Wireless communication methods and related devices

Abstract

This application provides a wireless communication method and related apparatus. The method constructs merged model parameters using model parameters from CSI prediction models trained locally by multiple UEs, and sends these merged model parameters to UEs that need to update their local CSI prediction models. This allows UEs to directly update their local CSI prediction models using the received merged model parameters without requiring local training. Furthermore, the merged model, obtained from the local CSI prediction models of multiple UEs, improves the generalization of the CSI prediction model and its versatility across different UEs. Moreover, merging the model parameters of local models from multiple UEs with similar CSI data distributions ensures the rationality of the models involved in the merging process.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a wireless communication method and related apparatus. Background Technology

[0002] In 5G mobile communication systems, network devices need to obtain Channel State Information (CSI) between terminal devices and network devices, and then perform resource scheduling for uplink / downlink data transmission based on the CSI.

[0003] The CSI prediction scheme based on artificial intelligence (AI) / machine learning (ML) is to use some historical CSI information and / or current CSI information as input into the CSI prediction model, and output the prediction result of CSI at future time.

[0004] However, training and updating the CSI prediction model locally on the terminal device may result in poor generalization of the CSI prediction model, and the update speed is slow and requires a large amount of data. Summary of the Invention

[0005] In view of the above, this application provides a wireless communication method and related apparatus to solve at least some of the aforementioned problems, and the disclosed technical solution is as follows:

[0006] In a first aspect, this application provides a wireless communication method applied to a first terminal. The method includes: sending a model update request message, the model update request message being used to request an update of the CSI prediction model parameters of the first terminal, the model update request message including a first CSI data distribution feature sequence, the first CSI data distribution feature sequence being used to characterize the distribution features of the CSI data collected by the first terminal; receiving a model update instruction message sent by a network device, the model update instruction message being used to instruct the first terminal to update the model parameters of the CSI prediction model, the model update instruction message including fusion model parameters corresponding to the CSI prediction model, the fusion model parameters being obtained by the network device based on the model parameters of the CSI prediction model in at least one second terminal; and responding to model update information by updating the model parameters of the CSI prediction model based on the fusion model parameters carried in the model update instruction message.

[0007] In this way, a merged model parameter is constructed using the model parameters of CSI prediction models trained locally by multiple UEs. This merged model parameter is then sent to the UEs that need to update their local CSI prediction models. The UEs can directly update their local CSI prediction models using the received merged model parameter without needing to perform local training. Furthermore, the merged model, derived from the local CSI prediction models of multiple UEs, improves the generalization ability of the CSI prediction model and its versatility across different UEs. Specifically, the merged model parameter is obtained by merging the model parameters of the local models of multiple UEs with similar CSI data distributions, thus ensuring the rationality of the models involved in the merging process.

[0008] In one possible implementation of the first aspect, the process of obtaining the first CSI data distribution feature sequence includes: hash encoding the obtained CSI data to obtain the first CSI data distribution feature sequence. Encoding the UE's CSI data in this way can significantly reduce the data volume, thereby reducing communication overhead. Furthermore, using the distance between the encoded feature sequences to determine the similarity of data distribution simplifies the judgment process and also provides a basis for rationally selecting models for model merging.

[0009] In one possible implementation of the first aspect, hash encoding the obtained CSI data to obtain a first CSI data distribution feature sequence includes: generating K random complex vectors, where K Represents the number of carrier frequencies; extracted from the CSI dataset obtained from the first terminal. L Each of the following CSI tensors is recently obtained and converted into a CSI feature complex vector. The random complex vectors have the same dimension as the CSI feature complex vectors. Each CSI feature complex vector is then compared with... K The inner product of several random complex vectors is obtained. K × L Each inner product operation result is converted into a binary bit; K × L The first CSI data distribution feature sequence is obtained by arranging the binary bits in sequence.

[0010] In one possible implementation of the first aspect, each CSI feature complex vector is respectively compared with... K The inner product of several random complex vectors is obtained. K × L Each inner product operation result is converted into a binary bit, including converting each inner product operation result into a binary bit according to the following formula:

[0011]

[0012] in, Represents a random complex vector The k-th complex vector in Indicates UE i The CSI feature complex vector obtained by CSI tensor transformation at time t; express and The binary bits obtained after performing the inner product calculation express and The real part of the result of the inner product operation is greater than 0; Let L represent the most recent L acquisition times, and S represent the total number of CSI tensors contained in the CSI dataset.

[0013] In one possible implementation of the first aspect, K × L The first CSI data distribution feature sequence is obtained by arranging the binary bits in sequence, including: arranging them according to the following formula:

[0014]

[0015] in, Indicates UE i The CSI data distribution feature sequence corresponding to the CSI dataset.

[0016] Secondly, this application also provides a wireless communication method applied to a network device. The method includes: receiving a model update request message sent by a first terminal, the model update request message being used to request an update of the model parameters of the first terminal's CSI prediction model, the model update request message including a first CSI data distribution feature sequence, the first CSI data distribution feature sequence being used to characterize the distribution features of the CSI data collected by the first terminal; sending a similarity discrimination indication message to a second terminal, the similarity discrimination indication message being used to instruct the second terminal to determine the similarity between its own CSI data distribution features and the first CSI data distribution feature sequence of the first terminal, the similarity discrimination indication message including a model merging process ID and the first CSI data distribution feature sequence; and receiving a similarity discrimination response message sent by the second terminal, the similarity discrimination response message including the second terminal's device ID and model merging... The similarity between the process ID and the CSI data distribution feature sequence of the second terminal and the first terminal is determined. Based on the similarity in each similarity judgment response message, a list of candidate terminals participating in model merging is determined, including the device ID corresponding to the candidate terminal. A model upload instruction message is sent, which instructs the second terminal to report CSI prediction model parameters. The model upload instruction message includes the model merging process ID and the list of candidate terminals. A model upload response message is received, which includes the model parameters of the CSI prediction model in the second terminal. The model parameters sent by each second terminal are fused to obtain fused model parameters. A model update instruction message is sent to the first terminal, which instructs the first terminal to update the model parameters of the CSI prediction model. The model update instruction message includes the model merging process ID and the fused model parameters.

[0017] In one possible implementation of the second aspect, the method further includes: if no similarity judgment response message is received within a preset waiting time threshold after sending the similarity discrimination indication message, sending a model training indication message to the first terminal, the model training indication message being used to instruct the first terminal to perform local training and updates for the CSI prediction model, the model training indication message including the model merging process ID.

[0018] In one possible implementation of the second aspect, a candidate terminal list for model merging is determined based on the similarity in each similarity judgment response message. This includes selecting a preset number of second terminals in descending order of similarity to obtain the candidate terminal list. In this way, by using the CSI data distribution similarity, the model parameters of the CSI prediction models of other terminals with similar local CSI data distributions to the terminal initiating the model update request can be updated, improving the rationality and usability of model merging.

[0019] In one possible implementation of the second aspect, a candidate terminal list is obtained by selecting a preset number of second terminals in descending order of similarity, including: selecting a preset number of second terminals in ascending order of the distance between the CSI data distribution feature sequence of each second terminal and the first CSI data distribution feature sequence.

[0020] In one possible implementation of the second aspect, a candidate terminal list is obtained by selecting a preset number of second terminals according to the order of increasing distance between the CSI data distribution feature sequence of each second terminal and the first CSI data distribution feature sequence. This includes: determining a first number of second terminals that send similarity discrimination response messages; if the first number is less than a preset threshold, then determining all second terminals that send similarity discrimination response messages to form a candidate terminal list; if the first number is greater than the preset threshold, then selecting a preset threshold number of terminals from the first number to form a candidate terminal list, according to the order of increasing distance between the CSI data distribution feature sequence of each second terminal and the first CSI data distribution feature sequence. This reasonably limits the number of terminals participating in model merging, avoiding excessively long times for obtaining fusion model parameters due to an excessive number of participating terminals, and improving the speed at which network devices respond to model update requests.

[0021] In one possible implementation of the second aspect, the model parameters sent by each second terminal are fused to obtain fused model parameters, including: obtaining the average model parameters corresponding to the model parameters sent by each second terminal using the arithmetic mean method, and obtaining the average variance value between each received model parameter and the average model parameter; normalizing the negative exponent of each average variance value to obtain the weight coefficient corresponding to each received model parameter; and performing a weighted summation on each received model parameter to obtain the fused model parameters.

[0022] Thirdly, this application also provides a wireless communication method applied to a second terminal. The method includes: receiving a similarity discrimination indication message sent by a network device, the similarity discrimination indication message being used to instruct the second terminal to determine the similarity between its own CSI data distribution feature sequence and the first CSI data distribution feature sequence of a first terminal, the similarity discrimination indication message including a model merging process ID and a first CSI data distribution feature sequence, the first CSI data distribution feature sequence being used to characterize the distribution features of the CSI data collected by the first terminal; responding to the similarity discrimination indication message, obtaining the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal; and, if the similarity meets a preset condition, sending a similarity judgment to the network device. The system responds to the similarity judgment response message, which includes the device ID of the second terminal, the model merging process ID, and the similarity between the CSI data distribution feature sequence of the second terminal and the first terminal. It also receives a model upload instruction message from the network device, which instructs the second terminal to report CSI prediction model parameters. This message includes the model merging process ID and a candidate terminal list, obtained by the network device based on the similarity between the CSI data distribution feature sequences of each second terminal and the first terminal. After determining that the second terminal is included in the candidate terminal list, the system sends the model parameters of its local CSI prediction model to the network device. These model parameters enable the network device to obtain the fusion model parameters corresponding to the CSI prediction model.

[0023] In one possible implementation of the third aspect, responding to the similarity discrimination indication message and obtaining the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal includes: responding to the similarity discrimination indication message and obtaining the model merging process ID and the first CSI data distribution feature sequence; obtaining the retention time of the CSI prediction model of the second terminal; if the retention time is greater than a preset retention time threshold, then obtaining the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal. This ensures that the CSI prediction model of the second terminal obtains a sufficient amount of model monitoring data, ensuring that the model is stable before participating in model updates of other terminals, thus guaranteeing the stability of the models participating in the merging.

[0024] In one possible implementation of the third aspect, obtaining the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal includes: calculating the Hamming distance between the CSI data distribution feature sequence of the second terminal and the first CSI data distribution feature sequence of the first terminal, wherein the Hamming distance is negatively correlated with the similarity.

[0025] In one possible implementation of the third aspect, the similarity meets preset conditions, including: the Hamming distance is less than a preset distance threshold.

[0026] Fourthly, this application also provides a communication device, including a processing module and a transceiver module, the communication device being used to perform the method as described in any one of the first to third aspects.

[0027] Fifthly, this application also provides a communication device, comprising: a memory for storing computer instructions; and a processor for executing the computer program or computer instructions stored in the memory, causing the communication device to perform the method as described in any one of the first to third aspects.

[0028] Sixthly, this application also provides a computer storage medium for storing a computer program, which, when executed, implements the method as described in any one of the first to third aspects.

[0029] In a seventh aspect, this application also provides a computer program product, wherein the computer program, when run, causes the methods of any one of the first to third aspects to be performed. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the communication system.

[0031] Figure 2 A flowchart illustrating a wireless communication method provided in an embodiment of this application;

[0032] Figure 3 A flowchart illustrating another wireless communication method provided in an embodiment of this application;

[0033] Figure 4 A flowchart illustrating yet another wireless communication method provided in this application embodiment;

[0034] Figure 5 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0035] Figure 6 This is a schematic diagram of another communication device provided in an embodiment of this application. Detailed Implementation

[0036] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to be a limitation of this application. As used in the specification and appended claims of this application, the singular expressions "a," "an," "the," "the," "the," and "this" are intended to also include expressions such as "one or more," unless the context clearly indicates otherwise. It should also be understood that in the embodiments of this application, "one or more" refers to one, two, or more; "and / or" describes the relationship between related objects, indicating that three relationships may exist; for example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0037] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0038] The "multiple" mentioned in the embodiments of this application refers to two or more. It should be noted that in the description of the embodiments of this application, terms such as "first" and "second" are used only for the purpose of distinguishing descriptions and should not be construed as indicating or implying relative importance, nor should they be construed as indicating or implying order.

[0039] The technical solutions of this application can be applied to various communication systems, such as 3GPP communication systems, fourth-generation (4G) mobile communication technologies, such as Long Term Evolution (LTE) systems, 5G mobile communication systems, 5G New Radio (NR) communication systems, New Radio Vehicle-to-Everything (NRV2X) systems, and also to systems with hybrid LTE and 5G networks, or non-terrestrial network (NTN) systems, device-to-device (D2D) communication systems, machine-to-machine (M2M) communication systems, Internet of Things (IoT) systems, and other next-generation communication systems, such as sixth-generation (6G) communication systems and other communication systems evolving after 5G. These can also be non-3GPP communication systems, and this application does not limit them. The following description uses a 5G communication system (e.g., an NR communication system) as an example.

[0040] To facilitate understanding of the embodiments of this application, Figure 1 The application scenario of this application will be illustrated using the communication system architecture shown below. Figure 1 As shown, the communication system includes network device 101 and terminal device 102. It is understood that... Figure 1 This application only illustrates one possible communication system architecture that can be applied to an embodiment of the present application. In other possible scenarios, the communication system architecture may also include other devices.

[0041] The terminal device 102 can communicate with one or more core networks or the Internet via a radio access network (RAN). The terminal device can also be called user equipment (UE), mobile station (MS), mobile terminal (MT), etc.

[0042] Terminal devices can be mobile terminal devices, such as mobile phones (or "cellular" phones), computers, and data cards. These can be portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile devices that exchange voice and / or data with the wireless access network. Examples include personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), tablets, and computers with wireless transceiver capabilities. Wireless terminal equipment can also be referred to as a system, subscriber unit, subscriber station, mobile station, mobile station (MS), remote station, access point (AP), remote terminal, access terminal, user terminal, user agent, subscriber station (SS), customer premises equipment (CPE), terminal, user equipment (UE), mobile terminal (MT), etc. Terminal equipment can also be wearable devices and next-generation communication systems, such as terminal equipment in 5G communication systems or terminal equipment in future evolved communication systems.

[0043] Network device 101 can be a device in a wireless network. For example, a network device can be a RAN node (or device) that connects terminal devices to the wireless network, and can also be called a base station. Currently, some RAN devices may include: next-generation base stations (gNodeB), transmission reception points (TRP), evolved Node Bs (eNB), radio network controllers (RNC), Node Bs (NB), base station controllers (BSC), base transceiver stations (BTS), home base stations (e.g., home evolved Node B, or home Node B, HNB), base band units (BBU), or wireless fidelity (Wi-Fi) access points, etc., in a network architecture. In addition, in a network structure, network devices may include centralized unit (CU) nodes, distributed unit (DU) nodes, or RAN devices that include both CU and DU nodes.

[0044] In some deployments, a gNB may include a CU and a DU. A gNB may also include a radio unit (RU). The CU implements some of the gNB's functions, and the DU implements others. For example, the CU implements radio resource control (RRC) and packet data convergence protocol (PDCP) layer functions, while the DU implements radio link control (RLC), media access control (MAC), and physical (PHY) layer functions. Since RRC layer information ultimately becomes PHY layer information, or is derived from PHY layer information, in this architecture, higher-layer signaling, such as RRC or PDCP layer signaling, can be considered to be sent by the DU, or by the DU+RU. It is understood that network devices can be CU nodes, DU nodes, or devices including both CU and DU nodes. Furthermore, the CU can be classified as a network device in the access network (RAN) or a network device in the core network (CN); this is not a limitation. In the ORAN system, CU can also be called O-CU, DU can also be called open (O)-DU, CUCP can also be called O-CU-CP, CU-UP can also be called O-CUP-UP, and RU can also be called O-RU.

[0045] Specifically, network devices can send configuration information to terminal devices (e.g., carried in scheduling messages and / or indication messages). The terminal devices then configure their networks based on this information, aligning the network configurations of the network devices and terminal devices. Alternatively, network configurations can be aligned using preset configurations on both the network devices and terminal devices. In essence, "alignment" means that when network devices and terminal devices exchange messages, they have a consistent understanding of the carrier frequency for sending and receiving messages, the type of the message, the meaning of the fields carried in the message, and other configurations of the message.

[0046] Furthermore, in other possible cases, the network device can be any other device that provides wireless communication functionality to the terminal device. The embodiments of this application do not limit the specific technology or device form used in the network device. For ease of description, the embodiments of this application are not limited.

[0047] Network equipment may also include core network equipment, such as access and mobility management function (AMF), user plane function (UPF), or session management function (SMF).

[0048] In this application embodiment, the device for implementing the function of the network device can be the network device itself, or it can be a device capable of supporting the network device in implementing that function, such as a chip system, which can be installed in the network device. The following description uses the example of a network device as the device for implementing the function of the network device to illustrate the technical solution provided in this application embodiment.

[0049] Please see Figure 2 The diagram illustrates a flowchart of a wireless communication method provided in an embodiment of this application, which can be applied to... Figure 1 In the communication system shown, such as Figure 2 As shown, the method may include the following steps:

[0050] S101, the first terminal obtains the first CSI data distribution feature sequence based on the local CSI dataset.

[0051] The first terminal is the terminal device that needs to update the local CSI prediction model. The local CSI dataset includes historical CSI information and / or current CSI information obtained by the first terminal.

[0052] CSI information may include CSI measurement results obtained from CSI measurements (also known as RS resource channel measurement, channel measurement, or pilot channel measurement, etc.) based on the channel state information reference signal (CSI-RS). The CSI measurement results may include at least one of the following: channel response information, channel-quality indicator (CQI), downlink data transmission rank index (RI), and precoding matrix index (PMI).

[0053] The CSI dataset obtained by the first terminal is hash-encoded to obtain a binary encoded sequence of CSI data distribution characteristics, i.e., the first CSI data distribution characteristic sequence. In this way, only the encoded characteristic sequence needs to be transmitted to the network device without transmitting the dataset entity, thereby reducing air interface communication overhead. Moreover, it provides a basis for simplifying the determination of the similarity of CSI dataset distributions between different UEs.

[0054] S102, the first terminal sends a model update request message to the network device.

[0055] The corresponding network device receives the model update request message. The model update request message is used to request available CSI prediction model parameters and may carry the first CSI data distribution feature sequence.

[0056] S103, the network device sends a similarity discrimination indication message to the second terminal.

[0057] The corresponding second terminal receives a similarity determination indication message. The similarity determination indication message is used to instruct the terminal to determine the similarity between the local CSI data distribution characteristics and the CSI data distribution characteristics of the first terminal.

[0058] For example, the similarity discrimination indication message may carry a model merging process ID and a first CSI data distribution feature sequence. The model merging process ID is a unique identifier uniformly assigned by the network device to each model merging process; different model merging processes correspond to different IDs. For example, the network device might be the UE. i The unique identifier assigned to the merging process of the initiated CSI prediction model is ID1; when UEj initiates a CSI prediction model merging request, the network device assigns a unique identifier to the merging process of that model as ID2. The first CSI data distribution feature sequence is used to enable other terminals to compare their local CSI data distribution with that of the first terminal.

[0059] In one exemplary embodiment, the network device may broadcast a similarity determination indication message. The second terminal includes one or more terminals capable of receiving the similarity determination indication message. For example, in a scenario where only one terminal receives the similarity determination indication message, the second terminal includes one terminal; in a scenario where multiple terminals receive the similarity determination indication message, the second terminal includes multiple terminals.

[0060] S104, the second terminal determines the similarity of CSI data distribution based on the CSI data distribution feature sequence corresponding to the local CSI dataset and the first CSI data distribution feature sequence.

[0061] After receiving the similarity determination instruction message, the second terminal parses the message to obtain the model merging process ID and the first CSI data distribution feature sequence. Further, the second terminal performs hash encoding on the local CSI dataset to obtain the corresponding CSI data distribution feature sequence, also known as the second CSI data distribution feature sequence. Finally, the similarity between the second CSI data distribution feature sequence and the first CSI data distribution feature sequence is obtained. For example, the Hamming distance between the two feature sequences can be obtained. If the Hamming distance is greater than a preset threshold, the two CSI data distribution feature sequences are determined to be dissimilar. If the Hamming distance is less than or equal to the preset threshold, the two CSI data distribution feature sequences are determined to be similar.

[0062] S105, the second terminal sends a similarity discrimination response message to the network device.

[0063] The corresponding network device receives a similarity determination response message. This message is used to respond to the similarity determination indication message sent by the network device.

[0064] After determining that its CSI data distribution characteristics are similar to those of the first terminal, the second terminal sends a similarity discrimination response message to the network device. This message may carry the second terminal's device ID, model merging process ID, and the distance between the second and first CSI data distribution characteristic sequences. If the second terminal's CSI data distribution characteristics are determined to be dissimilar to those of the first terminal, no similarity discrimination response message is sent to the network device.

[0065] The similarity of CSI data distribution is determined by calculating the distance of CSI data distribution characteristics among UEs, which simplifies the discrimination process and improves the accuracy of selecting model parameters for model merging.

[0066] S106, the network device determines the list of candidate terminals to participate in model merging based on the similarity discrimination response messages sent by each second terminal.

[0067] All second terminals that receive a similarity discrimination indication message from the network device will determine the similarity between their own CSI data distribution characteristics and the CSI data distribution characteristics of the first terminal. If the CSI data distribution characteristics of the two terminals are determined to be similar, they will send a similarity discrimination response message carrying the distance between the two CSI data distribution characteristic sequences to the network device. In other words, the network device may receive similarity discrimination response messages from multiple second terminals.

[0068] In one exemplary embodiment, the network device may wait to receive a similarity determination response message after sending a similarity determination indication message, and then detect whether a similarity determination response message has been received when the waiting time reaches a time threshold. In this embodiment, the network device receives at least one similarity determination response message sent by a second terminal.

[0069] Furthermore, the network device determines the distance between the CSI data distribution characteristics carried in the response messages based on the similarity of each data point, and selects a preset number of UEs in ascending order to form a candidate terminal list for participating in the model merging.

[0070] In one exemplary embodiment, the preset quantity can be Where N is the number of actual responding UEs. This represents the preset maximum number of responding UEs threshold. If N > Then choose Each responding UE forms a candidate terminal list; if N≤ Then select N responding UEs to form a candidate terminal list.

[0071] In another exemplary embodiment, a ratio can be set between the number of terminals included in the candidate terminal list and the actual number of received similarity discrimination response messages, such as 80% rounded up, and a preset number can be determined based on this ratio and the actual number of responding UEs.

[0072] The candidate terminal list includes the terminal IDs of all terminals participating in the CSI prediction model merging. The terminal ID is a unique identifier for the terminal device, which can be an International Mobile Equipment Identity (IMEI), an International Mobile Subscriber Identity (IMSI), or a Temporary Mobile Subscriber Identity (TMSI), etc.

[0073] S107, the network device sends a model upload instruction message.

[0074] The corresponding second terminal receives a model upload instruction message, which instructs the terminal to report the model parameters of the specified model to the network device.

[0075] For example, a network device can send a model upload instruction message via broadcast. This message may carry a model merging process ID and a list of candidate terminals.

[0076] S108, the second terminal sends a model upload response message to the network device.

[0077] The corresponding network device receives the model upload response message, which is a response to the model upload instruction message.

[0078] Each second terminal that receives the model upload instruction message parses the message to obtain the model merging process ID and a candidate terminal list. Next, the second terminal checks if the model merging process ID carried in the model upload instruction message is the same as the model merging process ID carried in its own similarity judgment response message. If they are the same, it further checks if its own device ID (i.e., the second terminal's ID) is included in the candidate terminal list. If it is, it determines that it needs to respond to the model upload instruction message and further sends a model upload response message containing the model merging process ID and local CSI prediction model parameters to the network device. If it does not contain the model upload instruction message, it determines that it does not need to respond to the model upload instruction message, i.e., it does not need to send a model upload response message to the network device.

[0079] S109, the network device merges the model parameters according to the model parameter sets sent by each second terminal to obtain the fused model parameter set.

[0080] For example, in one possible scenario, the network device receives model parameter sets sent by multiple second terminals. In this scenario, the network device can use a low-complexity, training-free model merging algorithm to merge the model parameters of the CSI prediction models sent by each second terminal to obtain a fused model parameter set.

[0081] In another possible scenario, the network device receives only a set of model parameters sent by a second terminal. In this scenario, the received set of model parameters can be directly used as the fused model parameter set.

[0082] S110, the network device sends a model update instruction message to the first terminal.

[0083] The corresponding first terminal receives a model update instruction message. This message instructs the terminal to update the model parameters of a specified local model. The model update instruction message may carry the model merging process ID and the fused model parameter set.

[0084] S111, the first terminal updates the parameters of the local CSI prediction model according to the fusion model parameter set in the model update instruction message.

[0085] After receiving the model update instruction message, the first terminal parses the instruction message to obtain the model merging process ID and the fusion model parameter set. It then determines whether the model merging process ID in the instruction message is the same as the model merging process ID in the sent model merging request. If they are the same, the local CSI prediction model parameters are updated based on the fusion model parameter set carried in the instruction message to obtain the updated CSI prediction model.

[0086] S112, the first terminal uses the updated model to make CSI predictions and sends the CSI prediction results to the network device.

[0087] The first terminal uses the updated CSI prediction model to perform CSI prediction, obtains the CSI prediction result, and sends it to the network device so that the network device can perform resource scheduling based on the CSI prediction result sent by the terminal.

[0088] The wireless communication method provided in this embodiment utilizes the model parameters of CSI prediction models trained locally by multiple UEs to construct merged model parameters, and sends the merged model parameters to the UEs that need to update their local CSI prediction models. This allows the UEs to directly update their local CSI prediction models using the received merged model parameters without requiring local training. Furthermore, the merged model obtained from the local CSI prediction models of multiple UEs improves the generalization of the CSI prediction model and enhances its versatility across different UEs. Specifically, in the process of constructing the merged model using the local models of multiple UEs, the model parameters of the local models of multiple UEs with similar CSI data distributions are merged to obtain the merged model parameters, thereby ensuring the rationality of the models involved in the merging. Further, when determining the similarity of CSI data distribution between UEs, the local CSI dataset of each UE is encoded to obtain a CSI data distribution feature sequence. The distance between the CSI data distribution feature sequences corresponding to the two UEs is calculated, and the similarity of the CSI data distributions of the two UEs is further determined based on the distance. Encoding the UE's CSI data can significantly reduce the amount of data, thereby reducing communication overhead. Furthermore, using the distance between encoded feature sequences to determine the similarity of data distribution simplifies the judgment process and provides a basis for rationally selecting models for model merging.

[0089] Please see Figure 3 The flowchart illustrates another wireless communication method provided in an embodiment of this application, which is applicable to... Figure 1 The communication system shown, such as Figure 3 As shown, the method may include the following steps:

[0090] S201, the first terminal obtains the first CSI data distribution feature sequence based on the local CSI dataset.

[0091] S202, the first terminal sends a model update request message to the network device.

[0092] S203, the network device sends a similarity discrimination indication message to the second terminal.

[0093] Please refer to the implementation process of S201~S203 in this embodiment. Figure 2The relevant content of S101~S103 will not be repeated here.

[0094] S204, the network device detects that the waiting time has reached the time threshold Twait and no similarity discrimination response message has been received, and sends a model training instruction message to the first terminal.

[0095] After sending a similarity discrimination indication message, the network device waits to receive a similarity discrimination response message. If the waiting time reaches the time threshold and no similarity discrimination response message is received from any terminal, it is considered that there are no model parameters that can be merged.

[0096] In one exemplary embodiment, the network device may maintain a timer. When the network device sends a similarity discrimination indication message, the timer is started and begins counting. When the countdown reaches a time threshold, the timer stops. If the network device does not receive any similarity discrimination response message before the timer stops, it determines that there are currently no model parameters that can be merged, and notifies the first terminal to update the model through model training.

[0097] S205, the first terminal responds to the model training instruction message and performs local CSI prediction model training.

[0098] After receiving the model training instruction message, the first terminal performs model training on the local CSI prediction model to update the model.

[0099] S206, the first terminal performs CSI prediction according to the updated model and sends the CSI prediction results to the network device.

[0100] The wireless communication method provided in this embodiment, after detecting that there are no models that can participate in model merging, notifies the terminal to update the model through local training.

[0101] Please see Figure 4 This illustration shows a flowchart of another wireless communication method provided in an embodiment of this application. Figure 2 The following is a further detailed description based on the illustrated embodiments. Figure 4 As shown, the method may include the following steps:

[0102] S301, the first terminal generates a first CSI data distribution feature Hash encoding sequence based on the local CSI dataset.

[0103] The first terminal is a UE that needs to update its CSI prediction model. For example, when the first terminal detects an event that triggers a model update, it executes the action described in S301. The event that triggers a model update may include an event where the performance monitoring indicators of the CSI prediction model meet preset conditions. For example, if the squared generalized cosine similarity (SGCS cosine similarity) detected by the CSI prediction model is lower than a corresponding preset threshold, or if the normalized mean square error (NMSE) detected by the model is higher than a corresponding preset threshold, then the event that triggers a model update is determined to have occurred.

[0104] In an exemplary embodiment, the process of obtaining the CSI data distribution feature hash sequence is as follows:

[0105] A1. Randomly generate K complex vectors using the standard normal distribution or the Cauchy distribution. These complex vectors are K×N. r ×N t A complex vector of dimension 1 can be denoted as: Where K represents the number of carrier frequencies, and N r N represents the number of antenna ports of the UE. t This indicates the number of antenna ports on the network device; these random complex vectors are shared across all UEs.

[0106] A2, First Terminal (UE) i From local CSI dataset Extract the L most recent channel state tensors in sequence, and convert each channel state tensor into a K×N value. r ×N t The channel state feature is a complex vector of dimension 1, where the transformation formula is as follows:

[0107] (1)

[0108] In formula 1, Indicates UE i The channel state tensor at time t The K×N obtained by conversion r ×N t A complex vector; the function vec() is used to rearrange a matrix into a vector. In Represents K×N r ×N t The set of channel state tensors; S is the total number of channel tensors contained in the dataset (i.e., the total number of times CSI data was collected). This indicates that the data collection time range is the most recent L data collection times.

[0109] A3. Perform an inner product operation on each channel state feature complex vector obtained from A2 with the random complex vector obtained from A1, and convert it to binary bits. The inner product calculation result can be converted to binary bits using the following formula:

[0110] (2)

[0111] In Formula 2 express and The feature complex vectors corresponding to the binary bits obtained after performing inner product calculation. express and The real part of the inner product of the two vectors is greater than 0 (indicating a positive correlation between them), that is, if and If the real part of the inner product is greater than 0, then =1, otherwise =0.

[0112] A4, using the L binary bits obtained from A3, arrange them in the order shown in the following formula to obtain UE. i The CSI data distribution characteristics of the hash-encoded sequence.

[0113] (3)

[0114] in, Replace UE i Local CSI dataset As a UE i This serves as the basis for comparing the similarity of CSI data distribution characteristics with other UEs, thus avoiding the transmission of the entire CSI dataset. This reduces communication overhead and simplifies the algorithm for comparing the similarity of CSI data distribution among UEs.

[0115] In addition, K and L can be flexibly adjusted. The length is adjusted to achieve the required accuracy for describing the similarity of channel data distribution among UEs.

[0116] S302, the first terminal sends a model update request message to the network device.

[0117] The model update request message carries the first terminal (UE) i The CSI data distribution characteristics of the hash-encoded sequence.

[0118] S303, Network device broadcast similarity discrimination indication message.

[0119] The network device receives the model update request message sent by the first terminal and parses it to obtain the model merging process ID and UE carried in the request message. i After hashing the CSI data distribution characteristics into a sequence, a CSI data distribution similarity discrimination indication message is generated and broadcast. This indication message carries the model merging process ID and the UE. i The CSI data distribution characteristics of the hash-encoded sequence.

[0120] S304: Upon receiving the similarity discrimination instruction message, the second terminal determines whether the hold time of the current CSI prediction model is greater than or equal to a preset hold time threshold. If yes, proceed to S305; otherwise, end the current process.

[0121] Setting a preset hold time threshold ensures that the model has undergone a sufficient number of CSI measurement cycles (e.g., one cycle is 20ms) after being put into use, thereby obtaining a sufficient amount of model monitoring data. Only after ensuring the model is stable and usable will it be used to participate in the model update of other UEs, thus ensuring the stability of the models participating in the merging. For example, the preset hold time threshold can be set to more than 100ms, and this application does not impose any restrictions on this.

[0122] After receiving the similarity discrimination instruction message, the second terminal parses the message to determine that it instructs the terminal to perform CSI data distribution feature similarity discrimination. Then, it obtains the hold time of its local CSI prediction model and compares this hold time with a preset hold time threshold. The exemplary hold time of the CSI prediction model refers to the time difference between the current moment and the latest update time of the CSI prediction model.

[0123] In addition, the network device sends a similarity determination indication message via broadcast, so there may be multiple second terminals that receive the message. This step applies to all terminals that receive the message.

[0124] S305, the second terminal determines whether the Hamming distance between its own CSI data distribution feature hash encoding sequence and the first terminal's CSI data distribution feature hash encoding sequence is less than a preset distance threshold. If yes, proceed to S306; otherwise, end the current process.

[0125] After the second terminal determines that the retention time of its local CSI prediction model is greater than or equal to the preset retention time threshold, it calculates the Hamming distance between its own CSI data distribution feature Hash encoding sequence and the first terminal's CSI data distribution feature Hash encoding sequence, and determines whether the Hamming distance is less than the preset distance threshold. If it is, it indicates that the CSI data distribution features of the two terminals are similar, and S306 is executed; otherwise, it indicates that the CSI data distribution features of the two terminals are not similar.

[0126] In an exemplary embodiment, the Hamming distance can be calculated according to the following formulas (4) to (5):

[0127] (4)

[0128] Formula 4 is for and Performing modulo-2 addition (i.e., logical XOR operation) on each bit in the vector yields a new binary complex vector. ; Indicates UE i The corresponding CSI data distribution feature hash encoding sequence, Indicates UE j The corresponding CSI data distribution characteristics hash encoding sequence.

[0129] (5)

[0130] Formula 5 means that The number of "1"s in the UE is used as the UE. i and UE j CSI data distribution characteristics: Hamming distance between hash-coded sequences , express The b-th bit in the equation. The meanings of K and L can be found in the relevant explanation in Formula 1.

[0131] Furthermore, the UE can be determined using the following formula. i and UE j Data distribution similarity:

[0132] (6)

[0133] in, This indicates a preset distance threshold. Indicates UE i and UE j Boolean value for the similarity of CSI data distribution, where if UE i and UE j If the Hamming distance between the hash-encoded sequences of the CSI data distribution characteristics is less than the distance threshold, the corresponding Boolean value is true, and the corresponding numerical value is "1"; if the Hamming distance is greater than or equal to the distance threshold, the corresponding Boolean value is false, and the corresponding numerical value is "0".

[0134] S306, the second terminal sends a similarity discrimination response message to the network device.

[0135] After determining that its CSI data distribution characteristics are similar to those of the first terminal, the second terminal sends a similarity discrimination response message to the network device. The response message carries the device ID of the second terminal, the model merging process ID, and the aforementioned Hamming distance.

[0136] S307, the network device determines whether the waiting time is greater than or equal to the preset waiting time threshold; if so, it executes S308; otherwise, it waits for a certain period of time and then continues to execute S307.

[0137] In one exemplary embodiment, the network device may maintain a timer. When the network device sends a similarity determination indication message, the timer is started to count down. When the countdown reaches a time threshold, the timer stops counting down. If the network device detects that the waiting time has not reached the waiting time threshold, it waits for a certain period of time and then continues to detect whether the waiting time has reached the waiting time threshold. If the network device detects that the waiting time has reached the waiting time threshold, it continues to execute S308.

[0138] For example, the waiting time threshold can be any value within the range of greater than 100ms and less than 500ms. This application does not impose any special restrictions on the specific value of the waiting time threshold.

[0139] S308, the network device determines whether it has received a similarity discrimination response message; if it has received a similarity discrimination response message, it executes S309, otherwise it executes S316.

[0140] After the network device detects that the timer has stopped, it checks whether it has received a similarity discrimination response message from the terminal. If it has received a similarity discrimination response message, it determines that there is a terminal with similar CSI data distribution characteristics to the first terminal and continues to execute S309. If it has not received any similarity discrimination response message from any terminal, it determines that there is no terminal with similar CSI data distribution characteristics to the first terminal and continues to execute S316.

[0141] S309, the network device selects a preset number of second terminals to obtain a candidate terminal list in order of increasing Hamming distance.

[0142] After receiving similarity discrimination response messages from multiple terminals, the network device obtains the model merging process ID and Hamming distance carried in the response message. If the model merging process ID carried in the response message is the same as the model merging process ID in the similarity discrimination indication message, the Hamming distances are sorted in ascending order, and a preset number of second terminals are selected to form a candidate terminal list, which includes the device ID of each selected second terminal.

[0143] In one exemplary embodiment, the preset quantity can be Where N is the number of UEs that actually responded. This represents the preset maximum number of responding UEs threshold. If N > Then choose Each responding UE forms a candidate terminal list; if N≤ Then select N responding UEs to form a candidate terminal list.

[0144] S310, Network device broadcasts model upload instruction message.

[0145] After determining the candidate terminal list, the network device broadcasts a model upload instruction message, which instructs the terminals to upload model parameters. This message may carry a model merging process ID and the candidate terminal list.

[0146] S311, the second terminal determines whether it needs to respond to the model upload instruction message; if so, it executes S311, otherwise it ends the current process.

[0147] After receiving the model upload instruction message broadcast by the network device, the second terminal parses the instruction message to obtain the model merging process ID and the candidate terminal list. Next, the second terminal determines whether the model merging process ID in the model upload instruction message is the same as the model merging process ID carried in its own similarity judgment response message. If they are different, no response to the model upload instruction message is needed; if they are the same, the second terminal further checks whether its own device ID is included in the candidate terminal list. If it is, a response to the model upload instruction message is required; otherwise, no response is needed.

[0148] S312, the second terminal sends the model parameters of the local CSI prediction model to the network device.

[0149] After the second terminal determines the model upload instruction message that needs to be received, it sends the model merging process ID and CSI prediction model parameters to the network device. .

[0150] S313, The network device executes a low-complexity, no-training-required model merging algorithm to obtain merged model parameters from the received CSI prediction model parameters.

[0151] In one exemplary embodiment, the main idea of ​​the low-complexity, training-free model merging algorithm includes:

[0152] B1. The average model parameters for each response UE are obtained using the arithmetic mean method. Here, the responding UE refers to the terminal that responds to the model upload instruction message sent by the network device.

[0153] For example, obtaining the average model parameters The formula is as follows:

[0154] (7)

[0155] in, This indicates the number of UEs included in the candidate terminal list. For the meaning of this, please refer to the relevant content in S309.

[0156] B2, calculate the average variance between the model parameters and the average model parameters for each response UE using the following formula ( );

[0157] (8)

[0158] in, This represents the average variance between the model parameters of the i-th UE and the average model parameters; This represents the number of parameters in the model parameter set. It is assumed that the models of each UE are isomorphic, that is, the number of model parameters of each UE is the same. Represents the model parameter set The first in One parameter, Represents the first in the average model parameter set One parameter.

[0159] B3. The weighting coefficients are obtained by normalizing the negative exponent of the average variance of the model parameters of each response UE using the following formula. ( ):

[0160] (9)

[0161] in, The negative exponent represents the average variance of the model parameters for the i-th UE. This indicates the number of UEs in the candidate terminal list.

[0162] B4. The model parameters are weighted using the following formula by applying the weight coefficients corresponding to each candidate terminal to obtain the fusion model parameters. :

[0163] (10)

[0164] in, This represents the model parameter set for the i-th UE. This represents the weight coefficients of the model parameters for the i-th UE. This indicates the number of UEs in the candidate terminal list.

[0165] S314, the network device sends a model update instruction message to the first terminal.

[0166] Network devices obtain fusion model parameters from multiple candidate UEs. Subsequently, a model update instruction message is sent to the terminal that initiated the model merging request (i.e., the first terminal in this embodiment). This message is used to instruct the UE to update the UE's local model parameters. This message may carry the model merging process ID and the fused model parameters. .

[0167] S315, the first terminal updates the model parameters of the local CSI prediction model based on the fusion model parameters.

[0168] After receiving the model update instruction message, the first terminal parses the message to obtain the model merging process ID and the fused model parameters. Update the model parameters in the local CSI prediction model to the fusion model parameters. , thus obtaining the updated model.

[0169] S316, the network device sends a model training instruction message to the first terminal.

[0170] If the network device determines in S308 that it has not received any similarity discrimination response message from any terminal, it sends a model training instruction message to the terminal that initiated the model update request. This message is used to instruct the terminal to perform model training and update the model parameters locally. This message may carry the model merging process ID.

[0171] S317, the first terminal performs local CSI prediction model training and updates model parameters.

[0172] After receiving the model training instruction message, the first terminal performs model training on its local CSI prediction model to update the model parameters.

[0173] S318, the first terminal uses the updated CSI prediction model to make CSI predictions and sends the CSI prediction results to the network device.

[0174] After updating the CSI prediction model using the fusion model parameters at the first terminal, or after training and updating the model locally, the updated model will be used to make CSI predictions, and the CSI prediction results will be sent to the network device.

[0175] In an exemplary embodiment, the model update request message, similarity discrimination response message, and model upload response message sent by the terminal can be carried through fields in the RRC layer uplink signaling. For example, newly defined fields (or reused existing fields) in the RRC layer CSI reporting configuration signaling RRC CSI-ReportConfig can carry these messages. The network device identifies the specific message content by parsing the corresponding fields in the RRC CSI-ReportConfig signaling.

[0176] In an exemplary embodiment, the similarity discrimination indication message, model upload indication message, model update indication message, and model training indication message sent by the aforementioned network device can carry these messages through newly defined fields (or reused existing ones) in the RRC layer downlink signaling. For example, newly defined fields in the RRC layer CSI resource configuration signaling RRC CSI-ResourceConfig can carry these messages. The terminal identifies the specific message content by parsing the corresponding fields in RRC CSI-ResourceConfig.

[0177] Corresponding to the above-described wireless communication method embodiments, this application also provides communication device embodiments.

[0178] Figure 5 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application. The communication device may be a terminal device, a device in a terminal device (such as a chip), or a device that can be used in conjunction with a terminal device; or the communication device may be a network device, a device in a network device (such as a chip), or a device that can be used in conjunction with a network device.

[0179] like Figure 5 As shown, the communication device may include a transceiver module 110 and a processing module 120. Specifically, the processing module 120 is used to process data, which may be data received by the transceiver module 110, and the processed data may also be sent by the transceiver module 110.

[0180] The processing module 120 is used to execute the processing flow of the terminal device or network device in the above-described communication method embodiment.

[0181] In one implementation, the communication device may correspond to the UE in the above method embodiments and may be used to execute the various steps and / or processes executed by the UE in the above method embodiments.

[0182] In another implementation, the communication device may correspond to the network device in the above method embodiments and may be used to execute the various steps and / or processes executed by the network device in the above method embodiments.

[0183] For other possible implementations of the communication device, please refer to the relevant descriptions of the terminal device or network device functions mentioned above, which will not be repeated here.

[0184] Figure 6 This is another schematic block diagram of the communication device provided in the embodiments of this application.

[0185] The communication device can be a UE, a network device, a chip, chip system, or processor that implements the above methods. This communication device can be used to implement the methods described in the above method embodiments; please refer to the descriptions in the above method embodiments for details.

[0186] like Figure 6 As shown, the communication device may include one or more processors 201, which may also be referred to as processing units or processing modules, and can implement certain control functions. The processor 201 can be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, while the central processing unit can be used to control the communication device (e.g., base station, baseband chip, user, user chip), execute software programs, and process data from the software programs.

[0187] In an alternative design, the processor 201 may also store instructions and / or data, which can be executed by the processor 201 to cause the communication device to perform the methods described in the above method embodiments.

[0188] In another alternative design, the communication device may include a communication interface 202 for implementing receiving and transmitting functions. For example, the communication interface 202 may be a transceiver circuit, interface, interface circuit, or transceiver. The transceiver circuit, interface, interface circuit, or transceiver for implementing receiving and transmitting functions may be separate or integrated. The aforementioned transceiver circuit, interface, interface circuit, or transceiver may be used for reading and writing code / data, or it may be used for transmitting or relaying signals.

[0189] Optionally, the communication device may include one or more memories 203, which may store instructions that can be executed on the processor 201, causing the communication device to perform the methods described in the above method embodiments. Optionally, the memories 203 may also store data. Optionally, the processor 201 may also store instructions and / or data. The processor 201 and the memories 203 may be provided separately or integrated together.

[0190] It should be understood that, in one possible design, the steps in the method embodiments provided in this application can be implemented by integrated logic circuits in the processor's hardware or by instructions in software form. The steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are not provided here.

[0191] In one implementation, the communication device can correspond to the UE in the above method embodiments and can be used to execute the various steps and / or processes executed by the UE in the above method embodiments. The processor 201 can be used to execute instructions stored in the memory 203, and when the processor 201 executes the instructions stored in the memory, the processor 201 is used to execute the various steps and / or processes of the above method embodiments corresponding to the terminal.

[0192] In another implementation, the communication device may correspond to the network device in the above method embodiments and may be used to execute the various steps and / or processes executed by the network device in the above method embodiments. The processor 201 may be used to execute instructions stored in the memory 203, and when the processor 201 executes the instructions stored in the memory, the processor 201 is used to execute the various steps and / or processes of the above method embodiments corresponding to the network device.

[0193] It is understood that the aforementioned processor can be one or more chips. For example, the processor can be a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a system-on-chip (SoC), a central processor unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips.

[0194] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0195] This application also provides a computer-readable storage medium storing instructions that, when executed on one or more computing devices, cause the one or more computing devices to perform the wireless communication method described in the above embodiments.

[0196] Computer-readable storage media can be non-transitory computer-readable storage media, such as read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage devices.

[0197] This application also provides a computer program product. When executed by one or more computing devices, the computer program product enables the computing devices to execute any of the aforementioned wireless communication methods. The computer program product can be a software installation package. When any of the aforementioned wireless communication methods needs to be used, the computer program product can be downloaded and executed on a computer.

[0198] This application also provides a processor, including: an input circuit, an output circuit, and a processing circuit. The processing circuit receives signals through the input circuit and transmits signals through the output circuit, causing the processor to execute the wireless communication method described in the above embodiments.

[0199] In specific implementation, the processor can be one or more chips, the input circuit can be input pins, the output circuit can be output pins, and the processing circuit can be transistors, gate circuits, flip-flops, and various logic circuits. The input signal received by the input circuit can be received and input by, for example, but not limited to, a receiver, and the signal output by the output circuit can be output to, for example, but not limited to, a transmitter and transmitted by the transmitter. Furthermore, the input circuit and the output circuit can be the same circuit, which is used as the input circuit and the output circuit at different times. This application does not limit the specific implementation of the processor and various circuits.

[0200] This application also provides a chip system including one or more processors for calling and executing instructions stored in a memory, thereby causing the wireless communication method described in the above embodiments to be executed. The chip system may be composed of a chip or may include chips and other discrete devices. The chip system may include input circuitry or interfaces for transmitting information or data, and output circuitry or interfaces for receiving information or data.

[0201] In the embodiments of this application, the terms and English abbreviations are exemplary examples given for ease of description and should not be construed as limiting the application in any way. This application does not preclude the possibility of defining other terms that can achieve the same or similar functions in existing or future agreements.

[0202] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated.

[0203] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0204] It should be understood that in the various embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0205] In summary, the above description is merely a preferred embodiment of the technical solution of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A wireless communication method, characterized in that, Applied to a first terminal, the method includes: Send a model update request message to the network device. The model update request message is used to request an update to the CSI prediction model parameters of the first terminal. The model update request message includes a first CSI data distribution feature sequence, which is used to characterize the distribution features of the CSI data collected by the first terminal. The network device receives a model update instruction message, which instructs the first terminal to update the model parameters of the CSI prediction model. The model update instruction message includes fusion model parameters corresponding to the CSI prediction model. The fusion model parameters are obtained by the network device by fusing the model parameters of the CSI prediction models of each terminal in the candidate terminal list. The candidate terminal list is obtained by the network device based on the similarity between the CSI data distribution characteristics sent by each terminal and the first CSI data distribution characteristic sequence. The candidate terminal list includes at least one second terminal, and the CSI data distribution characteristics of each second terminal are similar to the first CSI data distribution characteristic sequence. In response to the model update instruction message, the model parameters of the local CSI prediction model of the first terminal are updated directly based on the fusion model parameters carried in the model update instruction message, without the need for local training.

2. The method according to claim 1, characterized in that, The process of obtaining the first CSI data distribution feature sequence includes: The obtained CSI data is hash-encoded to obtain the distribution feature sequence of the first CSI data.

3. The method according to claim 2, characterized in that, The step of hash-encoding the obtained CSI data to obtain the first CSI data distribution feature sequence includes: Generate K random complex vectors, where K represents the carrier frequency; L most recently obtained CSI tensors are extracted from the CSI dataset obtained from the first terminal and converted into CSI feature complex vectors respectively. The random complex vectors have the same dimension as the CSI feature complex vectors. Each of the CSI feature complex vectors is subjected to an inner product operation with K random complex vectors to obtain K×L inner product operation results, and each inner product operation result is converted into binary bits; The first CSI data distribution feature sequence is obtained by arranging K×L binary bits in sequence.

4. The method according to claim 3, characterized in that The step of performing inner product operations on each of the CSI feature complex vectors with K random complex vectors to obtain K×L inner product results, and converting each inner product result into binary bits, includes: Convert the result of each inner product operation to binary bits using the following formula: in, Represents a random complex vector The k-th complex vector in Indicates UE i The CSI feature complex vector obtained by CSI tensor transformation at time t; express and The binary bits obtained after performing the inner product calculation express and The real part of the result of the inner product operation is greater than 0; Let L represent the most recent L acquisition times, and S represent the total number of CSI tensors contained in the CSI dataset.

5. The method according to claim 4, characterized in that, The step of arranging K×L binary bits in sequence to obtain the first CSI data distribution feature sequence includes: Arrange according to the following formula: in, Indicates UE i The CSI data distribution feature sequence corresponding to the CSI dataset.

6. A wireless communication method, characterized in that, Applied to network devices, the method includes: The system receives a model update request message sent by a first terminal. The model update request message is used to request an update to the model parameters of the CSI prediction model of the first terminal. The model update request message includes a first CSI data distribution feature sequence, which is used to characterize the distribution features of the CSI data collected by the first terminal. A similarity discrimination indication message is sent to the second terminal. The similarity discrimination indication message is used to instruct the second terminal to determine the similarity between its own CSI data distribution characteristics and the first CSI data distribution characteristic sequence of the first terminal. The similarity discrimination indication message includes the model merging process ID and the first CSI data distribution characteristic sequence. The system receives a similarity discrimination response message sent by the second terminal. The similarity discrimination response message includes the device ID of the second terminal, the model merging process ID, and the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal. Based on the similarity in each of the similarity judgment response messages, a list of candidate terminals to participate in model merging is determined, and the list of candidate terminals includes the device IDs corresponding to the candidate terminals; Send a model upload instruction message, which is used to instruct the second terminal to report CSI prediction model parameters. The model upload instruction message includes the model merging process ID and the candidate terminal list. Receive a model upload response message, the model upload response message including the model parameters of the CSI prediction model in the second terminal; The model parameters sent by each second terminal are fused to obtain fused model parameters; A model update instruction message is sent to the first terminal. The model update instruction message is used to instruct the first terminal to update the model parameters of the CSI prediction model. The model update instruction message includes the model merging process ID and the fusion model parameters.

7. The method according to claim 6, characterized in that, The method further includes: If no similarity judgment response message is received within a preset waiting time threshold after the similarity discrimination indication message is sent, a model training indication message is sent to the first terminal. The model training indication message is used to instruct the first terminal to perform local training and update for the CSI prediction model. The model training indication message includes the model merging process ID.

8. The method according to claim 6 or 7, characterized in that, The step of determining the candidate terminal list for model merging based on the similarity of the response messages includes: A candidate terminal list is obtained by selecting a preset number of second terminals in descending order of similarity.

9. The method according to claim 8, characterized in that, The step of selecting a preset number of second terminals in descending order of similarity to obtain a candidate terminal list includes: A candidate terminal list is obtained by selecting a preset number of second terminals in ascending order of the distance between the CSI data distribution feature sequence of each second terminal and the first CSI data distribution feature sequence.

10. The method according to claim 9, characterized in that, The step of selecting a preset number of second terminals to obtain a candidate terminal list, based on the ascending order of the distance between the CSI data distribution feature sequence of each second terminal and the first CSI data distribution feature sequence, includes: Determine the first number of second terminals that send the similarity discrimination response message; If the first number is less than a preset number threshold, then all second terminals that send the similarity discrimination response message are determined to form the candidate terminal list; If the first quantity is greater than the preset quantity threshold, then the preset quantity threshold of second terminals are selected from the first quantity to form the candidate terminal list in ascending order of the distance between the CSI data distribution feature sequence of each second terminal and the first CSI data distribution feature sequence.

11. The method according to claim 6, characterized in that, The process of fusing the model parameters sent by each second terminal to obtain fused model parameters includes: The arithmetic mean method is used to obtain the average model parameters corresponding to the model parameters sent by each second terminal, and the average variance value between each received model parameter and the average model parameter is obtained. Normalize the negative exponents of each mean variance value to obtain the weight coefficients corresponding to each received model parameter; The fusion model parameters are obtained by weighted summation of each received model parameter.

12. A wireless communication method, characterized in that, Applied to a second terminal, the method includes: The second terminal receives a similarity discrimination indication message sent by a network device. The similarity discrimination indication message is used to instruct the second terminal to determine the similarity between its own CSI data distribution feature sequence and the first CSI data distribution feature sequence of the first terminal. The similarity discrimination indication message includes a model merging process ID and the first CSI data distribution feature sequence, which is used to characterize the distribution features of the CSI data collected by the first terminal. In response to the similarity discrimination indication message, the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal is obtained; If the similarity meets the preset conditions, a similarity discrimination response message is sent to the network device. The similarity discrimination response message includes the device ID of the second terminal, the model merging process ID, and the similarity between the CSI data distribution feature sequence of the second terminal and the first terminal. The network device receives a model upload instruction message, which instructs the second terminal to report CSI prediction model parameters. The model upload instruction message includes the model merging process ID and a candidate terminal list. The candidate terminal list is obtained by the network device based on the similarity of the CSI data distribution feature sequences of each second terminal and the first terminal. After determining that the second terminal is included in the candidate terminal list, the model parameters of the local CSI prediction model are sent to the network device. The model parameters are used to enable the network device to obtain the fusion model parameters corresponding to the CSI prediction model.

13. The method according to claim 12, characterized in that, The step of responding to the similarity discrimination indication message and obtaining the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal includes: In response to the similarity discrimination indication message, obtain the model merging process ID and the first CSI data distribution feature sequence; The retention time of the prediction model of the second terminal is obtained. If the retention time is greater than a preset retention time threshold, the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal is obtained.

14. The method according to claim 12, characterized in that, The step of obtaining the similarity between the CSI data distribution feature sequences of the second terminal and the first terminal includes: Calculate the Hamming distance between the CSI data distribution feature sequence of the second terminal and the first CSI data distribution feature sequence of the first terminal, wherein the Hamming distance is negatively correlated with the similarity.

15. The method according to claim 14, characterized in that, The similarity meets preset conditions, including: the Hamming distance is less than a preset distance threshold.

16. A communication device, characterized in that, The communication device includes a processing module and a transceiver module, and is used to perform the method as described in any one of claims 1 to 15.

17. A communication device, characterized in that, include: Memory, used to store computer instructions; A processor for executing a computer program or computer instructions stored in the memory, causing the communication device to perform the method as described in any one of claims 1 to 15.

Images