure. [Figure 8C] Figure 8C schematically illustrates the information exchange of a second implementation form of the second embodiment of the present disclosure. [Figure 9] Figure 9 schematically illustrates exemplary information exchange for selecting UEs to participate in federative learning according to the embodiments of this disclosure and for dividing the UE groups used for federative learning. [Figure 10] Figure 10 schematically shows the conceptual arrangement of electronic equipment on the federated learning control device side according to an embodiment of the present disclosure. [Figure 11] Figure 11 schematically shows the conceptual operation flow of the federated learning control device side according to an embodiment of the present disclosure. [Figure 12] Figure 12 schematically shows the conceptual arrangement of electronic devices on the terminal device side according to an embodiment of the present disclosure. [Figure 13A] Figure 13A schematically shows the conceptual operation flow of the terminal device side according to the embodiment of this disclosure. [Figure 13B] Figure 13B schematically illustrates another conceptual operation flow on the terminal device side according to an embodiment of this disclosure. [Figure 13C]Figure 13C schematically illustrates another conceptual operation flow on the terminal device side according to an embodiment of this disclosure. [Figure 14] Figure 14 schematically shows the overall architecture of a wireless communication system to which an embodiment of the present disclosure is applied. [Figure 15] Figure 15 is a block diagram showing an exemplary configuration of an information processing device that can be used in an embodiment of the present disclosure. [Figure 16] Figure 16 is a block diagram showing a first example of an exemplary configuration of a gNB to which the technology of this disclosure can be applied. [Figure 17] Figure 17 is a block diagram showing a second example of an exemplary configuration of a gNB to which the technology of this disclosure can be applied. [Figure 18] Figure 18 is a block diagram showing an exemplary configuration of a smartphone to which the technology of this disclosure can be applied. [Figure 19] Figure 19 is a block diagram showing an exemplary arrangement of a car navigation device to which the technology of this disclosure can be applied.

[0019] The embodiments described herein are subject to various modifications and alternative forms, the specific embodiments of which are shown as examples in the drawings and described in detail herein. However, it should be understood that the drawings and their detailed descriptions do not limit the embodiments to any particular form of this disclosure, but rather include all modifications, equivalents, and alternative forms that fall within the spirit and scope of the claims. [Modes for carrying out the invention]

[0020] The following describes typical applications in various aspects of the apparatus and methods described herein. These examples are provided solely to add context and aid in understanding the embodiments described. Therefore, it will be apparent to those skilled in the art that the embodiments described below can be implemented without some or all of the specific details. In other cases, well-known process steps are not described in detail to avoid unnecessarily obscuring the embodiments described. Other applications are possible, and the solutions of this disclosure are not limited to these examples.

[0021] Typically, a wireless communication system includes at least terminal equipment and control equipment.

[0022] In this disclosure, the term “control equipment” has the full scope of its ordinary meaning and may be control equipment on the access network side of a wireless communication system, or equipment on the core network side of a wireless communication system. For example, control equipment in this disclosure may be a wireless communication station that is part of a wireless communication system or radio system to perform communications. Examples of base stations may be, for example, an eNB for the 4G communication standard, a gNB for the 5G NR communication standard, a remote wireless head, a wireless access point, a drone control tower, or a communication device that performs a similar function. Alternatively, for example, control equipment in this disclosure may be one or more network element devices on the core network side of a wireless communication system. For example, a single network element device may perform a single function, a single network element device may perform multiple functions, multiple network element devices may perform a single function, or multiple network element devices may work together to perform multiple functions. In this disclosure, network element device may refer to one or more software and / or hardware modules. In particular, control equipment in this disclosure may be one or more network element devices that perform one or more functions relating to federated learning on the core network side of a wireless system.

[0023] In this disclosure, the terms “Terminal Equipment” or “User Equipment (UE)” have the full scope of their ordinary meanings and include terminal equipment that is at least part of a wireless communication system or radio system for communication. For example, terminal equipment may be terminal equipment or components thereof such as a mobile phone, laptop computer, tablet computer, in-vehicle communication equipment, wearable device, or sensor. In this disclosure, “Terminal Equipment” and “User Equipment” (which may be abbreviated as “UE”) are interchangeable, or “Terminal Equipment” may be implemented as part of “User Equipment”.

[0024] In this disclosure, the expression “the terminal device uploads / transmits (information / data) to the control device” may mean that the terminal device directly transmits information / data to the control device (for example, if the control device is a base station), or that the terminal device transmits information / data to one or more intermediate devices, and that one or more intermediate devices forward the information / data to the control device (for example, if the control device is a network element device), and that one or more intermediate devices may include a base station and, where appropriate, one or more other network element devices in the core network.

[0025] In this disclosure, the expression "the control device notifies / transmits (information / data) to the terminal device" may mean that the control device directly transmits information / data to the terminal device (for example, if the control device is a base station), or that the control device transmits information / data to one or more intermediate devices, and that one or more intermediate devices forward the information / data to the terminal device (for example, if the control device is a network element device), and that one or more intermediate devices may include a base station and, where appropriate, one or more other network element devices in the core network.

[0026] As introduced in the background technology section, federative learning is a desirable AI / ML solution to adopt in wireless communication systems.

[0027] Figure 1 shows a conventional associative learning network structure.

[0028] In the conventional federative learning network structure with K nodes shown in Figure 1, the specific flow of federative learning is as follows: (1) The server determines the initial global model and sends the initial global model to each node. (2) Each node receives the global model from the server (for example, the W shown in Figure 1) t Based on this, the local data D1, D2, ..., D of the node k By using the global model for training (for example, by training the global model), the local model W1 t W1 t , , , W K t To obtain. (3) Each node uploads the trained local model to the server. (4) The server aggregates the local models collected from each node according to equation (1) to form a new global model W t+1 To obtain.

number

number

[0029] This federated learning network structure is applied to a wireless communication system. For example, in this case, each UE may function as one node in the federated learning network structure, and network element devices or base stations (hereinafter abbreviated as control devices) that realize functions related to federated learning in the core network may function as servers in the federated learning network structure. Each UE uploads its local model to the federated learning server using the uplink (if a network element device functions as the server, the UE may also transmit the local model to the base station, and the base station may transfer the local model to the network element device). The federated learning server also transmits a global model to each UE participating in the learning using the downlink (if a network element device functions as the server, the network element device may also transmit the global model to the base station, and the base station may transfer the local model to each UE).

[0030] However, due to the inherent uncertainty in wireless communication systems, when applying this conventional federated learning network structure to wireless communication systems, there are cases where the local models of some UEs (User Entities) fail to be successfully uploaded to the control device. For example, as shown in Figure 2, the upload of UE#1's local model fails. In this case, when the control device updates the global model, it aggregates only the received local models, for example, as shown for UE#2-UE#K in Figure 2. Therefore, the global model generated in the iterative process only considers the local models of some UEs, and the final converged model may be biased towards certain UEs (e.g., UEs with good communication quality). On the other hand, the number of local models adopted during aggregation decreases. These two factors ultimately lead to a decrease in the overall federated learning performance of the wireless communication system.

[0031] In view of this, this disclosure proposes a solution to improve federated learning in wireless communication systems. This disclosure considers compensating for the lack of communication quality in the uplinks (particularly the uplinks) of some UEs by transmitting local models using sidelinks between UEs in a federated learning network, in addition to the conventional uplinks and downlinks. Specifically, according to this disclosure, a control device can receive multiple models associated with multiple UEs in a wireless communication system. These multiple models include at least one aggregated model. For example, each model of the at least one aggregated model is generated by a corresponding UE by aggregating its local model with each local model from one or more other UEs, and each local model is obtained by training on the local data of a corresponding UE.

[0032] (First example) According to a first embodiment of the present disclosure, at least two UEs in a wireless communication system may constitute a UE group used for federated learning. The UE group may include one central UE and at least one decentralized UE, the at least one decentralized UE which may communicate with the central UE via a sidelink. In other words, the UE group may be a centralized UE group. The central UE may receive local models (hereinafter referred to as first local models for convenience of explanation) from other UEs in the UE group via sidelink communication, and each received local model is obtained by training on its local data by a corresponding decentralized UE in the group. After receiving the first local models from each decentralized UE in the group, the central UE may aggregate its own local model (hereinafter referred to as a second local model for convenience of explanation) and each received first local model to generate an aggregated model, and transmit the aggregated model to a control device of the wireless communication system (for example, a network element device or base station that implements the federated learning function in the core network described above). Furthermore, each decentralized UE can transmit its first local model to the central UE via a sidelink, and at the same time, upload the first local model to, for example, the control equipment of the wireless communication system via an uplink. In this way, the success rate of each decentralized UE's local model's contribution to the global model can be further ensured in the federated learning process.

[0033] Figure 3 is a schematic diagram showing the first embodiment. As shown in Figure 3, assume that the wireless communication system may include K UEs. Here, UE#1 to UE#M may form one centralized UE group. For example, in this UE group, UE#M may function as the central UE, and the other UEs may function as non-central UEs. UE#M may receive local models from other UEs in the UE group via sidelink communication (indicated by a dashed line in Figure 3). For example, as shown in Figure 3, UE#1 may transmit the local model W1 t obtained by performing training based on the local data of UE#1 to UE#M via sidelink and may also upload the local model W1 t individually to the control device. UE#2 may transmit the local model W2 t obtained by performing training based on the local data of UE#2 to UE#M via sidelink and may also upload the local model W2 t individually to the control device. After receiving the local models of each non-central UE in the group, UE#M may aggregate the local model obtained by performing training based on the local data of UE#M and the local models received from other non-central UEs to generate an aggregated model W M t+1 / 2 and may upload the aggregated model to the control device.

[0034] According to this disclosure, one or more UEs with poor communication quality (e.g., particularly uplink communication quality) may form a centralized UE group with neighboring UEs with good communication quality (e.g., particularly uplink communication quality). For example, one or more UEs with particularly poor uplink communication quality may form a centralized UE group with UEs that have good sidelink communication quality and good uplink communication quality. Furthermore, after considering communication quality, the computing power of the UEs may be considered to determine whether such UEs can be called the central UE of a centralized UE group. A control device may divide the centralized UE group and determine the central UE based on multiple pieces of information about each UE. The division process and the information on which the division is based will be described in detail later. According to this disclosure, one or more UEs with good communication quality (e.g., particularly uplink communication quality) may not participate in any UE group and may independently upload their local models as in conventional federated learning.

[0035] The basic concept of the first embodiment of this disclosure has been explained in conjunction with Figure 3, and the global model aggregation process according to the first embodiment of this disclosure will now be explained in detail.

[0036] Suppose a wireless communication system contains K UEs. First, let's simply assume that UE#1 and UE#2 constitute a single centralized UE group (for example, let's assume UE#2 is the central UE), and that the other UEs independently upload their local models to the control equipment. In this case, the aggregated model uploaded by UE#2 can be expressed as equation (3).

[0037]

number

[0038] Here, n1 and n2 are the data amounts for UE#1 and UE#2, respectively, and W1 t and W2 t These are the local models for UE#1 and UE#2, respectively.

[0039] When a control device performs aggregation based on the received model, it may aggregate the received local model and the aggregated model in such a way that the local model of the same UE is not aggregated repeatedly. This avoids bias in the contribution of each UE's local model to the global model. Therefore, the global model can be expressed as equation (4).

[0040]

number

[0041] Here, N represents the total amount of data from UE#1 to UE#K, and can be expressed as equation (5). k This is the amount of data in UE#k, U k ={0,1} and U k =1 indicates that the UE#k model upload was successful, U k =0 indicates that the upload of the UE#k model failed. Similarly, U1={0,1} means that U1=1 indicates that the upload of the UE#1 model was successful, and U1=0 indicates that the upload of the UE#1 model failed. U2={0,1} means that U2=1 indicates that the upload of the UE#2 model was successful, and U2=0 indicates that the upload of the UE#2 model failed.

[0042]

number

[0043] As reflected in equations (4) to (5), if the upload of the aggregated model in UE#2 is successful, the control device does not consider the local models uploaded individually by UE#1, and the control device only considers the local models uploaded individually by UE#1 if the upload of the aggregated model in UE#2 fails. Therefore, according to the first embodiment of this disclosure, as long as the upload of at least one model in UE#1 and UE#2 is successful, the model in UE#1 can contribute to the global model without bias in the degree of contribution.

[0044] Specifically, if UE#1 accesses its local model W1 via the uplink... t The probability of failure to send is q1, and the probability of success is p1 = 1 - q1, and UE#2 is aggregated via the uplink to the model W2 t+1 / 2 Let the probability of failure to send a signal be q2, and the probability of success be p2 = 1 - q2. For one iteration of associative learning, if we adopt a conventional associative learning solution (i.e., associative learning without sidelinks), the local model W1 of UE#1 t The probability that this ultimately contributes to the success of the global model is 1-q1. In contrast, when adopting the federative learning solution of the first embodiment of this disclosure, the local model W1 of UE#1 t The probability that this ultimately contributes to the success of the global model is 1 - q1q2. Therefore, the local model W1 of UE#1 t The probability that this ultimately contributes to the success of the global model improves by q1(1-q2). With multiple iterations of associative learning, such improvements may become more pronounced. Furthermore, since the aggregated model generated by aggregation and the local model have the same model size, the solution of the first embodiment does not result in a significant increase in the amount of data transmitted over the uplink in a wireless communication network.

[0045] More generally, a wireless communication system may include K UEs, and of these K UEs, M UEs, i.e., UE#1 to UE#M, constitute a single centralized UE group (for example, let's assume UE#M is the central UE), and the other UEs independently upload their local models to the control equipment. In this case, the aggregated model uploaded by UE#M can be expressed as equation (6).

[0046]

number

[0047] Here, n m This is the data volume of UE#m, and W m t This is the local model of UE#m.

[0048] To prevent the same UE's local model from being repeatedly aggregated, the global model can be expressed as shown in equation (7).

[0049]

number

[0050] Here, N can represent the total amount of data from UE#1 to UE#K, and can be expressed by equation (8), n k This is the amount of data in UE#k, and U k ={0,1} and U k =1 indicates that the UE#k model upload was successful, U k =0 indicates that the model upload for UE#k failed, and similarly, U m / U M ={0,1} and U m / U M =1 indicates that the upload of the UE#m / UE#M model was successful. m / U M =0 indicates that the upload of the UE#m / UE#M model failed.

[0051]

number

[0052] As reflected in equations (7) to (8), if the upload of the aggregated model of UE#M is successful, the control device does not consider the local models uploaded individually by UE#1 to UE#M-1, and only if the upload of the aggregated model of UE#M fails does the control device consider the local models uploaded individually by UE#1 to UE#M-1. Therefore, according to the first embodiment of this disclosure, as long as the upload of at least one model in UE#m (m=1,···,M-1) and UE#M is successful, the model of UE#m can contribute to the global model without bias.

[0053] Specifically, if UE#m accesses its local model W via the uplink, m t The probability of failure to send is q m Therefore, the probability of success is p m = 1-q m Therefore, UE#M aggregates via the uplink to the model W M t+1 / 2 The probability of failure to send is q M Therefore, the probability of success is p M = 1-q M Let's assume that, for one iteration of associative learning, if we adopt a conventional associative learning solution (i.e., associative learning without side links), then the local model W of UE#m m t The probability that this ultimately contributes to the success of the global model is 1-q m In contrast, when adopting the federated learning solution of this disclosure, the local model W of UE#m m t The probability that this will ultimately contribute to the success of the global model is 1-q m q M Therefore, the model W of UE#m m t The probability that this will ultimately contribute to the success of the global model is qm (1-q M ) To improve.

[0054] According to this disclosure, the control device needs to aggregate the received models (which may include, for example, aggregated models and local models in the first embodiment) so that the same UE's local models are not aggregated repeatedly. Therefore, the control device needs to be able to determine which local models corresponding to which non-centralized UEs the aggregated model transmitted by the central UE has aggregated. In other words, the control device needs to know which local models need to be aggregated with the received aggregated model in order to generate a global model. According to this disclosure, a first implementation is provided in which the UE itself determines which local models to aggregate and reports to the control device information indicating which local models have been aggregated along with the aggregated model, and a second implementation is provided in which the control device determines which local models the UE needs to aggregate and sends an aggregation instruction to the UE.

[0055] Figures 4A to 4B schematically illustrate the information exchange in the first implementation form of the first embodiment of the present disclosure.

[0056] As shown in Figures 4A-4B, in the first implementation of the first embodiment, the central UE (e.g., UE#M in Figures 4A-4B) may further upload information identifying each UE corresponding to the aggregated local model, along with the aggregated model. For example, the information identifying each UE corresponding to the aggregated local model may be a UE identifier. For example, the central UE may transmit the UE identifier at the same time as transmitting the aggregated model, and may encode both the aggregated model and the UE identifier in the same message. Alternatively, the central UE may transmit the aggregated model and the UE identifier in separate messages, respectively. Note that the central UE may not have received any local models. In this case, the central UE may upload its local models to the control device along with its identifier.

[0057] The central UE may aggregate all received local models to generate an aggregated model and upload the identifier of each UE corresponding to the aggregated local model to the control device. For example, as shown in Figure 4A, if all local models in a UE group are successfully sent to UE#M, UE#M aggregates the local models of all UEs in that UE group (including UE#M's own local model) and uploads the aggregated model W to the control device. M t+1 / 2 The aggregated UE IDs, i.e., {1,2,···,M}, may also be uploaded. This allows the control device to understand that the local models of UE#1 to UE#M have been aggregated into the aggregated model. In response to the received information, the control device may discard the local models individually uploaded by UE#1 to UE#M-1 and aggregate the received aggregated model with the local models of other UEs that are not divided into UE groups in the wireless communication system. This aggregates the local models of each UE participating in federated learning to generate a global model, thus avoiding repetition.

[0058] If not all local models within the UE group are successfully transmitted to UE#M, for example, if the transmission of the local model from UE#1 to UE#M fails as shown in Figure 4B, UE#M aggregates all the locally models of the UEs that were successfully received, as well as its own local model, and sends the aggregated model W to the control device. M t+1 / 2 The aggregated UE ID, i.e., {2,···,M}, may also be uploaded. This allows the control device to understand that the local models of UE#2 to UE#M are aggregated in the aggregated model, but the model of UE#1 is not. In response to the received information, the control device discards the local models individually uploaded by UE#2 to UE#M-1 and aggregates the received local model of UE#1, the aggregated model uploaded by UE#M, and the local models of other UEs that are not divided into UE groups in the wireless communication system, thereby creating a global model W t+1 You may generate this.

[0059] Specifically, if the transmission of the local model from UE#1 to UE#M fails, the aggregated model uploaded by UE#M may be transformed from equation (6) to equation (6-1) above.

[0060]

number

[0061] The global model generated by the control device may be transformed from equation (7) to equation (7-1) above.

[0062]

number

[0063] The total data size N for UE#1 to UE#M can also be expressed by transforming equation (8) into equation (8-1).

[0064]

number

[0065] Figures 5A to 5B schematically illustrate the information exchange in a second implementation form of the first embodiment according to this disclosure.

[0066] As shown in Figures 5A-5B, in the second implementation of the first embodiment, each non-centralized UE in the UE group may transmit information to the control device indicating the success or failure of the transmission of the local model to the central UE. Each non-centralized UE may transmit the information in any appropriate manner. This allows the control device to determine whether the transmission of the local model from the non-centralized UE to the central UE (also called a sidelink transmission of the local model) was successful. For example, the non-centralized UE may provide feedback to the control device only if the transmission of the local model to the central UE has failed. Furthermore, if the control device has not received a feedback message regarding the failure of the sidelink transmission of the local model from the non-centralized UE to the central UE, it defaults to considering the transmission of the local model from the non-centralized UE to the central UE to have been successful. Alternatively, for example, the non-centralized UE may provide feedback to the control device indicating whether the sidelink transmission of the local model was successful each time the local model is transmitted to the central UE, regardless of success or failure. Preferably, any method that guarantees communication reliability (e.g., a retransmission mechanism, a repeat transmission mechanism, etc.) may be employed to ensure that the control device receives information indicating whether the sidelink transmission of the local model was successful. Preferably, the decentralized UE may use a separate message from the message uploading the local model to the control device, as shown in Figures 5A-5B, to individually feed back to the control device whether the sidelink transmission of the local model was successful. Selectively, the decentralized UE may also send information indicating whether the sidelink transmission of the local model was successful along with the local model (for example, if the decentralized UE has monitored that the uplink communication quality is good).

[0067] In response to messages received from at least each non-centralized UE indicating whether the sidelink transmission of a local model was successful, the control device may determine which local models the central UE needs to aggregate and send an aggregation instruction to the central UE instructing it whether to aggregate its local models with one or more local models from other non-centralized UEs, and, if necessary, which one or more other non-centralized local models to aggregate with its local models. For example, the control device may simply respond to messages received from each non-centralized UE indicating whether the sidelink transmission of a local model was successful and instruct the central UE to aggregate all successfully received local models and its own model. Alternatively, for example, the control device may, based on a combination of the successfully received local models from each non-centralized UE and the messages indicating whether the sidelink transmission of the local models received from each non-centralized UE was successful, instruct the central UE to aggregate only the local models that the control device has not received but which have successfully transmitted sidelinks.

[0068] For example, as shown in Figure 5A, if all local models within a UE group have been successfully transmitted to UE#M, the control device may send an aggregation instruction to UE#M instructing UE#M to aggregate the local models of UE#1 to UE#M. In response to the aggregation instruction, UE#M generates an aggregated model W, which is the aggregated local model of UE#1 to UE#M. M t+1 / 2 A model may be generated and uploaded to the control device. The control device may discard the local models uploaded individually by UE#1 to UE#M-1 and aggregate the received aggregated model with the local models of other UEs that are not divided into UE groups in the wireless communication system. In this way, the local models of each UE participating in federated learning are aggregated to form a global model W, so as not to be repeated. t+1 Generates.

[0069] Furthermore, for example, if all local models within a UE group are successfully sent to UE#M, and the control device successfully receives local models from UE#1 to UE#M-1, the control device may send an aggregation instruction to UE#M, instructing UE#M not to aggregate the local models but to upload its own local models individually. The control device then aggregates the local models corresponding to each received UE to create a global model W t+1 You may generate this.

[0070] If none of the local models within the UE group have been successfully transmitted to UE#M, for example, if the transmission of the local model from UE#1 to UE#M fails, as shown in Figure 5B, the control device may send an aggregation instruction to UE#M instructing UE#M to aggregate the local models of UE#2 to UE#M. In response to the aggregation instruction, UE#M generates an aggregated model W, which is the aggregated local model of UE#2 to UE#M. M t+1 / 2 The control device may generate and upload a global model W by discarding the local models individually uploaded by UE#2 to UE#M-1 and aggregating the received local model of UE#1, the aggregated model uploaded by UE#M, and the local models of other UEs that are not divided into UE groups in the wireless communication system. t+1 You may generate this.

[0071] Furthermore, for example, if the transmission of a local model from UE#1 to UE#M fails, and the control device successfully receives local models from, for example, UE#1 and UE#2, the control device may send an aggregation instruction to UE#M, instructing UE#M to aggregate the local models from UE#3 to UE#M. In response to the aggregation instruction, UE#M will send an aggregated model W, which is the aggregated local model from UE#3 to UE#M. M t+1 / 2The control device may generate and upload the following: The control device then aggregates the received local model of UE#1, the local model of UE#2, the aggregated model uploaded by UE#M, and the local models of other UEs that are not divided into UE groups in the wireless communication system to form the global model W. t+1 You may generate this.

[0072] A first embodiment of the present disclosure has been described in detail with reference to the drawings. For the sake of explanation, the above describes the case in which there is one centralized UE group in the wireless communication system. It should be understood that any number of centralized UE groups may exist depending on the actual requirements.

[0073] (Second example) According to a second embodiment of the present disclosure, at least two UEs in a wireless communication system may constitute a UE group used for federated learning. The UE group may include at least two UEs communicating with each other via sidelinks. In other words, any two UEs within the UE group may send and receive information via sidelinks (i.e., sidelink communication between any two UEs may be bidirectional). The UE group of the second embodiment of the present disclosure is a decentralized UE group. In the UE group, each UE may receive at least one local model from at least one other UE in the UE group via sidelink communication, each of which is obtained by one of the other UEs training on its local data. Each UE in the UE group may, after receiving at least one local model from at least one other UE in the group, generate an aggregated model by aggregating its own local model, which it has trained on its own local data, with at least one of the received local models, and upload the aggregated model to the control equipment of the wireless communication system (e.g., a network element device or base station that implements the federated learning function in the core network described above). Alternatively, each UE in the UE group may transmit its own local model to other UEs in the UE group via sidelink communication (e.g., broadcast) for the generation of an aggregated model by the other UEs.

[0074] Figure 6 is a schematic diagram showing a second embodiment. As shown in Figure 6, suppose the wireless communication system includes K UEs. Here, UE#1 to UE#M may constitute a single decentralized UE group. For example, in this UE group, each UE may transmit its local model to other UEs in the UE group via sidelink communication (shown by dashed lines in Figure 6) (e.g., broadcast), and receive local models from each of the other UEs in the UE group. For example, as shown in Figure 6, UE#1 aggregates the local model obtained by training based on UE#1's local data and each local model received from the other UEs in the UE group to form an aggregated model W1 t+1 / 2 The aggregated model may be obtained and uploaded to the control device. UE#2 to UE#M use a similar method to obtain their respective aggregated models W2 t+1 / 2 ~W M t+1 / 2 The aggregated model may then be uploaded to the control device.

[0075] In particular, sidelink transmission of local models may fail. In other words, each UE within the same UE group may not necessarily be able to receive local models from all other UEs. In this specification, a failure to transmit a local model via a sidelink to a UE and / or a failure to receive a local model via a sidelink to a UE within a UE group is abbreviated as "a failure to transmit a local model via a sidelink for a particular UE." For example, if the transmission of a local model from UE#1 to UE#2 fails, a failure to transmit a local model via a sidelink for both UE#1 and UE#2 is considered to exist, regardless of whether the transmission of a local model from UE#2 to UE#1 was successful or not.

[0076] If there are sidelink transmission failures for one or more UEs, these one or more UEs may be excluded from the original UE group, and the other UEs in the original UE group may aggregate only the local models of the remaining UEs. In other words, for each UE for which there are no sidelink transmission failures for its local model, that UE aggregates the local model it has trained on its local data with the one or more local models it has received that correspond to one or more other UEs for which there are no sidelink transmission failures for its local model.

[0077] An excluded UE may upload its local model to the control device individually. Alternatively, the excluded UEs may reconfigure one or more centralized and / or decentralized UE groups (i.e., UE groups according to the first embodiment and / or UE groups according to the second embodiment). For example, when reconfiguring a decentralized UE group, each UE may aggregate its own local model with the local models of one or more other excluded UEs to generate a new aggregated model and upload this new aggregated model to the control device. In other words, when reconfiguring a decentralized UE group, each UE may aggregate its own local model with the local models of one or more other UEs that have received and experienced failures in sidelink transmissions of their local models to generate a new aggregated model and upload this new aggregated model to the control device. For example, when reconfiguring a centralized group, the new central UE may generate a new aggregated model similar to that of a UE in a decentralized UE group and upload this new aggregated model to the control device. Furthermore, for example, when reconfiguring a centralized UE group, the decentralized UEs in the new UE group may send their local models to the new centralized UE to generate a new aggregated model, which can then be uploaded to the control device.

[0078] To make it clear, the purpose of excluding a UE from a UE group is, for example, to simplify the operation of a control device when generating a global model. Because a control device needs to aggregate one or more received local models and / or one or more aggregated models so as not to repeatedly aggregate the local models of the same UE, if the aggregated model uploaded by each UE in a UE group aggregates multiple different local models corresponding to multiple different UEs, the control device requires relatively complex statistical and computational operations to generate a proper global model. However, this disclosure is not limited to the implementation form of excluding a UE from a UE group if the side-link transmission of a local model fails. For example, as an alternative form, each UE may aggregate all received local models and, for each UE in which the transmission of its local model to any UE fails, that UE may separately upload its local model to the control device.

[0079] According to this disclosure, multiple UEs with poor communication quality (e.g., particularly uplink communication quality) may form a decentralized UE group. For example, multiple UEs with particularly poor uplink communication quality may form a decentralized UE group. Alternatively, one or more UEs with poor communication quality (e.g., particularly uplink communication quality) may form a decentralized UE group with one or more neighboring UEs with good communication quality (e.g., particularly uplink communication quality). A control device may divide the decentralized UE group based on multiple pieces of information about each UE. The division process and the information on which the division is based will be described in detail later. Similar to the first embodiment, one or more UEs with good communication quality (e.g., particularly uplink communication quality) may not participate in any UE group and may independently upload their local model as in conventional federated learning.

[0080] The basic concept of the second embodiment of this disclosure has been explained in conjunction with Figure 6, and the global model aggregation process according to the second embodiment of this disclosure will now be explained in detail.

[0081] Suppose a wireless communication system contains K UEs. First, of these K UEs, UE#1 and UE#2 form a decentralized UE group, and UE#1 is trained based on its local data to create a local model W1. t After obtaining the local model W1 t The data is sent to UE#2, and UE#2 then trains the local model W2 based on that local data. t After obtaining the local model W2 t Let's simply assume that this is sent to UE#1. Then, UE#1 and UE#2 each perform model aggregation. The local models that UE#1 and UE#2 aggregate are the same two groups of models (i.e., each group of models is the same as UE#1's local model W1). t and UE#2 local model W2 t (including) Aggregation model W1 uploaded by UE#1 t+1 / 2 and aggregate model W2 uploaded by UE#2 t+1 / 2 This can be expressed as equation (9).

[0082]

number

[0083] Here, n1 and n2 are the data sizes for UE#1 and UE#2, respectively.

[0084] When a control device performs aggregation based on the received model, it may aggregate the received local model and the aggregated model so that the local model of the same UE is not aggregated repeatedly. This avoids the contribution of each UE's local model to the global model being different. Therefore, the global model can be expressed as equation (10).

[0085]

number

[0086] Here, N can represent the total data volume of UE#1 to UE#K, which is expressed by Equation (11), and n k is the data volume of UE#k, and U k ={0, 1}, and U k =1 indicates that the upload of the model of UE#k has been successful, and U k =0 indicates that the upload of the model of UE#k has failed. Similarly, U1={0, 1}, U1=1 indicates that the upload of the model of UE#1 has been successful, and U1=0 indicates that the upload of the model of UE#1 has failed. U2={0, 1}, U2=1 indicates that the upload of the model of UE#2 has been successful, and U2=0 indicates that the upload of the model of UE#2 has failed.

[0087]

Number

[0088] As reflected in Equations (10) to (11), when generating the global model, the control device does not repeatedly aggregate the aggregated models uploaded by UE#1 and UE#2. That is, when both UE#1 and UE#2 have successfully uploaded the aggregated model, the control device discards one of the aggregated models and only uses one aggregated model representing the local models of UE#1 and UE#2 respectively for generating the global model. Also, as reflected in Equations (10) to (11), as long as at least one of the model uploads in UE#1 and UE#2 is successful, the models of UE#1 and UE#2 can contribute to the global model.

[0089] Specifically, assume that the failure probability of UE#1 transmitting the model via the uplink is q1, and the success probability is p1 = 1 - q1. The failure probability of UE#2 transmitting the model via the uplink is q2, and the success probability is p2 = 1 - q2. For one iteration of federated learning, when adopting the conventional federated learning solution (i.e., federated learning without sidelink), the local model W1 of UE#1 tThe probability that it finally contributes to the global model successfully is 1 - q1. In contrast, when adopting the federated learning solution of the second embodiment of the present disclosure, the local model W1 of UE#1 t The probability that it finally contributes to the global model successfully is 1 - q1q2. Therefore, the local model W1 of UE#1 t The probability that it finally contributes to the global model successfully is improved by q1(1 - q2). Similarly, the probability that the model W2 of UE#2 t The probability that it finally contributes to the global model successfully is improved by q2(1 - q1). Such improvement may become more prominent through multiple iterations of federated learning. Also, since the aggregated model generated by aggregation and the local model have the same model size, the solution of the second embodiment does not cause a significant increase in the amount of data transmitted via the uplink in the wireless communication network.

[0090] More generally, if a wireless communication system includes K UEs, and among these K UEs, M UEs, that is, UE#1 to UE#M, constitute one decentralized UE group, and each UE transmits its local model to other UEs and aggregates its own local model with the local models from other UEs. Assume that the local models aggregated by UE#1 to UE#M are models of the same M groups (that is, the models of each group all include the local model W1 of UE#1 t ~ the local model W of UE#M M t ), the aggregated model uploaded by each UE can be expressed as Equation (12).

[0091]

Equation

[0092] Here, n m is the data volume of UE#m, and W m t is the local model of UE#m.

[0093] To prevent the same UE's local model from being repeatedly aggregated, the global model can be expressed as shown in equation (13).

[0094]

number

[0095] Here, N can represent the total amount of data from UE#1 to UE#K, and is expressed by equation (14), n k This is the amount of data in UE#k, and U k ={0,1} and U k =1 indicates that the UE#k model upload was successful, U k =0 indicates that the model upload for UE#k failed, and similarly, U m ={0,1} and U m =1 indicates that the UE#m model upload was successful, U m =0 indicates that the UE#m model upload failed.

[0096]

number

[0097] As reflected in equations (13) to (14), when generating the global model, the control device does not repeatedly aggregate the aggregate models uploaded by UE#1 to UE#M. That is, if UE#1 to UE#M all successfully upload their aggregate models, the control device uses a single aggregate model, which is an aggregate of each of UE#1 to UE#M's local models, to generate the global model. Also, as reflected in equations (13) to (14), as long as at least one model upload in UE#1 to UE#M is successful, the models of UE#1 to UE#M can contribute to the global model.

[0098] Specifically, if the probability of UE#m failing to send the model via the uplink is q m Therefore, the probability of success is pm = 1-q m Let's assume that, for one iteration of associative learning, if we adopt a conventional associative learning solution (i.e., associative learning without side links), then the local model W of UE#m m t The probability that this will ultimately contribute to the success of the global model is 1-q m In contrast, when adopting the federated learning solution of the second embodiment of this disclosure, the local model W of UE#m m t The probability that this will ultimately contribute to the success of the global model is,

number

number

[0099] Similar to the first embodiment, the control device needs to be able to determine which local models corresponding to which UEs have been aggregated into the aggregated model transmitted by each UE. In other words, the control device needs to be able to determine which local models need to be aggregated with the received aggregated model in order to generate a global model. Similar to the first embodiment, a first implementation can be provided in which the UE itself determines which local models to aggregate and reports to the control device information indicating which local models have been aggregated along with the aggregated model, and a second implementation can be provided in which the control device determines which local models the UE needs to aggregate and sends an aggregation instruction to the UE.

[0100] Figures 7A to 7C schematically illustrate the information exchange of the first implementation form of the second embodiment of the present disclosure.

[0101] As shown in Figures 7A-7C, in the first implementation of the second embodiment, each UE within the UE group first trains based on its own local data and then sets its own local model (W1 t W2 t , , , W M t The local model is obtained. Then, each UE transmits its respective local model to the other UEs in the UE group. For example, as shown in Figures 7A-7C, each UE may broadcast its own local model within the UE group. In the first implementation of the second embodiment, each UE in the UE group may decide which local models to aggregate based on the sidelink transmission result (i.e., success or failure) of the local models of each UE in the group. Therefore, as shown in Figures 7A-7C, after transmitting a local model within the UE group, each UE may transmit to the other UEs in the UE group the transmission result of that local model, i.e., information indicating whether the transmission of that local model to each other UE in the UE group was successful or unsuccessful. For example, as shown in Figures 7A-7C, each UE may broadcast the sidelink transmission result of its local model within the UE group. In other words, each UE broadcasts to the UE group via sidelink communication information indicating whether the transmission of its local model to each other UE in the UE group was successful or unsuccessful. Simultaneously, each UE may receive information via sidelink communication from at least one other UE in the UE group indicating whether the transmission of the local model from that at least one other UE to each UE in the UE group was successful or unsuccessful.

[0102] Each UE may transmit information in any appropriate manner so that other UEs can determine whether the sidelink transmission of the UE's local model was successful. For example, each UE may broadcast only that the transmission of its local model to one or more other UEs failed (in which case, the transmission of the local model from that UE to the remaining UEs can be considered successful by default). Alternatively, each UE may broadcast the results of its sidelink transmission of its local model to each other UE. Preferably, any method that guarantees communication reliability (e.g., a retransmission mechanism, a repeat transmission mechanism, etc.) may be employed to transmit information indicating whether the sidelink transmission of the local model was successful.

[0103] In the first implementation of the second embodiment, after receiving local model transmission result information from other UEs, each UE in the UE group may determine for itself which received local models to aggregate.

[0104] For example, for each UE for which there are no sidelink transmission failures of the local model related to it (i.e., the UE successfully transmits the local model to all other UEs in the UE group and successfully receives the local model from each of the other UEs in the UE group), as described above, the UE may determine which one or more UEs are excluded from the UE group, i.e., which one or more UEs in the original UE group have sidelink transmission failures of the local model. For example, this determination may be based on the results of sidelink transmissions broadcast by other UEs. After deciding which local models to aggregate, the UE may generate an aggregated model accordingly and, along with the aggregated model, further upload information identifying each UE corresponding to each aggregated local model. For example, the information identifying each UE corresponding to each aggregated local model may be a UE identifier. For example, the UE may transmit the UE identifier at the same time as transmitting the aggregated model, and may encode both the aggregated model and the UE identifier in the same message, for example. Alternatively, the UE may transmit the aggregated model and the UE identifier in separate messages, respectively.

[0105] For example, as described above, for UEs where a sidelink transmission failure of a local model exists (i.e., transmission of a local model to any UE in a UE group fails, and local models to any other UEs in that UE group have not been successfully received), these UEs may individually upload their local models to the control device, or they may reconfigure one or more centralized and / or decentralized UE groups. For example, these UEs may decide how to act according to default rules, for example, by default uploading their local models to the control device individually, or by default negotiating themselves to reconfigure one or more centralized and / or decentralized UE groups. Alternatively, these UEs may decide how to act after waiting for further notification from the control device. For example, the control device may further notify these UEs whether to individually transmit their local models or to reconfigure the UE groups (and how to configure the UE groups). For example, the control device may make a decision based on the aggregation status of the received aggregated models, or it may instruct these UEs to send further information (e.g., information indicating which of these excluded UEs' local models were successfully transmitted) to assist in that decision.

[0106] Figure 7A shows the case where all sidelink transmissions of local models are successful. In this case, each UE aggregates the local models of all UEs within its UE group (including its own local model) and sends the aggregated model W1 to the control device. t+1 / 2 W2 t+1 / 2 , , , W M t+1 / 2The aggregated UE IDs, i.e., {1,2,···,M}, may also be uploaded. This allows the control device to understand that the local models of UE#1 to UE#M have been aggregated into each of these aggregated models. In response to the received information, the control device may aggregate the aggregated model uploaded by any one of UE#1 to UE#M with the local models of other UEs that are not divided into UE groups in the wireless communication system. This aggregates the local models of each UE participating in federated learning to form a global model W, so as not to be repetitive. t+1 Generates.

[0107] Figures 7B and 7C illustrate the case where the sidelink transmission from UE#2 to UE1 fails. In this case, as shown in Figures 7B and 7C, all other UEs except UE#1 and UE#2 decide to exclude UE#1 and UE#2 based on the broadcasted sidelink transmission results within the UE group, and accordingly aggregate the local models (including their own local models) of all other UEs in the UE group and send the aggregated model W3 to the control device. t+1 / 2 , , , W M t+1 / 2 The aggregated UE IDs, i.e., {3,···,M}, may also be uploaded. This allows the control device to understand that the local models of UE#3 to UE#M have been aggregated into each of these aggregated models. UE#1 and UE#2 may individually upload their local models to the control device and transmit their identifiers along with their local models, as shown in Figure 7B. Alternatively, UE#1 and UE#2 may reconfigure a centralized UE group with UE#2 as the central UE, as shown in Figure 7C. Specifically, as shown in Figure 7C, UE#2 aggregates its own local model with the local model of UE#1 to form aggregated model W2 t+1 / 2 The aggregated model and the IDs {1,2} of the aggregated UE may be generated and uploaded to the control device. UE#1 may individually upload its own local model and ID to the control device.

[0108] Specifically, if the transmission of the local model from UE#2 to UE#1 fails, the aggregated model uploaded by UE#1 and other UEs besides UE#2 may be transformed from equation (12) to equation (12-1).

[0109]

number

[0110] As shown in Figure 7B, if UE#1 and UE#2 each upload their local models individually, the global model generated by the control device may be transformed from equation (13) to equation (13-1).

[0111]

number

[0112] Furthermore, the total data size N of UE#1 to UE#K may be transformed from equation (14) to equation (14-1).

[0113]

number

[0114] As shown in Figure 7C, when UE#1 and UE#2 reconfigure a centralized UE group with UE#2 as the central UE, the global model generated by the control equipment may be transformed from equation (13) to equation (13-2).

[0115]

number

[0116] Furthermore, the total data size N of UE#1 to UE#K may be transformed from equation (14) to equation (14-2).

[0117]

number

[0118] Figures 8A to 8C schematically illustrate the information exchange of a second implementation form of the first embodiment according to this disclosure.

[0119] As shown in Figures 8A-8C, in the second implementation of the second embodiment, unlike the first implementation, each UE uploads information to the control device indicating the success or failure of its local model's transmission to other UEs in the UE group, rather than broadcasting the sidelink transmission results of its local model to each UE in the UE group. Each UE may transmit the information in any suitable manner. This allows the control device to know whether all sidelink transmissions of the UE's local model have been successful. For example, each UE may notify the control device only that the transmission of its local model to one or more other UEs has failed. Alternatively, each UE may notify the control device of the sidelink transmission results of its local model to each other UE. Preferably, any method that guarantees communication reliability (e.g., a retransmission mechanism, a repeat transmission mechanism, etc.) may be used to transmit information indicating whether the sidelink transmission of the local model has been successful.

[0120] In response to messages received from at least each UE indicating whether the sidelink transmission of the local model was successful, the control device may determine which local models each UE needs to aggregate, and send aggregation instructions to each UE instructing it whether it should aggregate its local model with one or more local models of other UEs, and, if so, which one or more other local models it should aggregate with its local model.

[0121] For example, as shown in Figure 8A, if all sidelink transmissions of the local models are successful, the control device may send aggregation instructions to each UE in the UE group, instructing UE#1 to UE#M to aggregate their local models. In response to the aggregation instructions, UE#1 to UE#M each create an aggregated model W1 in which their local models are aggregated. t+1 / 2 W2 t+1 / 2 , , , W M t+1 / 2 A global model W may be generated and uploaded to the control device. The control device may aggregate the aggregated model uploaded by any one of UE#1 to UE#M with the local models of other UEs that are not divided into UE groups in the wireless communication system. This aggregates the local models of each UE participating in federated learning to avoid repetition and create a global model W. t+1 Generates.

[0122] If not all local models within a UE group have been successfully sent to UE#M, the control unit may, based on the sidelink transmission results received from each UE, decide which UEs to remove from the original UE group and how the removed UEs should behave. For example, the control unit may decide that these removed UEs need to upload their local models individually. Alternatively, for example, the control unit may reconfigure one or more centralized or decentralized UE groups with these UEs. For example, the control unit may decide whether to reconfigure the groups and how to reconfigure them based on information such as the latest sidelink communication quality, uplink and downlink communication quality, and computing power of these UEs.

[0123] For example, if the transmission of the local model from UE#2 to UE#1 fails, as shown in Figure 8B, the control device may send aggregation instructions to UE#1 to UE#M respectively, instructing UE#1 and UE#2 to upload their respective local models individually, and instructing the other UEs to aggregate the local models of UE#3 to UE#M. In response to the aggregation instructions, UE#1 and UE#2 will each individually upload their local model W1 t and W2 t The local models of UE#3 to UE#M may be uploaded. Furthermore, other UEs may generate an aggregated model and upload it to the control device. The control device aggregates the received local model of UE#1, the local model of UE#2, the aggregated model uploaded by UE#3 to UE#M, and the local models of other UEs not divided into UE groups in the wireless communication system to form a global model W. t+1 You may generate this.

[0124] Furthermore, for example, if the transmission of a local model from UE#2 to UE#1 fails, as shown in Figure 8C, the control device may send aggregation instructions to UE#1 to UE#M respectively to form a centralized UE group, and accordingly instruct UE#1 and UE#2 to upload their models, and instruct the other UEs to aggregate the local models of UE#3 to UE#M. In response to the aggregation instruction, UE#1 may upload its local model individually, and UE#2 may aggregate its own local model with UE#1's local model to form the aggregated model W2 t+1 / 2 The UE may generate and upload to the control device, and other UEs may generate an aggregated model in which the local models of UE#3 to UE#M are aggregated and upload to the control device. The control device aggregates the received local model of UE#1, the aggregated model uploaded by UE#2, the aggregated models uploaded by UE#3 to UE#M, and the local models of other UEs that are not divided into UE groups in the wireless communication system to form a global model W t+1 You may generate this.

[0125] A second embodiment of the present disclosure has been described in detail with reference to the drawings. For the sake of explanation, the above describes the case in which there is one decentralized UE group in the wireless communication system. It should be understood that any number of centralized UE groups may exist depending on the actual requirements.

[0126] The above describes the first and second embodiments, respectively. According to this disclosure, the first and second embodiments may be implemented in combination, and one or more centralized UE groups and one or more decentralized UE groups may exist simultaneously. More generally, in a wireless system, one or more centralized UE groups, one or more decentralized UE groups, and one or more UEs that are not divided into any UE group may exist in any suitable combination.

[0127] As described in detail by linking the first and second embodiments, the federated learning network of the present disclosure increases the contribution of each UE's local model to the global model by transmitting local models using side links between UEs, in addition to conventional uplinks and downlinks. With the introduction of side links, the solution of the present disclosure may select UEs to participate in federated learning from the wireless communication network based on several additional factors in addition to conventional selection factors. These will be described in detail below.

[0128] According to this disclosure, the control equipment comprises at least, Information regarding the sidelink status of the UE, including information that can indicate at least whether a sidelink exists between the UE and one or more UEs in the wireless communication system, and the communication quality of the sidelink, UE location information and, Information indicating the communication quality of the UE's uplink and downlink, Information indicating UE's computing power, The UEs that participate in federative learning may be selected based on one or more of the following: information indicating the amount of user data to generate the local model for the UE.

[0129] For example, information regarding communication quality may be any information that can indicate communication quality, such as Reference Signal Receiving Power (RSRP), Reference Signal Receiving Quality (RSRQ), Reference Signal Strength Indicator (RSSI), or Signal to Noise Ratio (SNR).

[0130] For example, location information may be any applicable information indicating the UE's location, such as location information obtained by any positioning technology (e.g., Global Positioning System (GPS), Assisting-Global Navigation Satellite System (A-GNSS), motion sensor positioning, etc.), uplinked / downlink-Time Difference of Arrival (UL / DL-TDOA), multi-round trip time (Multi-RTT), and new radio enhanced cell ID (NR-Enhanced Cell ID (E-CID)).

[0131] For example, information indicating the computing power of the UE may include, for instance, CPU usage and parameters representing the performance of the processing itself.

[0132] The control device may obtain this information in any suitable manner. For example, one or more of this information may be directly uploaded to the control device by the UE. Alternatively, for example, if the control device is a network element responsible for federated learning functions in the core network, one or more of this information may be obtained from a base station associated with the control device. For example, the network element may request one or more of this information from the base station. Alternatively, for example, if the control device is a base station, one or more of this information (e.g., the UE's location information and the communication quality of the UE's uplink and downlink) may be calculated by the control device itself based on measurements. According to this disclosure, a control device can select UEs to participate in federated learning by integrating the above information, and advantageously, some of these UEs may not be selected to participate in federated learning based on conventional federated learning. For example, the control device may divide the UE group based on the above information to allow more UEs to participate in federated learning.

[0133] For example, for one or more UEs where basic conditions such as computing power and user data volume satisfy the requirements for federated learning participation, but the uplink and downlink communication quality does not satisfy the requirements for federated learning participation, (1) In the vicinity of one or more UEs, A UE that meets the requirements for participation in federated learning, such as computing power, user data volume, and uplink / downlink communication quality. A UE with which one or more UEs have sidelinks that meet the required communication quality, If there is a UE with sufficient computing power to support aggregate operations on the model, that UE can be formed into a centralized UE group and participate in federated learning. (2) If there are side links between the one or more UEs that meet the required quality of communication, and the computing power of the one or more UEs is sufficient to support aggregate operations on the model, the one or more UEs may form a decentralized UE group and participate in federated learning. (3) If there are sidelinks between the one or more UEs that meet the required communication quality, and there are sidelinks between the one or more UEs that meet the required communication quality between them and another one or more UEs that meet the basic requirements and uplink and downlink communication quality requirements for participation in federated learning, and the computing power of the one or more UEs and the other one or more UEs is sufficient to support aggregate operations on the model, then the one or more UEs may form a decentralized UE group with the other one or more UEs and participate in federated learning.

[0134] Furthermore, for example, if one or more UEs (e.g., those that may have important local data) particularly wish to participate in federated learning but whose uplink / downlink communication quality does not meet conventional federated learning requirements, the control equipment can ensure that this one or more UEs can participate in federated learning by dividing the UE group based on the above information.

[0135] for example, (1) In the vicinity of one or more of these UEs, Uplink and downlink communication quality meets the requirements for participation in federated learning. A UE with which one or more UEs have sidelinks that meet the required communication quality, If there is a UE whose computing power is sufficient to support aggregate operations on the model, even if the amount of user data for that UE is small (for example, if it does not meet the conventional criteria for participating in federated learning), that UE will be allowed to form a centralized UE group with one or more other UEs and participate in federated learning. (2) If there are side links between the one or more UEs that meet the required quality of communication, and the computing power of the one or more UEs is sufficient to support aggregate operations on the model, the one or more UEs may form a decentralized UE group and participate in federated learning. (3) If there are side links between the one or more UEs with communication quality that meet the requirements, and there are side links between the one or more UEs with communication quality that meet the requirements for participation in federated learning, and the computing power of both the one or more UEs and the other one or more UEs is sufficient to support aggregate operations on the model, then the one or more UEs can participate in federated learning by forming a decentralized UE group with the other one or more UEs, even if the amount of user data of the other one or more UEs is small (for example, not meeting the conventional criteria for participation in federated learning).

[0136] The above examples illustrate two ways in which UEs that do not meet the conventional federated learning conditions can participate in federated learning as described in this disclosure by dividing the UE group. It should be understood that even for each UE that meets the conventional federated learning conditions, the UE may be divided into a group according to the first or second embodiment of this disclosure based on one or more of the above information. For example, for one or more UEs whose uplink and downlink communication quality meets the requirements for participating in federated learning, but whose communication quality is low, (1) In the vicinity of one or more UEs, UE with high uplink and downlink communication quality, A UE with which one or more UEs have sidelinks that meet the required communication quality, If there is a UE with sufficient computing power to support aggregate operations on the model, that UE can be formed into a centralized UE group and participate in federated learning. (2) If there are side links between the one or more UEs that meet the required quality of communication, and the computing power of the one or more UEs is sufficient to support aggregate operations on the model, the one or more UEs may form a decentralized UE group and participate in federated learning. (3) If there are sidelinks between the one or more UEs that meet the required communication quality, and there are sidelinks between the one or more UEs that meet the required communication quality and one or more UEs with high uplink and downlink communication quality, and the computing power of both the one or more UEs and the other one or more UEs is sufficient to support aggregate operations on the model, then the one or more UEs can form a decentralized UE group with the other one or more UEs and participate in federated learning.

[0137] In particular, according to this disclosure, a single UE is not divided into different UE groups used for federative learning; that is, a single UE does not simultaneously participate in multiple UE groups used for federative learning. Furthermore, according to this disclosure, the division of UE groups may be adjusted dynamically.

[0138] Figure 9 schematically illustrates exemplary information exchange for selecting UEs to participate in federated learning according to an embodiment of the present disclosure and for dividing the group of UEs to be used for federated learning. As shown in Figure 9, each UE in the wireless communication network may upload information about itself to a control device. As described in detail above, this information may include at least one or more of the following: information about the UE's sidelink status, the UE's location, information indicating the communication quality of the UE's uplinks and downlinks, information indicating the UE's computing power, and information indicating the amount of user data for generating the UE's local model. For example, this information may be transmitted directly to the control device (e.g., if the control device is a base station) or received by the control device via transmission to other devices (if the control device is a network element of the core network, via the base station and, where appropriate, other network elements in the core network). Although not explicitly shown in Figure 9, the control device may also decide on some of this information itself.

[0139] Next, the control device may select UEs to participate in federated learning and divide them into UE groups based on one or more of these pieces of information. The control device may then notify each UE of instructions to participate in federated learning and / or information regarding the division of the federated learning groups. For example, this information may include information indicating whether a UE participates in federated learning individually or as part of a UE group, information indicating whether the UE group in which the UE participates is a centralized UE group or a decentralized UE group, and information indicating the role of the UE in the UE group (e.g., whether it is a central UE or not). For example, this information may be transmitted directly to each UE (e.g., if the control device is a base station) or received by each UE via transmission by other devices (if the control device is a network element of the core network, via the base station and, where appropriate, other network elements in the core network). The control device may notify this information in any suitable manner.

[0140] The specific solutions of this disclosure are described in detail with reference to the drawings. Advantageously, this disclosure can improve the overall efficiency of federated learning by increasing the contribution rate of each UE's (especially UEs with poor uplink communication quality) local model to the global model without increasing the amount of uplink data on the UE, and by avoiding the global model becoming biased towards UEs with good communication quality. Furthermore, this disclosure can appropriately relax the criteria for participating in federated learning, allowing more UEs to participate in federated learning, or ensuring that desired UEs can participate in federated learning.

[0141] The following describes the conceptual layout and conceptual operation flow of the control and terminal equipment to which the solutions of this disclosure apply, with reference to the drawings.

[0142] Figure 10 schematically shows the conceptual arrangement of electronic equipment on the federated learning control device side according to an embodiment of the present disclosure.

[0143] As shown in Figure 10, the electronic device 10 may include a processing circuit 102. The processing circuit 102 may be configured to receive a plurality of models associated with a plurality of terminal devices in a wireless communication system, wherein the plurality of models include at least one aggregated model. As described in detail above, each of the at least one aggregated model may be generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by training on the local data of a corresponding terminal device. In particular, as described with reference to the first embodiment, the plurality of models may further include one or more local models, wherein at least one of the one or more local models is the same as at least one of the local models aggregated into the at least one aggregated model.

[0144] The processing circuit 102 may be in the form of a general-purpose processor or a dedicated processing circuit such as an ASIC. For example, the processing circuit 102 may be composed of a circuit (hardware) or a central processing unit (e.g., a central processing unit (CPU)). The processing circuit 102 may also hold a program (software) for operating the circuit (hardware) or the central processing unit. This program may be stored in memory (e.g., located in memory 104), or on an externally connected storage medium, or it may be downloaded via a network (e.g., the Internet).

[0145] In one selective implementation, the processing circuit 102 may include a terminal device selection unit. For example, the control information determination unit may determine, for example, which terminal devices can participate in federated learning based on information about the terminal devices. Alternatively, for example, the control information determination unit may decide, based on information about the terminal devices, to divide at least some of the terminal devices participating in federated learning into one or more terminal device groups according to the first embodiment and / or one or more terminal device groups according to the second embodiment.

[0146] In one selective implementation, the processing circuit 102 may further include an aggregation instruction generation unit. This aggregation instruction generation unit may determine an aggregation instruction, for example, in response to local model sidelink transmission result information uploaded by the terminal device. As described above, the aggregation instruction may instruct the terminal device whether its local model should be aggregated with the local models of one or more other terminal devices, and, if aggregation is necessary, with which one or more local models of other terminal devices.

[0147] Selectively, the electronic device 10 may further include a memory 104, shown by a dotted line in the figure, and a communication unit 106. The electronic device 10 may also further include other components not shown, such as a network interface, a processor, a controller, a radio frequency link, and a baseband processing unit. The processing circuit 102 may be associated with the memory 104 and / or the communication unit 106. For example, to access data, the processing circuit 102 may be directly or indirectly connected to the memory 104 (for example, with other components connected in between). Alternatively, for example, the processing circuit 102 may be directly or indirectly connected to the communication unit 106. This allows for the transmission and reception of signals via the communication unit 106.

[0148] Memory 104 can store various information determined and / or generated by the processing circuit 102 (e.g., local / aggregated models uploaded by terminal devices, sidelink transmission results of local models of terminal devices, information for selecting and / or grouping terminal devices to participate in federated learning), programs and data used for the operation of the electronic device 10, and data transmitted by the communication unit 106. Memory 104 can be located inside the processing circuit 102 or outside the electronic device 10, and is therefore depicted with a dashed line. Memory 104 may be volatile memory and / or non-volatile memory. For example, memory 104 may include, but is not limited to, random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), read-only memory (ROM), and flash memory.

[0149] The communication unit 106 may be configured to communicate with terminal equipment under the control of the processing circuit 102. In one example, the communication unit 106 may be implemented as a transmitter or transceiver including communication components such as an antenna array and / or a radio frequency link. In one example, the communication unit 106 may be implemented as a communication component that communicates via a cable.

[0150] Figure 10 shows that the processing circuit 102 is separated from the communication unit 106, but the processing circuit 102 may be implemented to include the communication unit 106. For example, it may be implemented in combination with the communication control unit. The processing circuit 102 may also be implemented to include one or more other components in the electronic device 10, or the processing circuit 102 may be implemented as the electronic device 10 itself. In actual implementation, the processing circuit 102 may be implemented as a chip (e.g., an integrated circuit module including a single chip), a hardware component, or a complete product.

[0151] It should be noted that each of the above means is merely a logic module categorized based on the specific function it realizes, and does not restrict the specific implementation form; for example, it may be implemented in software, hardware, or a combination of software and hardware. When actually implemented, each of the above means may be realized as an independent physical entity, or by a single entity (e.g., a processor (CPU or DSP, etc.), an integrated circuit, etc.). Furthermore, the dashed lines in the diagram indicate that each of the above means does not necessarily have to exist in reality, and the operation / function realized by them may be realized by the processing circuit itself. The processing circuit may also include other means for realizing the various functions of the control device described above.

[0152] Figure 11 schematically shows a conceptual operation flow 100 of the electronic device 10 on the control device side according to an embodiment of the present disclosure.

[0153] The operation of the electronic device 10 begins at S1002.

[0154] In S1004, the electronic device 10 receives a plurality of models associated with a plurality of terminal devices in a wireless communication system, and the plurality of models include at least one aggregated model. As described in detail above, each of the at least one aggregated model may be generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by training on the local data of a corresponding terminal device.

[0155] The operation of the electronic device 10 ends at S1006.

[0156] Note that the operation steps of the electronic device 10 on the control device side shown in Figure 11 are merely illustrative. The operation of the electronic device 10 may further include additional or alternative steps described in detail above. For example, the electronic device 10 may receive information from a terminal device indicating the sidelink transmission result of the terminal device's local model, and generate and transmit an aggregate instruction based at least on such information. Alternatively, for example, before S1004, the electronic device 10 may receive information from terminal devices in the wireless communication system to determine which terminal devices will participate in joint learning, divide at least some of these terminal devices into one or more terminal device groups according to the first embodiment and / or one or more terminal device groups according to the second embodiment, and notify each terminal device accordingly. The electronic device may also aggregate the received models to generate a global model.

[0157] Figure 12 schematically shows the conceptual arrangement of the electronic device 20 on the terminal device side according to an embodiment of the present disclosure.

[0158] As shown in Figure 12, the electronic device 20 may include a processing circuit 202. The processing circuit 202 may be configured to perform the operation of a decentralized terminal device according to the first embodiment, the operation of a centralized terminal device according to the first embodiment, and the operation of a terminal device according to the second embodiment, as described in detail above.

[0159] The processing circuit 202 may be in the form of a general-purpose processor or a dedicated processing circuit such as an ASIC. For example, the processing circuit 202 may be composed of a circuit (hardware) or a central processing unit (e.g., a central processing unit (CPU)). The processing circuit 202 may also hold a program (software) for operating the circuit (hardware) or the central processing unit. This program may be stored in memory (e.g., located in memory 204), or on an externally connected storage medium, or it may be downloaded via a network (e.g., the Internet).

[0160] In one implementation, the processing circuit 202 may include a local model generation unit. This local model generation unit may generate a local model of the electronic device 20 by training based on local data.

[0161] In one implementation, the processing circuit 202 may further include an aggregated model generation unit. This aggregated model generation unit may, for example, generate an aggregated model by aggregating the local model generated by the local model generation unit and the local model of other terminal devices received via the sidelink.

[0162] Selectively, the electronic device 20 may further include a memory 204, shown by a dotted line in the figure, and a communication unit 206. The electronic device 20 may also include other components not shown, such as a radio frequency link, a baseband processing unit, a network interface, a processor, and a controller. The processing circuit 202 may be associated with the memory 204 and / or the communication unit 206. For example, to access data, the processing circuit 202 may be directly or indirectly connected to the memory 204 (for example, with other components connected in between). Alternatively, for example, the processing circuit 202 may be directly or indirectly connected to the communication unit 206. This allows for the transmission and reception of radio signals via the communication unit 206.

[0163] Memory 204 can store various information determined and / or generated by the processing circuit 202 (e.g., local data, local models generated by the electronic device 20 itself, local models of other terminal devices, global models transmitted by control devices, etc.), programs and data used for the operation of the electronic device 20, data transmitted by the communication unit 206, etc. Memory 204 can be located inside the processing circuit 202 or outside the electronic device 20, and is therefore depicted with a dashed line. Memory 204 may be volatile memory and / or non-volatile memory. For example, memory 204 may include, but is not limited to, random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), read-only memory (ROM), and flash memory.

[0164] The communication unit 206 may be configured to communicate with terminal equipment under the control of the processing circuit 202. In one example, the communication unit 206 may be implemented as a transmitter or transceiver including communication components such as an antenna array and / or a radio frequency link.

[0165] Figure 12 shows that the processing circuit 202 is separated from the communication unit 206, but the processing circuit 202 may be implemented to include the communication unit 206. For example, it may be implemented in combination with the communication control unit. The processing circuit 202 may also be implemented to include one or more other components in the electronic device 10, or the processing circuit 202 may be implemented as the electronic device 10 itself. In actual implementation, the processing circuit 102 may be implemented as a chip (e.g., an integrated circuit module including a single chip), a hardware component, or a complete product.

[0166] It should be noted that each of the above means is merely a logic module categorized based on the specific function it realizes, and does not restrict the specific implementation form; for example, it may be implemented in software, hardware, or a combination of software and hardware. When actually implemented, each of the above means may be realized as an independent physical entity, or by a single entity (e.g., a processor (CPU or DSP, etc.), an integrated circuit, etc.). Furthermore, the dashed lines in the diagram indicate that each of the above means does not necessarily have to exist in reality, and the operation / function realized by them may be realized by the processing circuit itself. The processing circuit may also include other means for realizing the various functions of the control device described above.

[0167] Figure 13A schematically shows a conceptual operation flow 200 when the electronic device 20 according to this disclosure functions as a decentralized terminal device (hereinafter referred to as the first terminal device) in the first embodiment.

[0168] The operation of the first terminal device begins at S2002.

[0169] In S2004, the first terminal device uploads the first local model, which was obtained by training based on local data, to the control device of the wireless communication system. In S2006, the first terminal device transmits the first local model via sidelink communication to the second terminal device of the wireless communication system (i.e., the central terminal device in the terminal device group to which the first terminal device belongs, as described above) for the second terminal device to generate an aggregated model.

[0170] The operation of the first terminal device ends at S2008.

[0171] Note that the operation steps of the first terminal device on the control device side shown in Figure 13A are merely illustrative. The operation of the first terminal device may further include additional or alternative steps described in detail above. For example, the first terminal device may upload and transmit information to the control device indicating the success or failure of the transmission of the first local model to the second terminal device. Also, the execution order of the operations shown in Figure 13A is merely illustrative, and for example, the operations in S2004 and S2006 may be executed in reverse order or synchronously.

[0172] Figure 13B schematically shows a conceptual operation flow 210 when the electronic device 20 according to this disclosure functions as a central terminal device (hereinafter referred to as the second terminal device) in the first embodiment. The operation of the second terminal device begins at S2102.

[0173] In S2104, the second terminal device receives, via sidelink communication, at least one first local model obtained by training on its local data by one of the first terminal devices of the wireless communication system, each of which is trained by one of the first terminal devices.

[0174] In S2106, the second terminal device generates an aggregated model by aggregating the second local model obtained by training based on local data with the at least one first local model, and uploads the aggregated model to the control device of the wireless communication system.

[0175] The operation of the second terminal device ends at S2108.

[0176] Note that the operation steps of the second terminal device on the control device side shown in Figure 13B are merely illustrative. The operation of the second terminal device may further include additional or alternative steps described in detail above. For example, the second terminal device may upload information that identifies each terminal device corresponding to the aggregated local model, along with the aggregated model. For example, the second terminal device may receive aggregated instructions from the control device.

[0177] Figure 13C schematically shows a conceptual operation flow 220 when the electronic device 20 according to this disclosure functions as a terminal device (hereinafter referred to as the third terminal device) in the second embodiment. The operation of the third terminal device begins with S2202.

[0178] In S2204, the third terminal device may receive, via sidelink communication, at least one local model obtained by training each of the at least one other terminal devices in the terminal device group based on its local data.

[0179] In S2206, the third terminal device may generate an aggregated model by aggregating the third local model obtained by training based on local data and at least one of the received local models, and upload the aggregated model to the control device.

[0180] In S2208, the third terminal device may transmit the third local model to each other terminal device in the terminal device group via sidelink communication, allowing each other terminal device to generate an aggregated model.

[0181] The operation of the third terminal device ends at S2210.

[0182] Note that the operation steps of the third terminal device on the control device side shown in Figure 13C are merely illustrative. The operation of the third terminal device may further include additional or alternative steps described in detail above. For example, the third terminal device may transmit / receive the sidelink transmission results of each local model in the terminal device group and decide which local models to aggregate based on these results. Alternatively, for example, the third terminal device may upload the sidelink transmission results of its local model in the terminal device group to the control device and receive aggregation instructions. Furthermore, the execution order of the operations shown in Figure 13C is merely illustrative, and for example, the operations in S2204 and S2208 may be performed in reverse order or synchronously.

[0183] Furthermore, whether as a first terminal device, a second terminal device, or a third terminal device, the electronic device 20 can upload information regarding one or more of the following to the control device: sidelink status, communication quality, location, computing power, and user data volume, and can receive instructions from the control device to participate in joint learning or information regarding the division of terminal device groups.

[0184] As described above, the control equipment of this disclosure may be control equipment on the access network side of a wireless communication system (e.g., a base station), or equipment on the core network side of a wireless communication system (e.g., network element equipment). Accordingly, the operation of the control equipment described above may be performed by the base station, by the network element equipment, or even partially by the base station and partially by the network element equipment. In the latter case, which specific operations of these operations the base station and the network element equipment perform may be determined on a case-by-case basis or defined by an appropriate standard.

[0185] In particular, the control device according to the present disclosure may be one or more network element devices for realizing an application function (AF) in a core network. The following describes the information transmission path of the present disclosure in such an implementation form with reference to FIG. 14.

[0186] FIG. 14 schematically shows the overall architecture of a wireless communication system to which an embodiment of the present disclosure is applied. Each network element / function on the core network side related to FIG. 14 is as follows. - UPF: User Plane Function - NEF: Network Exposure Function - NRF: Network Repository Function - UDM: Unified Data Management - PCF: Policy Control Function - AF: Application Function - NSSF: Network Slice Selection Function - AUSF: Authentication Server Function - AMF: Access and Mobility Management Function - SMF: Session Management Function

[0187] Under the schematic overall architecture of Figure 14, for example, the models uploaded by each UE (e.g., local models or aggregated models) may be uploaded to the AF, which functions as a control device, via the gNB and UPF. The AF, which functions as a control device, may also obtain information about the UE's sidelink status from the NEF. For example, information about the UE's sidelink status may be stored in the NEF or reported from the UE to the AF via the NEF through any suitable path shown in Figure 14. In addition, other information described above, such as information uploaded by the UE to the control device (e.g., UE location information, uplink and downlink communication quality information, computing power, amount of user data for generating local models, etc.) and information transmitted to the UE by the control device (e.g., aggregated instructions, global models, information about UE group divisions, etc.), may be transmitted between the UE and the AF through any feasible path shown in Figure 14.

[0188] Note that Figure 14 shows only an exemplary overall architecture of a wireless communication system to which embodiments of this disclosure apply. The architecture of the wireless communication system may further include any appropriate functional / network elements and / or interfaces.

[0189] It should be understood that the machine-executable instructions in the machine-readable storage medium or program product according to the embodiments of this disclosure may be configured to perform operations corresponding to the embodiments of the above-described devices and methods. When referring to the embodiments of the above-described devices and methods, the embodiments of the machine-readable storage medium or program product will be obvious to those skilled in the art and will not be described further. Machine-readable storage medium or program product that holds or contains the above-described machine-executable instructions is also within the scope of this disclosure. Such storage mediums may include, but are not limited to, floppy disks, optical disks, magneto-optical disks, memory cards, memory sticks, and the like.

[0190] Furthermore, it should be understood that the above-described series of processes and devices may be implemented by software and / or firmware. When implemented by software and / or firmware, the programs constituting this software are installed from a storage medium or network to a computer having a dedicated hardware configuration, for example, the general-purpose computer / computer system 1300 shown in Figure 15, and this computer can perform various functions when various programs are installed. Figure 15 is a block diagram showing an exemplary configuration of a computer / computer system, which is an information processing device that can be adopted in embodiments of this disclosure. In one example, the computer may correspond to the exemplary terminal device described above in this disclosure. In another example, the computer may correspond to a network element device that functions as the exemplary terminal device described above in this disclosure. When functioning as a network element device, it is shown as a single block structure diagram, but the functions of the computer / computer system 1300 may be implemented as a distributed system. For example, some processes may be executed on one processor while other processes are executed on other remote processors. Other elements of the computer / computer system 1300 may be distributed similarly. Furthermore, the functions disclosed herein may be implemented on a single server or device that can be connected to each other via a network. Furthermore, it is not necessary to include one or more components of system 1300. In Figure 15, the central processing unit (CPU) 1301 executes various processes based on programs stored in read-only memory (ROM) 1302 or programs loaded from storage 1308 into random access memory (RAM) 1303. The RAM 1303 also stores data necessary for the CPU 1301 to execute various processes as needed.

[0191] The CPU 1301, ROM 1302, and RAM 1303 are connected to each other via bus 1304. The input / output interface 1305 is also connected to bus 1304.

[0192] The input unit 1306, which includes a keyboard and mouse, the output unit 1307, which includes a display such as a cathode ray tube (CRT) or liquid crystal display (LCD) and speakers, the storage unit 1308, which includes a hard disk, and the communication unit 1309, which includes a network interface card such as a LAN card or modem, are connected to the input / output interface 1305. The communication unit 1309 performs communication processing via a network, such as the Internet.

[0193] If necessary, drive 1310 is also connected to input / output interface 1305. Removable media 1311, such as magnetic disks, optical disks, magneto-optical disks, or semiconductor memory, are installed on drive 1310 as necessary, and computer programs read from them are installed on storage 1308 as necessary.

[0194] When the above series of processes are implemented by software, the programs that make up the software are installed from a network, such as the internet, or from a storage medium, such as removable media 1311.

[0195] Those skilled in the art should understand that such a storage medium is not limited to the removable media 1311 shown in Figure 15, which stores the program and is distributed separately from the device to provide the program to the user. Examples of removable media 1311 include magnetic disks (including floppy disks®), optical disks (including optical disk read-only memory (CD-ROM) and digital versatile disks (DVD)), magneto-optical disks (including MiniDisc (MD)®), and semiconductor memory. Alternatively, the storage medium may be a ROM 1302, a hard disk included in storage 1308, etc., which stores the program and is distributed to the user together with the device containing them. The technology disclosed herein can be applied to a variety of products.

[0196] For example, the electronic device 10 according to the embodiment of this disclosure can be implemented as various base stations or may be included in various base stations. For example, the electronic device 10 according to the embodiment of this disclosure can be implemented as various information processing devices employing the architecture shown in Figure 15 or may be included in various information processing devices. For example, the electronic device 20 according to the embodiment of this disclosure can be implemented as various terminal devices / user devices or may be included in various terminal devices / user devices.

[0197] For example, the base stations referred to in this disclosure may be implemented as any type of base station, such as evolved node B (gNB) such as macro gNBs and small gNBs. Small gNBs may be gNBs that cover cells smaller than macrocells, such as pico gNBs, micro gNBs, or femto gNBs. Alternatively, the base station may be implemented as any other type of base station, such as node Bs and base transceiver stations (BTS). The base station may include an entity (also called base station equipment) configured to control the radio communication and one or more remote radio heads (RRHs) located separately from the entity. Furthermore, various types of terminals described later can operate as base stations by performing base station functions temporarily or semi-permanently.

[0198] For example, the terminal equipment referred to in this disclosure may, in some examples, also be called user equipment and may be implemented as a mobile terminal (e.g., a smartphone, tablet personal computer (PC), notebook PC, portable game console, portable / dongle mobile router, and digital imaging device) or an in-vehicle terminal (e.g., a car navigation system). User equipment may be implemented as a terminal that performs machine-to-machine (M2M) communication (also called a machine-type communication (MTC) terminal). User equipment may also be a wireless communication module (e.g., an integrated circuit module including a single chip) mounted on each of the above terminals.

[0199] The following describes an example of this disclosure with reference to Figures 16 to 19.

[0200] [Examples related to base stations] It should be understood that the term "base station" in this disclosure has the full scope of its ordinary meaning and includes radio communication stations that are at least part of a radio communication system or radio system in order to perform communications. Examples of base stations include, for example, a base station may be one or both of a base station transceiver (BTS) and / or a base station controller (BSC) in a GSM® system, one or both of a radio network controller (RNC) and / or a Node B in a WCDMA® system, an eNB in ​​LTE and LTE-Advanced systems, a gNB, eLTE eNB, etc. appearing in a 5G communication system, or a corresponding network node in a future communication system, but not limited to these. Some functions of base stations in this disclosure may be implemented as entities with control functions over communications in D2D, M2M and V2V communication scenarios, or as entities with spectrum tuning functions in cognitive radio communication scenarios.

[0201] (Example 1) Figure 16 is a block diagram showing a first example of an exemplary configuration of a gNB to which the technology of the present disclosure can be applied. The gNB 1400 includes a plurality of antennas 1410 and base station equipment 1420. The base station equipment 1420 and each antenna 1410 can be connected to each other via RF cables. In one implementation, the gNB 1400 (or base station equipment 1420) herein may correspond to the electronic equipment 10 and / or electronic equipment 80 described above.

[0202] Each of the antennas 1410 includes one or more antenna elements (for example, multiple antenna elements included in a multi-input multiple-output (MIMO) antenna) and is used by the base station equipment 1420 to transmit and receive radio signals. As shown in Figure 16, the gNB 1400 may include multiple antennas 1410. For example, multiple antennas 1410 may be compatible with multiple frequency bands used by the gNB 1400.

[0203] The base station equipment 1420 includes a controller 1421, memory 1422, network interface 1423, and wireless communication interface 1425.

[0204] The controller 1421 may be, for example, a CPU or a DSP, and operates various higher-layer functions of the base station equipment 1420. For example, the controller 1421 generates data packets based on data in signals processed by the wireless communication interface 1425 and transmits the generated packets via the network interface 1423. The controller 1421 can bundle data from multiple baseband processors to generate bundle packets and transmit the generated bundle packets. The controller 1421 may have logic functions to perform controls such as radio resource control, radio bearer control, mobility management, reception control, and scheduling. This control can be performed in conjunction with nearby gNBs or core network nodes. The memory 1422 includes RAM and ROM and stores programs executed by the controller 421 and various types of control data (e.g., terminal lists, transfer power data, and scheduling data).

[0205] The network interface 1423 is a communication interface for connecting the base station device 1420 to the core network 1424. The controller 1421 can communicate with a core network node or another gNB via the network interface 1423. In this case, the gNB 1400 and the core network node or another gNB can be connected to each other by a logical interface (e.g., S1 interface and X2 interface). The network interface 1423 may be a wired communication interface or a wireless communication interface used for a wireless backhaul line. If the network interface 1423 is a wireless communication interface, compared with the frequency band used by the wireless communication interface 1425, the network interface 1923 can be used for wireless communication using a higher frequency band.

[0206] The wireless communication interface 1425 supports any cellular communication scheme (e.g., Long Term Evolution (LTE) and LTE-Advanced) and provides wireless connectivity to terminals in cells located on the gNB 1400 via the antenna 1410. The wireless communication interface 1425 may typically include, for example, a baseband (BB) processor 1426 and RF circuitry 1427. The BB processor 1426 can perform, for example, encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing, as well as various types of signal processing at different layers (e.g., L1, Media Access Control (MAC), Radio Link Control (RLC), Packet Data Aggregation Protocol (PDCP)). Instead of the controller 1421, the BB processor 1426 may have some or all of the above logic functions. The BB processor 1426 may be a memory in which a communication control program is stored, or it may be a module containing a processor and associated circuitry configured to execute the program. Program updates can change the functionality of the BB processor 1426. The module may be a card or bread inserted into a slot of the base station equipment 1420. Alternatively, the module may be a chip mounted on a card or bread. Simultaneously, the RF circuit 1427, including, for example, a mixer, filter, and amplifier, can transmit and receive radio signals via the antenna 1410. Figure 16 shows an example in which one RF circuit 1427 is connected to one antenna 1410, but the disclosure is not limited to this illustration, and one RF circuit 1427 may be connected to multiple antennas 1410 simultaneously.

[0207] As shown in Figure 16, the wireless communication interface 1425 may include multiple BB processors 1426. For example, multiple BB processors 1426 may be compatible with multiple frequency bands used by the gNB1400. As shown in Figure 16, the wireless communication interface 1425 may include multiple RF circuits 1427. For example, multiple RF circuits 1427 may be compatible with multiple antenna elements. Although Figure 16 shows an example in which the wireless communication interface 1425 includes multiple BB processors 1426 and multiple RF circuits 1427, the wireless communication interface 1425 may include a single BB processor 1426 or a single RF circuit 1427.

[0208] (Second example) Figure 17 is a block diagram showing a second example of an exemplary configuration of a gNB to which the technology of the present disclosure can be applied. The gNB 1530 includes a plurality of antennas 1540, a base station device 1550, and an RRH 1560. The RRH 1560 and each antenna 1540 can be connected to each other via an RF cable. The base station device 1550 and the RRH 1560 can be connected to each other via a high-speed line such as an optical fiber cable. In one implementation, the gNB 1530 (or base station device 1550) herein may correspond to the electronic devices 50 and / or 100 described above.

[0209] Each of the antennas 1540 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used by the RRH1560 to transmit and receive radio signals. As shown in Figure 17, the gNB1530 may include multiple antennas 1540. For example, multiple antennas 1540 may be compatible with multiple frequency bands used by the gNB1530.

[0210] The base station equipment 1550 includes a controller 1551, a memory 1552, a network interface 1553, a wireless communication interface 1555, and a connection interface 1557. The controller 1551, memory 1552, and network interface 1553 are the same as the controller 1421, memory 1422, and network interface 1423 described with reference to Figure 16.

[0211] The wireless communication interface 1555 supports any cellular communication scheme (e.g., LTE and LTE-Advanced) and provides wireless communication to a terminal located in the sector corresponding to the RRH1560 via the RRH1560 and antenna 1540. The wireless communication interface 1555 may typically include, for example, a BB processor 1556. The BB processor 1556 is the same as the BB processor 1426 described with reference to Figure 14, except that the BB processor 1556 is connected to the RF circuit 1564 of the RRH1560 via a connection interface 1557. As shown in Figure 17, the wireless communication interface 1555 may include multiple BB processors 1556. For example, multiple BB processors 1556 may be compatible with multiple frequency bands used by the gNB1530. Although Figure 17 shows an example in which the wireless communication interface 1555 includes multiple BB processors 1556, the wireless communication interface 1555 may include a single BB processor 1556.

[0212] The connection interface 1557 is an interface for connecting the base station equipment 1550 (wireless communication interface 1555) to the RRH1560. The connection interface 1557 may also be a communication module used for communication in the high-speed line described above that connects the base station equipment 1550 (wireless communication interface 1555) to the RRH1560.

[0213] The RRH1560 includes a connection interface 1561 and a wireless communication interface 1563.

[0214] The connection interface 1561 is an interface for connecting the RRH1560 (wireless communication interface 1563) to the base station equipment 1550. The connection interface 1561 may also be a communication module used for communication on the high-speed line described above.

[0215] The wireless communication interface 1563 transmits and receives radio signals via the antenna 1540. The wireless communication interface 1563 may typically include, for example, an RF circuit 1564. The RF circuit 1564 may include, for example, a mixer, filter, and amplifier, and be capable of transmitting and receiving radio signals via the antenna 1540. Figure 15 shows an example in which one RF circuit 1564 is connected to one antenna 1540, but the disclosure is not limited to this illustration, and one RF circuit 1564 may be connected to multiple antennas 1540 simultaneously.

[0216] As shown in Figure 17, the wireless communication interface 1563 may include multiple RF circuits 1564. For example, multiple RF circuits 1564 can support multiple antenna elements. Although Figure 17 shows an example in which the wireless communication interface 1563 includes multiple RF circuits 1564, the wireless communication interface 1563 may also include a single RF circuit 1564.

[0217] [Examples related to user equipment] (Example 1) Figure 18 is a block diagram showing an exemplary configuration of a smartphone 1600 to which the technology of the present disclosure can be applied. The smartphone 1600 includes a processor 1601, memory 1602, storage device 1603, external connection interface 1604, imaging device 1606, sensor 1607, microphone 1608, input device 1609, display device 1610, speaker 1611, wireless communication interface 1612, one or more antenna switches 1615, one or more antennas 1616, bus 1617, battery 1618, and auxiliary controller 1619. In one implementation, the smartphone 1600 (or processor 1601) herein may correspond to the electronic devices 50 and / or 100 described above.

[0218] The processor 1601 is, for example, a CPU or a system-on-a-chip (SoC) and can control the functions of the application layer and other layers of the smartphone 1600. The memory 1602 includes RAM and ROM and stores data and programs executed by the processor 1601. The storage device 1603 may include, for example, a storage medium such as semiconductor memory and a hard disk. The external connection interface 1604 is an interface for connecting external devices (e.g., memory cards and Universal Serial Bus (USB) devices) to the smartphone 1600.

[0219] The imaging device 1606 includes an image sensor (e.g., a charge-coupled device (CCD) and a complementary metal-oxide-semiconductor (CMOS)) and generates a captured image. Sensor 1607 may include a set of sensors such as a measuring sensor, a gyroscope, a geomagnetic sensor, and an accelerometer. Microphone 1608 converts sound input to the smartphone 1600 into an audio signal. Input device 1609 includes, for example, a touch sensor, a keypad, a keyboard, a button, or a switch configured to detect touches on the screen of the display device 1610 and receives operations or information input from the user. Display device 1610 includes a screen (e.g., a liquid crystal display (LCD) and an organic light-emitting diode (OLED) display) and displays the output image from the smartphone 1600. Speaker 1611 converts the audio signal output from the smartphone 1600 into sound.

[0220] The wireless communication interface 1612 supports any cellular communication method (e.g., LTE and LTE-Advanced) and performs wireless communication. The wireless communication interface 1612 may typically include, for example, a broadband processor 1613 and an RF circuit 1619. The broadband processor 1613 can perform various types of signal processing for wireless communication, such as encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing. At the same time, the RF circuit 1614, which includes, for example, a mixer, filter, and amplifier, can transmit and receive wireless signals via the antenna 1616. The wireless communication interface 1612 may be a single chip module on which the broadband processor 1613 and the RF circuit 1614 are integrated. As shown in Figure 18, the wireless communication interface 1612 may include multiple broadband processors 1613 and multiple RF circuits 1614. Figure 18 shows an example in which the wireless communication interface 1612 includes multiple BB processors 1613 and multiple RF circuits 1614, but the wireless communication interface 1612 may also include a single BB processor 1613 or a single RF circuit 1614.

[0221] In addition to the cellular communication method, the wireless communication interface 1612 can support other types of wireless communication methods, such as short-range wireless communication, proximity communication, and wireless local network (LAN) methods. In this case, the wireless communication interface 1612 may include a BB processor 1613 and an RF circuit 1614 for each wireless communication method.

[0222] Each of the antenna switches 1615 switches the destination of the antenna 1616 among multiple circuits included in the wireless communication interface 1612 (for example, circuits used for different wireless communication methods).

[0223] Each of the antennas 1616 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used to transmit and receive radio signals of the wireless communication interface 1612. As shown in Figure 18, the smartphone 1600 may include multiple antennas 1616. Although Figure 18 shows an example in which the smartphone 1600 includes multiple antennas 1616, the smartphone 1600 may also include a single antenna 1616.

[0224] The smartphone 1600 may also include antennas 1616 for each wireless communication method. In this case, the antenna switch 1615 may be omitted from the arrangement of the smartphone 1600.

[0225] Bus 1617 connects the processor 1601, memory 1602, storage device 1603, external connection interface 1604, imaging device 1606, sensor 1607, microphone 1608, input device 1609, display device 1610, speaker 1611, wireless communication interface 1612, and auxiliary controller 1619 to each other. Battery 1618 provides power to each block of the smartphone 1600 shown in Figure 18 via power lines. Power lines are partially indicated by dotted lines in the drawing. The auxiliary controller 1619 operates the minimum necessary functions of the smartphone 1600, for example, in sleep mode.

[0226] (Second example) Figure 19 is a block diagram showing an exemplary arrangement of a car navigation device 1720 to which the technology of the present disclosure can be applied. The car navigation device 1720 includes a processor 1721, memory 1722, a Global Positioning System (GPS) module 1724, a sensor 1725, a data interface 1726, a content player 1727, a storage medium interface 1728, an input device 1729, a display device 1730, a speaker 1731, a wireless communication interface 1733, one or more antenna switches 1736, one or more antennas 1737, and a battery 1738. In one implementation, the car navigation device 1720 (or processor 1721) herein may correspond to the electronic devices 50 and / or 100 described above.

[0227] The processor 1721 is, for example, a CPU or SoC, and can control the navigation and other functions of the car navigation device 1720. The memory 1722 includes RAM and ROM and stores data and programs executed by the processor 1721.

[0228] The GPS module 1724 measures the position (e.g., latitude, longitude, altitude) of the car navigation device 1720 using GPS signals received from GPS satellites. The sensor 1725 may include a set of sensors, such as a gyro sensor, a geomagnetic sensor, and an air pressure sensor. The data interface 1726 is connected to, for example, an in-vehicle network 1741 via a terminal (not shown) to acquire data generated by the vehicle (e.g., vehicle speed data).

[0229] The content player 1727 plays content stored on a storage medium (e.g., CDs and DVDs). This storage medium is inserted into the storage medium interface 1728. The input device 1729 includes, for example, a touch sensor, button, or switch configured to detect touches on the screen of the display device 1730, and receives operations or information input from the user. The display device 1730 includes, for example, an LCD or OLED display screen, and displays images of the navigation function or the played content. The speaker 1731 outputs sounds of the navigation function or the played content.

[0230] The wireless communication interface 1733 supports any cellular communication scheme (e.g., LTE and LTE-Advanced) and performs wireless communication. The wireless communication interface 1733 may typically include, for example, a broadband processor 1734 and an RF circuit 1735. The broadband processor 1734 can perform various types of signal processing for wireless communication, including, for example, encoding / decoding, modulation / demodulation, and multiplexing / demultiplexing. At the same time, the RF circuit 1735, including, for example, a mixer, filter, and amplifier, can transmit and receive wireless signals via the antenna 1737. The wireless communication interface 1733 may be a single chip module on which the broadband processor 1734 and the RF circuit 1735 are integrated. As shown in Figure 19, the wireless communication interface 1733 may include multiple broadband processors 1734 and multiple RF circuits 1735. Figure 19 shows an example in which the wireless communication interface 1733 includes multiple BB processors 1734 and multiple RF circuits 1735, but the wireless communication interface 1733 may also include a single BB processor 1734 or a single RF circuit 1735.

[0231] In addition to the cellular communication method, the wireless communication interface 1733 can support other types of wireless communication methods, such as short-range wireless communication, proximity communication, and wireless LAN. In this case, the wireless communication interface 1733 may include a BB processor 1734 and an RF circuit 1735 for each wireless communication method.

[0232] Each of the antenna switches 1736 switches the destination of the antenna 1737 among multiple circuits included in the wireless communication interface 1733 (for example, circuits used for different wireless communication methods).

[0233] Each of the antennas 1737 includes one or more antenna elements (for example, multiple antenna elements included in a MIMO antenna) and is used to transmit and receive radio signals of the wireless communication interface 1733. As shown in Figure 19, the car navigation device 1720 may include multiple antennas 1737. Although Figure 19 shows an example in which the car navigation device 1720 includes multiple antennas 1737, the car navigation device 1720 may include a single antenna 1737.

[0234] The car navigation device 1720 may also include antennas 1737 for each wireless communication method. In this case, the antenna switch 1736 may be omitted from the arrangement of the car navigation device 1720.

[0235] Battery 1738 supplies power to each block of the car navigation device 1720 shown in Figure 19 via power supply lines. The power supply lines are partially indicated by dotted lines in the drawing. Battery 1738 stores the power supplied from the vehicle.

[0236] The technology described herein may be implemented as an in-vehicle system (or vehicle) 1740 which includes one or more blocks of a car navigation device 1720, an in-vehicle network 1741, and a vehicle module 1742. The vehicle module 1742 generates vehicle data (e.g., vehicle speed, engine speed, fault information) and outputs the generated data to the in-vehicle network 1741.

[0237] The above describes exemplary embodiments of the present disclosure with reference to the drawings, but of course, the present disclosure is not limited to these examples. Those skilled in the art will understand that various changes and modifications can be made within the scope of the appended claims, and that such changes and modifications will fall within the scope of the art of the present disclosure.

[0238] It should be understood that the machine-executable instructions in the machine-readable storage medium or program product according to the embodiments of this disclosure may be configured to perform operations corresponding to the embodiments of the above-described devices and methods. When referring to the embodiments of the above-described devices and methods, the embodiments of the machine-readable storage medium or program product will be obvious to those skilled in the art and will not be described further. Machine-readable storage medium or program product that holds or contains the above-described machine-executable instructions is also within the scope of this disclosure. Such storage mediums may include, but are not limited to, floppy disks, optical disks, magneto-optical disks, memory cards, memory sticks, and the like.

[0239] Furthermore, it should be understood that the above series of processes and devices may be implemented by software and / or firmware. When implemented by software and / or firmware, the relevant programs constituting the relevant software are stored in the storage medium of the relevant device (for example, the memory 104 or 204 of the electronic device 10 shown in Figure 10, or the electronic device 20 shown in Figure 12), and when the program is executed, various functions can be performed.

[0240] For example, the multiple functions included in one means in the above embodiments may be implemented by separate devices. Alternatively, the multiple functions implemented by multiple means in the above embodiments may each be implemented by separate devices. Furthermore, one of the above functions may be implemented by multiple means. Of course, such configurations are within the scope of the technology of this disclosure.

[0241] In this specification, the steps described in the flowchart include not only processes that are executed chronologically in the order described, but also processes that are not necessarily executed chronologically but in parallel or individually. Furthermore, of course, the order of steps that are executed chronologically may also be changed as appropriate.

[0242] While the present disclosure and its merits have been described in detail, it should be understood that various modifications, substitutions, and transformations are possible, provided they do not fall outside the spirit and scope of the present disclosure, which is limited to the claims attached. Furthermore, the terms “including,” “incorporating,” or any other variant thereof in the embodiments of the present disclosure, by their non-exclusive nature, mean that a process, method, article, or apparatus containing a set of elements includes not only those elements but also other elements not explicitly stated, or elements specific to such a process, method, article, or apparatus. Unless further restrictions are imposed, an element limited by “including one…” does not exclude other identical elements from being included in a process, method, article, or apparatus containing such elements.

[0243] Furthermore, this disclosure may have the following configurations. (1) Control equipment used in wireless communication systems, The system includes a processing circuit configured to receive multiple models associated with multiple terminal devices in the wireless communication system, wherein the multiple models include at least one aggregated model. A control device in which each of the at least one aggregated model is generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by training on the local data of a corresponding terminal device. (2) The processing circuit further, The control device according to (1), further configured to receive identification information that identifies terminal devices corresponding to local models aggregated in each of the at least one aggregation models. (3) The processing circuit further, The control device according to (1), configured to receive local model sidelink transmission result information indicating the success or failure of transmitting a local model from one or more terminal devices to one or more other terminal devices, among some or all of the aforementioned plurality of terminal devices. (4) The processing circuit further, The control device according to (3), configured to transmit aggregation instructions in response to at least local model sidelink transmission result information, wherein the aggregation instructions instruct the terminal device whether its local model should be aggregated with the local models of one or more other terminal devices, and, if aggregation is necessary, which of the local models of one or more other terminal devices the local model of the terminal device should be aggregated with. (5) The processing circuit further, The control device according to (2) or (3), configured to aggregate one or more received local models and / or one or more aggregated models so that the same terminal device's local model is not aggregated repeatedly. (6) The control device according to any one of (1) to (4), wherein the plurality of models further comprises one or more local models, and at least one of the one or more local models is the same as at least one of the local models aggregated into the at least one aggregated model. (7) The processing circuit further, Regarding each of the aforementioned multiple terminal devices, Information regarding the sidelink status of the terminal device, indicating at least whether a sidelink exists between the terminal device and one or more terminal devices in the wireless communication system, and the communication quality of the sidelink, The location information of the terminal device and, Information indicating the communication quality of the uplink and downlink of the terminal device, Information indicating the computing power of the terminal device, A control device according to any one of (1) to (4), configured to obtain one or more of the following: information indicating the amount of user data for generating a local model of the terminal device. (8) The processing circuit further, The control device according to (7), which is configured to select one or more terminal devices from the plurality of terminal devices to participate in federated learning based on the information obtained, and / or to divide at least some of the terminal devices participating in federated learning into one or more terminal device groups used for federated learning, each terminal device group including at least two terminal devices that communicate via sidelinks. (9) The one or more terminal devices: One or more groups of first types of terminal equipment, each including one central terminal equipment and at least one decentralized terminal equipment communicating with the central terminal equipment via a side link, and / or It includes one or more groups of a second type of terminal equipment, each containing at least two terminal equipment that communicate with each other via sidelinks, The control device according to (8), wherein in a first type of terminal device group, a central terminal device aggregates a local model corresponding to the central terminal device and a local model corresponding to at least one non-centralized terminal device in the first type of terminal device group, received via a side link, and in a second type of terminal device group, each terminal device aggregates a local model corresponding to the terminal device and a local model corresponding to other terminal devices in the second type of terminal device group, received via a side link. (10) The processing circuit further, The control device according to (8), configured to transmit information relating to the division of one or more terminal device groups. (11) The control device is an electronic device for realizing a base station or application function (AF), as described in any one of (1) to (4). (12) A first terminal device used in a wireless communication system, The control device of the aforementioned wireless communication system uploads a first local model, which was obtained by training based on local data. A first terminal device including a processing circuit configured to transmit a first local model to a second terminal device of the wireless communication system via sidelink communication for the generation of an aggregated model by the second terminal device. (13) The processing circuit further, The first terminal device according to (12), wherein the control device is configured to upload information indicating the success or failure of the transmission of the first local model to the second terminal device. (14) A second terminal device used in a wireless communication system, via sidelink communication, each of the at least one first terminal devices of the wireless communication system receives at least one first local model that has been trained on its local data by one of the at least one first terminal devices, A second terminal device, comprising a processing circuit configured to generate an aggregated model by aggregating a second local model obtained by training on local data with the at least one first local model, and uploading the aggregated model to the control device of the wireless communication system. (15) The processing circuit further, A second terminal device as described in (14), configured to upload information that identifies each terminal device corresponding to the aggregated local model, together with the aggregated model. (16) The processing circuit further, The system receives an aggregation instruction that tells the second terminal device whether the second local model should be aggregated with one or more first local models, and if aggregation is necessary, which one or more first local models should be aggregated with the second local model. A second terminal device according to (14), configured to generate an aggregation model in response to the aggregation instruction. (17) A third terminal device used in a wireless communication system, wherein the third terminal device constitutes a terminal device group with one or more other terminal devices, via sidelink communication, each terminal device in the terminal device group receives at least one local model that has been trained by one of the other terminal devices based on its local data, A third local model, which is trained based on local data, and at least one of the received local models are aggregated to generate an aggregated model, and the aggregated model is uploaded to the control device. A third terminal device including a processing circuit configured to transmit a third local model to each other terminal device in the terminal device group via sidelink communication, for use by each other terminal device in generating an aggregated model. (18) The processing circuit further, A third terminal device according to (17), configured to transmit information identifying each terminal device corresponding to the aggregated local model, together with the aggregated model. (19) The processing circuit further, Information indicating the success or failure of the transmission of the third local model to each other terminal device in the terminal device group is transmitted via sidelink communication. A third terminal device according to (18), configured to receive information from at least one other terminal device in the terminal device group via sidelink communication indicating the success or failure of a local model transmission from the other terminal device to each terminal device in the terminal device group. (20) The processing circuit further, If there are no failures in the local model sidelink transmission for the third terminal device, The third terminal device according to (19), configured to generate an aggregated model by aggregating a third local model obtained by training on local data and one or more local models corresponding to one or more other terminal devices that have been received and for which there are no failures in sidelink transmission of the local model relating thereto. (21) The processing circuit further, If there is a failure in the local model sidelink transmission for the third terminal device, The third local model is uploaded to the control device individually, or The third local model and one or more local models corresponding to one or more other terminal devices for which there are failures in the sidelink transmission of the received local model are aggregated to generate a new aggregated model, and the new aggregated model is uploaded to the control device, or The third terminal device according to (19), configured to transmit a third local model to another terminal device in the terminal device group that has failed to transmit a sidelink of a local model relating to another terminal device, to provide for the generation of a new aggregated model, and to upload it to the control device. (22) The processing circuit further, The third terminal device according to (17), configured to transmit information to the control device indicating the success or failure of transmitting a third local model to each other terminal device in the terminal device group. (23) The processing circuit further, The system receives an aggregation instruction that tells the third terminal device whether the third local model should be aggregated with one or more local models, and if aggregation is necessary, which one or more local models should be aggregated with the third local model. A third terminal device according to (22), configured to generate an aggregation model in response to the aggregation instruction. (24) The processing circuit further, The aforementioned control device, Information regarding the sidelink status of the terminal device, indicating at least whether a sidelink exists between the terminal device and one or more terminal devices in the wireless communication system, and the communication quality of the sidelink, The location information of the aforementioned terminal device, Information indicating the communication quality of the uplink and downlink of the aforementioned terminal equipment, Information indicating the computing power of the aforementioned terminal device, The terminal device according to any one of (12), (14), or (17), uploads one or more of the information of the terminal device indicating the amount of user data for generating a local model, and uses this information to select and / or divide terminal devices into groups corresponding to terminal devices participating in federated learning, wherein the terminal device group includes at least two terminal devices that communicate via sidelinks. (25) The processing circuit further, A terminal device according to any one of (12), (14), or (17), configured to receive information relating to the division of one or more terminal device groups. (26) A method on the control equipment side used in a wireless communication system, The wireless communication system includes receiving multiple models associated with multiple terminal devices, wherein the multiple models include at least one aggregated model. A method wherein each of the at least one aggregated model is generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by training on the local data of a corresponding terminal device. (27) A method for a first terminal device used in a wireless communication system, Uploading a first local model, obtained by training based on local data, to the control device of the aforementioned wireless communication system, A method comprising transmitting a first local model to a second terminal device of the wireless communication system via sidelink communication for the generation of an aggregated model by the second terminal device. (28) A method for a second terminal device used in a wireless communication system, The system receives, via sidelink communication, at least one first local model obtained by training on local data by one of the first terminal devices of the wireless communication system, A method comprising generating an aggregated model by aggregating a second local model obtained by training on local data with the at least one first local model, and uploading the aggregated model to the control equipment of the wireless communication system. (29) A method for a third terminal device used in a wireless communication system, wherein the third terminal device constitutes a terminal device group with one or more other terminal devices, and the method is The terminal device group receives, via sidelink communication, at least one local model obtained by training on local data from at least one other terminal device in the terminal device group, each of which is trained by one of the other terminal devices. The process involves aggregating a third local model, which is trained based on local data, with at least one of the received local models to generate an aggregated model, and uploading the aggregated model to the control device. A method comprising transmitting a third local model to each other terminal device in the terminal device group via sidelink communication for the generation of an aggregated model by each other terminal device. (30) A non-temporary, computer-readable storage medium that stores executable instructions that, when executed, accomplish one of the methods described in (26) to (29). (31) A computer program product that, when executed, contains executable instructions that accomplish any one of the methods described in (26) to (29). (32) Processor and A device including a memory device that, when executed, stores executable instructions that carry out the method described in any one of (26) to (29).

Claims

1. A control device used in a wireless communication system, The system includes a processing circuit configured to receive multiple models associated with multiple terminal devices in the wireless communication system, wherein the multiple models include at least one aggregated model. A control device in which each of the at least one aggregated model is generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by training on the local data of a corresponding terminal device.

2. The processing circuit further, The control device according to claim 1, further configured to receive identification information that identifies terminal devices corresponding to local models aggregated in each of the at least one aggregation models.

3. The processing circuit further, The control device according to claim 1, configured to receive local model sidelink transmission result information indicating the success or failure of transmitting a local model from one or more terminal devices to one or more other terminal devices, among some or all of the aforementioned plurality of terminal devices.

4. The processing circuit further, The control device according to claim 3, configured to transmit an aggregation instruction in response to at least local model sidelink transmission result information, wherein the aggregation instruction instructs the terminal device whether its local model should be aggregated with the local models of one or more other terminal devices, and, if aggregation is necessary, which of the local models of one or more other terminal devices the local model of the terminal device should be aggregated with.

5. The processing circuit further, The control device according to claim 2 or 3, configured to aggregate one or more received local models and / or one or more aggregated models so that the local models of the same terminal device are not repeatedly aggregated.

6. The control device according to any one of claims 1 to 4, wherein the plurality of models further comprises one or more local models, and at least one of the one or more local models is the same as at least one of the local models aggregated into the at least one aggregated model.

7. The processing circuit further, Regarding each of the aforementioned multiple terminal devices, Information regarding the sidelink status of the terminal device, indicating at least whether a sidelink exists between the terminal device and one or more terminal devices in the wireless communication system, and the communication quality of the sidelink, The location information of the terminal device and, Information indicating the communication quality of the uplink and downlink of the terminal device, Information indicating the computing power of the terminal device, A control device according to any one of claims 1 to 4, configured to obtain one or more of the following: information indicating the amount of user data for generating a local model of the terminal device.

8. The processing circuit further, The control device according to claim 7, configured to select one or more terminal devices from the plurality of terminal devices to participate in federated learning based on the information obtained, and / or to divide at least some of the terminal devices participating in federated learning into one or more terminal device groups used for federated learning, wherein each terminal device group includes at least two terminal devices that communicate via sidelinks.

9. The aforementioned one or more terminal devices are One or more groups of first types of terminal devices, each including one central terminal device and at least one decentralized terminal device communicating with the central terminal device via a side link, and / or It includes one or more groups of a second type of terminal equipment, each containing at least two terminal equipment that communicate with each other via sidelinks, The control device according to claim 8, wherein in a first type of terminal device group, a central terminal device aggregates a local model corresponding to the central terminal device and a local model corresponding to at least one non-centralized terminal device in the first type of terminal device group, received via a side link, and in a second type of terminal device group, each terminal device aggregates a local model corresponding to the terminal device and a local model corresponding to other terminal devices in the second type of terminal device group, received via a side link.

10. The processing circuit further, The control device according to claim 8, configured to transmit information relating to the division of one or more terminal device groups.

11. The control device is an electronic device for realizing a base station or application function (AF), as described in any one of claims 1 to 4.

12. A first terminal device used in a wireless communication system, The control device of the aforementioned wireless communication system uploads a first local model, which was obtained by training based on local data. A first terminal device including a processing circuit configured to transmit a first local model to a second terminal device of the wireless communication system via sidelink communication for the generation of an aggregated model by the second terminal device.

13. The processing circuit further, The first terminal device according to claim 12, wherein the control device is configured to upload information indicating the success or failure of the transmission of the first local model to the second terminal device.

14. A second terminal device used in a wireless communication system, via sidelink communication, each of the at least one first terminal devices of the wireless communication system receives at least one first local model that has been trained by one of the at least one first terminal devices based on its local data, A second terminal device, comprising a processing circuit configured to generate an aggregated model by aggregating a second local model obtained by training on local data with the at least one first local model, and uploading the aggregated model to the control device of the wireless communication system.

15. The processing circuit further, The second terminal device according to claim 14, which is configured to upload information identifying each terminal device corresponding to the aggregated local model, together with the aggregated model.

16. The processing circuit further, The system receives an aggregation instruction that tells the second terminal device whether the second local model should be aggregated with one or more first local models, and if aggregation is necessary, which one or more first local models should be aggregated with the second local model. The second terminal device according to claim 14, configured to generate an aggregation model in response to the aggregation instruction.

17. A third terminal device used in a wireless communication system, wherein the third terminal device constitutes a terminal device group with one or more other terminal devices, and the third terminal device is via sidelink communication, each terminal device in the terminal device group receives at least one local model that has been trained by one of the other terminal devices based on its local data, A third local model, which is trained based on local data, and at least one of the received local models are aggregated to generate an aggregated model, and the aggregated model is uploaded to the control device. A third terminal device including a processing circuit configured to transmit a third local model to each other terminal device in the terminal device group via sidelink communication, for use by each other terminal device in generating an aggregated model.

18. The processing circuit further, The third terminal device according to claim 17, configured to transmit information identifying each terminal device corresponding to the aggregated local model, together with the aggregated model.

19. The processing circuit further, Information indicating the success or failure of the transmission of the third local model to each other terminal device in the terminal device group is transmitted via sidelink communication. The third terminal device according to claim 18, configured to receive information from at least one other terminal device in the terminal device group via sidelink communication indicating the success or failure of a local model transmission from that other terminal device to each terminal device in the terminal device group.

20. The processing circuit further, If there are no failures in the local model sidelink transmission for the third terminal device, The third terminal device according to claim 19, configured to generate an aggregated model by aggregating a third local model obtained by training on local data and one or more local models corresponding to one or more other terminal devices that have been received and for which there are no failures in sidelink transmission of the local model relating thereto.

21. The processing circuit further, If there is a failure in the local model sidelink transmission for the third terminal device, The third local model is uploaded to the control device individually, or The third local model and one or more local models corresponding to one or more other terminal devices for which there are failures in the sidelink transmission of the received local model are aggregated to generate a new aggregated model, and the new aggregated model is uploaded to the control device, or The third terminal device according to claim 19, configured to transmit a third local model to another terminal device in the terminal device group that has failed to transmit a sidelink of a local model relating to that other terminal device, to provide for the generation of a new aggregated model, and to upload it to the control device.

22. The processing circuit further, The third terminal device according to claim 17, configured to transmit information to the control device indicating the success or failure of transmitting a third local model to each other terminal device in the terminal device group.

23. The processing circuit further, The system receives an aggregation instruction that tells the third terminal device whether the third local model should be aggregated with one or more local models, and if aggregation is necessary, which one or more local models should be aggregated with the third local model. The third terminal device according to claim 22, configured to generate an aggregation model in response to the aggregation instruction.

24. The processing circuit further, The aforementioned control device, Information regarding the sidelink status of the terminal device, indicating at least whether a sidelink exists between the terminal device and one or more terminal devices in the wireless communication system, and the communication quality of the sidelink, The location information of the aforementioned terminal device, Information indicating the communication quality of the uplink and downlink of the aforementioned terminal equipment, Information indicating the computing power of the aforementioned terminal device, The terminal device according to any one of claims 12, 14, or 17, which uploads one or more of the information of the terminal device indicating the amount of user data for generating a local model, and uses this information to select and / or divide terminal devices into groups corresponding to terminal devices participating in federated learning, wherein the terminal device group includes at least two terminal devices that communicate via sidelinks.

25. The processing circuit further, A terminal device according to any one of claims 12, 14, or 17, configured to receive information relating to the division of one or more terminal device groups.

26. A method used on the control equipment side in a wireless communication system, The system includes receiving multiple models associated with multiple terminal devices in the wireless communication system, wherein the multiple models include at least one aggregated model. A method wherein each of the at least one aggregated model is generated by a corresponding terminal device by aggregating its local model with each local model from one or more other terminal devices, and each local model is obtained by training on the local data of a corresponding terminal device.

27. A method for a first terminal device used in a wireless communication system, Uploading a first local model, obtained by training based on local data, to the control device of the aforementioned wireless communication system, A method comprising transmitting a first local model to a second terminal device of the wireless communication system via sidelink communication, to be used by the second terminal device to generate an aggregated model.

28. A method for a second terminal device used in a wireless communication system, The system receives, via sidelink communication, at least one first local model obtained by training one of the at least one first terminal devices based on its local data from at least one first terminal device of the wireless communication system, A method comprising generating an aggregated model by aggregating a second local model obtained by training on local data with at least one first local model, and uploading the aggregated model to the control equipment of the wireless communication system.

29. A method for a third terminal device used in a wireless communication system, wherein the third terminal device constitutes a terminal device group with one or more other terminal devices, and the method is The terminal device group receives, via sidelink communication, at least one local model obtained by training on local data from at least one other terminal device in the terminal device group, The process involves aggregating a third local model, which is trained based on local data, with at least one of the received local models to generate an aggregated model, and then uploading the aggregated model to the control device. A method comprising transmitting a third local model to each other terminal device in the terminal device group via sidelink communication for the generation of an aggregated model by each other terminal device.

30. A non-temporary, computer-readable storage medium that stores executable instructions, when executed, that implement the method according to any one of claims 26 to 29.

31. A computer program product comprising executable instructions that, when executed, implement the method according to any one of claims 26 to 29.

32. Processor and A device comprising a memory device that, when executed, stores executable instructions that enable the method according to any one of claims 26 to 29.

Images