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

Abstract

This disclosure relates to a communication method, communication device, communication system, storage medium, and program product. The method includes: performing bilateral AI model performance testing on a first AI model and a second AI model, generating performance testing results, wherein the first AI model is a Channel State Information (CSI) recovery part model of a network device, and the second AI model is a CSI generation part model of a terminal; based on the performance testing results, sending first information to the terminal, the first information being used by the terminal to determine a third AI model, the third AI model being a CSI generation part model used for inference in the terminal. This reduces the complexity of bilateral model training collaboration and ensures the model inference performance of the bilateral AI models.

Description

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

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

[0002] Among related technologies, employing AI (Artificial Intelligence) technology can reduce terminal feedback overhead or improve CSI (Channel State Information) feedback accuracy, and this has been carried out in 3GPP (3rd Generation Partnership Project) standardization research. A bilateral AI / ML model based on a terminal-side CSI generation model and a network-side CSI recovery model can respectively achieve compressed feedback and recovery of CSI.

[0003] Summary of the Invention

[0004] To overcome the technical problem of poor inference performance of AI models in related technologies, this disclosure provides a communication method, communication device, communication system, storage medium, and program product.

[0005] According to a first aspect of the embodiments of this disclosure, a communication method is provided, performed by a network device, the method comprising:

[0006] A bilateral AI model performance test is performed on the first AI model and the second AI model to generate a performance test result. The first AI model is the Channel State Information (CSI) recovery part model of the network device, and the second AI model is the CSI generation part model of the terminal.

[0007] Based on the performance test results, a first message is sent to the terminal. The first message is used by the terminal to determine a third AI model, which is the CSI generation part model used for inference in the terminal.

[0008] According to a second aspect of the embodiments of this disclosure, a communication method is provided, executed by a terminal, the method comprising:

[0009] The network device receives first information sent by the network device, the first information being generated by the network device based on performance testing results, the performance testing results being generated by the network device after performing performance testing on a bilateral AI model of a first AI model and a second AI model, the first AI model being a CSI recovery part model of the network device, and the second AI model being a CSI generation part model of the terminal.

[0010] Based on the first information, a third AI model is determined, which is the CSI generation part model used for inference in the terminal.

[0011] According to a third aspect of the present disclosure, a communication device is provided for performing the communication method of any one of the first or second aspects of the present disclosure.

[0012] According to a fourth aspect of the present disclosure, a storage medium is provided that stores instructions that, when executed on a communication device, cause the communication device to perform a communication method as described in either the first or second aspect of the present disclosure.

[0013] According to a fifth aspect of the present disclosure, a program product is provided, comprising at least one of a program and instructions, wherein when the program or instructions are executed by a communication device, they implement the steps of the communication method described in the first aspect of the present disclosure, or when the program or instructions are executed by a communication device, they implement the steps of the communication method described in the second aspect of the present disclosure.

[0014] By adopting the above technical solution, at least the following beneficial technical effects can be achieved:

[0015] Both the first and second AI models undergo bilateral AI model performance testing, generating performance test results. The first AI model is the Channel State Information (CSI) recovery part model of the network device, and the second AI model is the CSI generation part model of the terminal. Based on the performance test results, first information is sent to the terminal. This first information is used by the terminal to determine the third AI model, which is the CSI generation part model used for inference in the terminal. This reduces the complexity of bilateral model training collaboration and ensures the model inference performance of the bilateral AI models. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings required for the description of the embodiments are introduced below. The following drawings are only some embodiments of this disclosure and do not impose specific limitations on the protection scope of this disclosure.

[0017] Figure 1A is a schematic diagram of the architecture of a communication system according to an embodiment of the present disclosure.

[0018] Figure 1B is a schematic diagram illustrating the implementation of CSI compressed feedback and recovery based on a bilateral AI / ML model according to an embodiment of the present disclosure.

[0019] Figure 2A is an interactive schematic diagram of a communication method according to an embodiment of the present disclosure.

[0020] Figure 2B is a schematic diagram of the bilateral model collaborative training process according to an embodiment of the present disclosure.

[0021] Figure 3A is an interactive schematic diagram of a communication method according to an embodiment of the present disclosure.

[0022] Figure 3B is an interactive schematic diagram of a communication method according to an embodiment of the present disclosure.

[0023] Figure 4 is a schematic diagram of the structure of a terminal according to an embodiment of the present disclosure.

[0024] Figure 5 is a schematic diagram of the structure of a network device according to an embodiment of the present disclosure.

[0025] Figure 6A is a schematic diagram of the structure of the communication device 6100 proposed in an embodiment of this disclosure.

[0026] Figure 6B is a schematic diagram of the structure of the chip 6200 proposed in the embodiment of this disclosure. Detailed Implementation

[0027] This disclosure provides a communication method, communication device, communication system, storage medium, and program product.

[0028] In a first aspect, embodiments of this disclosure provide a communication method executed by a network device, the method comprising:

[0029] A bilateral AI model performance test is performed on the first AI model and the second AI model to generate a performance test result. The first AI model is the Channel State Information (CSI) recovery part model of the network device, and the second AI model is the CSI generation part model of the terminal.

[0030] Based on the performance test results, a first message is sent to the terminal. The first message is used by the terminal to determine a third AI model, which is the CSI generation part model used for inference in the terminal.

[0031] In the above embodiments, the network device determines the inference AI model by instructing the terminal through the first information, thereby reducing the complexity of bilateral model training collaboration and ensuring the model inference performance of the bilateral AI model.

[0032] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes:

[0033] The second AI model was determined to be the reference AI model;

[0034] The first AI model is generated by training based on the second AI model and the first dataset.

[0035] In the above embodiments, training the counterpart AI model based on the reference AI model can improve the model performance of the counterpart AI model and ensure the consistency and accuracy of the inference results of the two-sided AI models.

[0036] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0037] Based on the performance test results, it is determined that the set performance requirements are met;

[0038] The first information is sent to the terminal, and the first information is also used to indicate that the third AI model is the second AI model.

[0039] In the above embodiments, when it is determined that the current AI model meets the set performance requirements, the network device instructs the terminal to use the current AI model for inference, which simplifies the collaborative training process of the two-sided AI models and improves the training efficiency of the AI ​​model.

[0040] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0041] Based on the performance test results, it is determined that the set performance requirements are not met;

[0042] Based on the first AI model and the second dataset, a fourth AI model is trained and generated, which is a CSI-generated partial model.

[0043] According to the fourth AI model, the first information is sent to the terminal. The first information includes the first model parameters of the fourth AI model. The first model parameters are used by the terminal to update and set the AI ​​model to generate the third AI model, or the first model parameters are used by the terminal to train and generate the third AI model based on the first model parameters and the third dataset.

[0044] In the above embodiments, the bilateral AI model is trained by combining a reference model, model parameter passing, and model training, taking advantage of the advantages of each method to ensure the model inference performance of the AI ​​model.

[0045] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes:

[0046] The second model parameters of the second AI model are sent to the terminal. The second model parameters are used by the terminal to update and set the AI ​​model and generate the second AI model.

[0047] In the above embodiments, the bilateral AI model is trained by passing model parameters, thereby reducing the complexity of collaborative training of the bilateral AI model, enhancing the robustness of the bilateral AI model, and ensuring the model inference performance of the AI ​​model.

[0048] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0049] Based on the performance test results, it is determined that the set performance requirements are not met;

[0050] The first information is sent to the terminal, and the first information is also used to instruct the terminal to train and generate the third AI model based on the second model parameters and the third dataset.

[0051] In the above embodiments, training the bilateral AI model using model training and model parameter passing can keep the model at a relatively low computational cost and ensure the model's feature enhancement and generalization ability.

[0052] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0053] Based on the performance test results, it is determined that the set performance requirements are not met;

[0054] Based on the first AI model and the second dataset, a fourth AI model is trained and generated, which is a CSI-generated partial model.

[0055] According to the fourth AI model, the first information is sent to the terminal. The first information includes the fifth dataset of the fourth AI model. The fifth dataset is used by the terminal to update and set the AI ​​model to generate the third AI model, or the fifth dataset is used by the terminal to train the model to generate the third AI model. The fifth dataset includes target CSI and CSI feedback.

[0056] In the above embodiments, the use of target CSI and CSI feedback can ensure the recovery accuracy of the bilateral AI model and improve the model inference performance and efficiency of the bilateral AI model.

[0057] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes:

[0058] The first AI model is determined to be the reference AI model;

[0059] The second AI model is trained and generated based on the first AI model and the second dataset.

[0060] In the above embodiments, training the counterpart AI model based on the reference AI model can improve the model performance of the counterpart AI model and ensure the consistency and accuracy of the inference results of the two-sided AI models.

[0061] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0062] Based on the performance test results, it is determined that the set performance requirements are met;

[0063] Based on the second AI model, the first information is sent to the terminal. The first information includes the second model parameters of the second AI model. The second model parameters are used by the terminal to update and set the AI ​​model to generate the third AI model.

[0064] In the above embodiments, the bilateral AI model is trained by passing model parameters, thereby reducing the complexity of collaborative training of the bilateral AI model, enhancing the robustness of the bilateral AI model, and ensuring the model inference performance of the AI ​​model.

[0065] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0066] Based on the performance test results, it is determined that the set performance requirements are not met;

[0067] According to the second AI model, the first information is sent to the terminal. The first information includes the second model parameters of the second AI model. The second model parameters are used by the terminal to train and generate the third AI model based on the second model parameters and the third dataset.

[0068] In the above embodiments, training the bilateral AI model using model training and model parameter passing can keep the model at a relatively low computational cost and ensure the model's feature enhancement and generalization ability.

[0069] In conjunction with some embodiments of the first aspect, in some embodiments, sending the first information to the terminal based on the performance detection result includes:

[0070] Based on the performance test results, it is determined that the set performance requirements are not met;

[0071] According to the second AI model, the first information is sent to the terminal. The first information includes a sixth dataset of the second AI model. The sixth dataset is used by the terminal to update and set the AI ​​model to generate the third AI model. Alternatively, the sixth dataset is used by the terminal to train the model to generate the third AI model. The sixth dataset includes target CSI and CSI feedback.

[0072] In the above embodiments, the use of target CSI and CSI feedback can ensure the recovery accuracy of the bilateral AI model and improve the model inference performance and efficiency of the bilateral AI model.

[0073] In conjunction with some embodiments of the first aspect, in some embodiments, the third dataset includes at least one of the following:

[0074] The dataset sent by the network device;

[0075] The dataset measured by the terminal.

[0076] In the above embodiments, the data type of the third dataset can include various types. The terminal trains the inference AI model based on the third dataset, thereby ensuring the consistency between the terminal-side inference AI model and the network-side inference AI model and improving the model inference performance of the bilateral AI model.

[0077] Secondly, embodiments of this disclosure provide a communication method executed by a terminal, the method comprising:

[0078] The network device receives first information sent by the network device, the first information being generated by the network device based on performance testing results, the performance testing results being generated by the network device after performing bilateral AI model performance testing on a first AI model and a second AI model, the first AI model being the CSI recovery part model of the network device, and the second AI model being the CSI generation part model of the terminal.

[0079] Based on the first information, a third AI model is determined, which is the CSI generation part model used for inference in the terminal.

[0080] In conjunction with some embodiments of the second aspect, in some embodiments, determining the third AI model based on the first information includes:

[0081] The second AI model was determined to be the reference AI model;

[0082] Based on the first information, the third AI model is determined to be the second AI model.

[0083] In conjunction with some embodiments of the second aspect, in some embodiments, the first information includes first model parameters of a fourth AI model, wherein the fourth AI model is generated by the network device based on the first AI model and a second dataset, and determining the third AI model based on the first information includes:

[0084] The AI ​​model is updated and configured based on the parameters of the first model to generate the third AI model; or...

[0085] The third AI model is trained and generated based on the parameters of the first model and the third dataset.

[0086] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes:

[0087] Receive the second model parameters sent by the network device, wherein the second model parameters are the model parameters of the second AI model;

[0088] The AI ​​model is updated and set according to the parameters of the second model, and the second AI model is generated.

[0089] In conjunction with some embodiments of the second aspect, in some embodiments, determining the third AI model based on the first information includes:

[0090] Based on the first information, determine the third dataset;

[0091] The third AI model is trained and generated based on the second model parameters and the third dataset.

[0092] In conjunction with some embodiments of the second aspect, in some embodiments, the first information includes a fifth dataset of the fourth AI model, the fifth dataset including target CSI and CSI feedback, the fourth AI model being trained and generated by the network device based on the first AI model and the second dataset, and determining the third AI model based on the first information includes:

[0093] The AI ​​model is updated and configured based on the fifth dataset to generate the third AI model; or...

[0094] The third AI model is generated by training the model based on the fifth dataset.

[0095] In conjunction with some embodiments of the second aspect, in some embodiments, the first information includes the second model parameters of the second AI model, and the step of determining the third AI model based on the first information includes:

[0096] The first AI model is determined to be the reference AI model;

[0097] Based on the second model parameters, the AI ​​model is updated and set to generate the third AI model.

[0098] In conjunction with some embodiments of the second aspect, in some embodiments, the first information includes the second model parameters of the second AI model, and the step of determining the third AI model based on the first information includes:

[0099] The first AI model is determined to be the reference AI model;

[0100] The third AI model is trained and generated based on the second model parameters and the third dataset.

[0101] In conjunction with some embodiments of the second aspect, in some embodiments, the first information includes a sixth dataset of the second AI model, the sixth dataset including target CSI and CSI feedback, and the step of determining the third AI model based on the first information includes:

[0102] Based on the sixth dataset, update the configured AI model to generate the third AI model; or...

[0103] The third AI model is generated by training the model based on the sixth dataset.

[0104] In conjunction with some embodiments of the second aspect, in some embodiments, the third dataset includes at least one of the following:

[0105] The dataset sent by the network device;

[0106] The dataset measured by the terminal.

[0107] Thirdly, embodiments of this disclosure provide a communication device for performing the communication method described in any one of the first aspects of this disclosure, or the communication method described in any one of the second aspects of this disclosure.

[0108] Fourthly, embodiments of this disclosure provide a communication system including a terminal and a network device, wherein the terminal is configured to implement the communication method described in any one of the first aspects of this disclosure, and the network device is configured to implement the communication method described in any one of the second aspects of this disclosure.

[0109] Fifthly, embodiments of this disclosure provide a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform a communication method as described in any one of the first aspects of this disclosure, or a communication method as described in any one of the second aspects of this disclosure.

[0110] In a sixth aspect, embodiments of this disclosure provide a program product comprising at least one of a program and instructions, wherein when the program or instructions are executed by a communication device, they implement the steps of the communication method described in the first aspect of this disclosure, or when the program or instructions are executed by a communication device, they implement the steps of the communication method described in the second aspect of this disclosure.

[0111] In a seventh aspect, embodiments of this disclosure provide a terminal, which includes at least one of a transceiver module and a processing module; wherein the terminal is used to execute an optional implementation of the first aspect.

[0112] Eighthly, embodiments of this disclosure provide a network device, which includes at least one of a transceiver module and a processing module; wherein the core network device is used to perform an optional implementation of the second aspect.

[0113] In a ninth aspect, embodiments of this disclosure provide a terminal, which includes one or more processors; wherein the terminal is used to execute an optional implementation of the first aspect.

[0114] In a tenth aspect, embodiments of this disclosure provide a network device comprising: one or more processors; wherein the core network device is configured to perform an optional implementation of the second aspect.

[0115] In one aspect, embodiments of this disclosure provide a communication system comprising: a terminal and a network device; wherein the terminal is configured to perform the method described in the optional implementation of the first aspect, and the network device is configured to perform the method described in the optional implementation of the second aspect.

[0116] In a twelfth aspect, embodiments of this disclosure provide a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform the method as described in the optional implementations of the first and second aspects.

[0117] In a thirteenth aspect, embodiments of this disclosure provide a program product that, when executed by a communication device, causes the communication device to perform the method as described in the optional implementations of the first and second aspects.

[0118] In a fourteenth aspect, embodiments of this disclosure provide a computer program that, when run on a computer, causes the computer to perform the methods described in alternative implementations of the first and second aspects.

[0119] In a fifteenth aspect, embodiments of this disclosure provide a chip or chip system. The chip or chip system includes processing circuitry configured to perform the methods described according to optional implementations of the first and second aspects above.

[0120] It is understood that the aforementioned communication equipment, communication system, storage medium, program product, etc., are all used to execute the methods proposed in the embodiments of this disclosure. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0121] This disclosure provides a communication method, communication device, communication system, storage medium, and program product. In some embodiments, terms such as information processing method and communication method may be used interchangeably.

[0122] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments. In all embodiments of this disclosure, unless otherwise specified or logically conflicting, the terminology and / or descriptions between the embodiments are consistent and can be mutually referenced. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0123] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0124] In this embodiment of the disclosure, unless otherwise stated, elements expressed in the singular form, such as "a," "an," "the," "the," "the," "the," "the," "the," "this," etc., can mean "one and only one," or "one or more," "at least one," etc. For example, when using articles such as "a," "an," "the," etc. in translation, the noun following the article can be understood as either a singular expression or a plural expression.

[0125] In the embodiments of this disclosure, "multiple" refers to two or more.

[0126] In some embodiments, the terms "at least one of A or B, at least one of A and B", "one or more", "a plurality of", "multiple" and the like can be used interchangeably.

[0127] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "in response to one case A, in response to another case B", etc., may include the following technical solutions depending on the situation: in some embodiments, A (execute A regardless of whether there is a branch B); in some embodiments, B (execute B regardless of whether there is a branch A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, both A and B are executed. The same applies when there are more branches such as A, B, C, etc.

[0128] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execute A regardless of whether a branch B exists); in some embodiments, B (execute B regardless of whether a branch A exists); in some embodiments, execution is selected from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, and C.

[0129] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not impose restrictions on the position, order, priority, quantity, or content of the descriptive objects. The description of the descriptive objects is found in the claims or the context of the embodiments, and the use of prefixes should not constitute unnecessary restrictions. For example, if the descriptive object is a "field," the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is a "level," the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the number of descriptive objects is not limited by ordinal numbers and can be one or more. For example, in "first device," the number of "devices" can be one or more. Furthermore, the objects modified by different prefixes can be the same or different. For example, if the object being described is "device", then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. Similarly, if the object being described is "information", then "first information" and "second information" can be the same information or different information, and their content can be the same or different.

[0130] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0131] In some embodiments, terms such as "time / frequency" and "time-frequency domain" refer to the time domain and / or frequency domain.

[0132] In some embodiments, terms such as “in response to…”, “in response to determining…”, “in the case of…”, “when…”, “when…”, “if…”, etc. can be used interchangeably. These descriptions all refer to the device making a corresponding action under certain objective circumstances. They do not necessarily limit the time, nor do they require the device to make a judgment action when implementing it, nor do they mean that there must be other limitations.

[0133] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0134] In some embodiments, devices, etc., may be interpreted as physical or virtual, and their names are not limited to those described in the embodiments. Terms such as “device,” “equipment,” “circuit,” “network element,” “network function,” “network device,” “function,” “node,” “unit,” “section,” “system,” “network,” “chip,” “chip system,” “entity,” and “subject” are interchangeable.

[0135] In some embodiments, "network" can be interpreted as devices included in a network (e.g., access network devices, core network devices, etc.).

[0136] In some embodiments, the terms "access network device (AN device)," "radio access network device (RAN device)," "base station (BS)," "radio base station," "fixed station," "node," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "antenna panel," "antenna array," "cell," "macro cell," "small cell," "femto cell," "pico cell," "sector," "cell group," "serving cell," "carrier," "component carrier," and "bandwidth part (BWP)" can be used interchangeably.

[0137] In some embodiments, the terms "terminal", "terminal device", "user equipment (UE)", "user terminal", "mobile station (MS)", "mobile terminal (MT)", "subscriber station", "mobile unit", "subscriber unit", "wireless unit", "remote unit", "mobile device", "wireless device", "wireless communication device", "remote device", "mobile subscriber station", "access terminal", "mobile terminal", "wireless terminal", "remote terminal", "handset", "user agent", "mobile client", and "client" can be used interchangeably.

[0138] In some embodiments, access network devices, core network devices, or network devices can be replaced by terminals. For example, embodiments of this disclosure can also be applied to structures where communication between access network devices, core network devices, or network devices and terminals is replaced by communication between multiple terminals (e.g., device-to-device (D2D), vehicle-to-everything (V2X), etc.). In this case, the structure can also be configured such that the terminal has all or part of the functions of the access network device. Furthermore, terms such as "uplink" and "downlink" can be replaced with terms corresponding to communication between terminals (e.g., "sidelink"). For example, uplink channel, downlink channel, etc., can be replaced with sidelink channel, and uplink link, downlink, etc., can be replaced with sidelink link.

[0139] In some embodiments, the terminal may be replaced by an access network device, a core network device, or a network device. In this case, the access network device, core network device, or network device may also be configured to have all or some of the functions of the terminal.

[0140] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0141] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0142] Furthermore, each element, each row, or each column in the table of this disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0143] Figure 1A is a schematic diagram of the architecture of a communication system according to an embodiment of the present disclosure. As shown in Figure 1A, the communication system 100 includes a terminal 101 and a network device 102.

[0144] In some embodiments, terminal 101 includes, for example, at least one of the following: mobile phone, wearable device, Internet of Things device, car with communication function, smart car, tablet computer, computer with wireless transceiver function, virtual reality (VR) terminal device, augmented reality (AR) terminal device, wireless terminal device in industrial control, wireless terminal device in self-driving, wireless terminal device in remote medical surgery, wireless terminal device in smart grid, wireless terminal device in transportation safety, wireless terminal device in smart city, and wireless terminal device in smart home, but is not limited thereto.

[0145] In some embodiments, network device 102 may be a node or device that connects a terminal to a wireless network. The network device may include at least one of the following in a 5G communication system: evolved Node B (eNB), next-generation eNB (ng-eNB), next-generation Node B (gNB), node B (NB), home node B (HNB), home evolved node B (HeNB), wireless backhaul device, radio network controller (RNC), base station controller (BSC), base transceiver station (BTS), base band unit (BBU), mobile switching center, base station in a 6G communication system, open RAN, cloud RAN, base station in other communication systems, and access node in a Wi-Fi system, but is not limited thereto.

[0146] In some embodiments, the technical solutions of this disclosure can be applied to the Open RAN architecture. In this case, the interfaces between or within network devices involved in the embodiments of this disclosure can be transformed into internal interfaces of Open RAN. The processes and information interactions between these internal interfaces can be implemented by software or programs.

[0147] In some embodiments, a network device may be composed of a central unit (CU) and a distributed unit (DU). The CU may also be called a control unit. The CU-DU structure can separate the protocol layer of the network device. Some of the protocol layer functions are centrally controlled by the CU, while the remaining part or all of the protocol layer functions are distributed in the DU, which is centrally controlled by the CU. However, this is not the only possibility.

[0148] It is understood that the communication system described in this disclosure is for the purpose of more clearly illustrating the technical solutions of this disclosure, and does not constitute a limitation on the technical solutions proposed in this disclosure. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions proposed in this disclosure are also applicable to similar technical problems.

[0149] The following embodiments of this disclosure can be applied to the communication system 100 shown in FIG1A, or to some of the main bodies, but are not limited thereto. The main bodies shown in FIG1A are illustrative. The communication system may include all or some of the main bodies in FIG1A, or it may include other main bodies outside of FIG1A. The number and form of each main body are arbitrary. Each main body may be physical or virtual. The connection relationship between the main bodies is illustrative. The main bodies may not be connected or may be connected. The connection can be in any way, it can be a direct connection or an indirect connection, it can be a wired connection or a wireless connection.

[0150] The embodiments disclosed herein can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G new radio (NR), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New radio access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20, Ultra-Wideband (UWB), Bluetooth (a registered trademark), Public Land Mobile Network (PLMN) networks, Device-to-Device (D2D) systems, Machine-to-Machine (M2M) systems, Internet of Things (IoT) systems, Vehicle-to-Everything (V2X) systems, systems utilizing other communication methods, and next-generation systems built upon them, etc. Furthermore, multiple systems can be combined (e.g., a combination of LTE or LTE-A with 5G).

[0151] Figure 1B is a schematic diagram illustrating CSI compression feedback and recovery based on a bilateral AI / ML model according to an embodiment of this disclosure. As shown in Figure 1B, the UE side generates a partial model (defined as an encoder) through CSI, compresses the downlink channel information H, and sends it to the gNB after quantization into a binary bit stream. The gNB side recovers the partial model (defined as a decoder) through CSI to recover H', which is approximately the same as the original downlink channel information.

[0152] In some embodiments, the CSI generation and recovery partial models need to be trained using a collected dataset. The types of training methods for the CSI generation and recovery partial models include the following three:

[0153] (1) The model is trained on one side (e.g., the terminal side or the network side), and then the trained part of the model is sent to the other side. For example, on the network side, a part of the model is recovered based on the standard CSI and the dataset, and a training CSI generated part of the model is trained and sent to the terminal side.

[0154] (2) Train the CSI generation part model and the CSI recovery part model on the terminal side and the network side respectively through joint training. Alternatively, after training part of the model on one end of the terminal side or the network side, train the other part of the bilateral model on the other end, wherein the parameters of the model that was trained first are not updated;

[0155] (3) First, the model is trained on one side, and then the training data or other auxiliary information is sent to the other side to train another part of the model. For example, this training method can be divided into: ① NW (Network) training first: The network side first trains the CSI generation part model and the CSI recovery part model, and then sends the dataset for training the CSI generation part model and / or other auxiliary information to the terminal side; ② UE training first: The terminal side first trains the CSI generation part model and the CSI recovery part model, and then sends the dataset for training the CSI recovery part model and / or other auxiliary information to the network side.

[0156] In some embodiments, to reduce or mitigate the training complexity of bilateral models, the following options for collaborative training between devices are proposed:

[0157] Method 1: Standardize the reference model structure and parameters;

[0158] Method 2: Dataset standardization;

[0159] Method 3: Standardize the model structure, and the model parameters are passed between the NW side and the UE side;

[0160] Method 4: Standardize the data format, and the data is transmitted between the NW side and the UE side;

[0161] Method 5: Standardize the model format, while the reference model is passed between the NW side and the UE side.

[0162] Method 1 can be divided into: (1) Standardized reference encoder; (2) Standardized reference decoder; (3) Standardized reference encoder and Standardized reference decoder.

[0163] In some embodiments, based on the model parameters transmitted to the UE or the behavior performed by the UE after modeling, methods 3 and 5 can be further classified as follows:

[0164] Method 3: (1) Method 3a: Receive the model parameters and then train the model to develop a different model or test.

[0165] Example, method 3a-1: Passing the encoder parameter;

[0166] Method 3a-2: Passing parameters to the decoder;

[0167] Method 3a-3: Pass the parameters for the encoder and the parameters for the decoder.

[0168] (2) Method 3b: The received model parameters are directly used for model inference.

[0169] Method 5 includes Method 5a, which involves training the received model to develop a different model or for testing. Example: Method 5a-1: Passing the encoder model;

[0170] Method 5a-2: Passing the decoder model;

[0171] Method 5a-3: Pass the encoder and decoder models.

[0172] In some embodiments, depending on the content of the transmitted data, method 4 can be divided into:

[0173] Method 4-1: The dataset consists of target CSI and CSI feedback;

[0174] Method 4-2: The dataset consists of the feedback CSI and the reconstructed target CSI;

[0175] Method 4-3: The dataset consists of the target CSI, the feedback CSI, and the recovered target CSI.

[0176] In some embodiments, considering factors such as bilateral collaboration performance, complexity, feasibility, and whether to expose private information on the NW side or UE side, the above-mentioned methods 1, 3a-1, 4-1, and 3b can be used as preferred candidate bilateral AI model collaborative training methods.

[0177] In some embodiments, for method 1 above, it is necessary to standardize one or both of the reference CSI generation partial model and the reference CSI recovery partial model. Because the AI ​​model and its parameters are already standardized, their flexibility and inference performance are limited.

[0178] For method 3a-1 above, although the model structure is standardized, the model parameters can be further updated through model training, offering some flexibility and performance improvement compared to method 1. However, its flexibility is limited by the standardized model structure, and retraining the model also introduces some latency.

[0179] In method 3b above, the model parameters received by the UE can be directly used for inference without retraining the model. However, if the data distribution characteristics on the UE side are different from the data characteristics used to train the AI ​​model on the NW side, directly using the received model for inference will affect inference performance.

[0180] Regarding method 4-1 above, this option is not constrained by the model structure and can flexibly train the model based on the dataset, achieving a certain level of inference performance. However, the UE side needs to receive a large dataset sent by the NW side for model training, which will consume a significant amount of wireless transmission resources.

[0181] Since the different bilateral AI model collaborative training options mentioned above have different advantages or disadvantages, this embodiment proposes to combine different bilateral AI model collaborative training options for bilateral model training. That is, the model is trained by combining standardized models, model parameter transfer and model training, so as to make full use of the advantages of each option and achieve better AI model inference performance with lower collaborative training complexity.

[0182] Figure 2A is an interactive schematic diagram of a communication method according to an embodiment of the present disclosure. As shown in Figure 2A, the embodiments of the present disclosure relate to a communication method, which includes:

[0183] In step S2101, the network device performs bilateral AI model performance testing on the first AI model and the second AI model, and generates performance testing results.

[0184] For example, network devices and terminals perform model inference based on a bilateral AI model. The terminal is configured with a second AI model, and the network device is configured with a first AI model. In this embodiment, the bilateral AI model is used for CSI prediction and CSI inference. For instance, the terminal inputs the measured downlink channel information H into the second AI model, which performs CSI prediction or CSI inference and sends the inference result to the network device. The network device inputs this inference result into the first AI model, and uses the first AI model to recover the downlink channel information, thereby recovering H', which approximates the downlink channel information.

[0185] In this model, the first AI model and the second AI model are corresponding AI models. The downlink channel information needs to be compressed based on the second AI model, and then the first AI model uses this compressed information to recover the approximate downlink channel information. Therefore, to ensure the consistency and accuracy of the model inference results, the first AI model and the second AI model must correspond to each other. For example, the opposing AI model can be trained based on a single-sided AI model. Optionally, the same type of AI model can be configured in both the network device and the terminal. For example, a standardized CSI recovery model and a standardized CSI generation model can be configured in the network device and the terminal respectively to ensure the model performance of the two-sided AI model.

[0186] In some embodiments, the first AI model is a CSI recovery part model for the network device, and the second AI model is a CSI generation part model for the terminal. The CSI generation part model is used by the terminal to acquire CSI through actual wireless channel measurements, and the CSI recovery part model is used by the network device to reconstruct and recover the terminal's channel state after receiving the CSI feedback from the terminal. For example, in this embodiment, an encoder model is used to represent the CSI generation part model, and a decoder model is used to represent the CSI recovery part model.

[0187] For example, a network device can perform performance checks on a first AI model and a second AI model based on a bilateral AI model performance testing standard. The network device can be configured with both a first and a second AI model, collect relevant training data, input the training data into the second AI model, input the second AI model's output into the first AI model, and compare the first AI model's output with the corresponding output dataset from the training dataset to obtain the bilateral AI model's performance testing result. Optionally, the bilateral AI model's performance testing can also be completed based on the interaction between the network device and the terminal. For example, the terminal sends the target CSI input of the second AI model and the output of the second AI model to the network device, which then inputs the second AI model's output into the first AI model, generates the first AI model's output, and compares this output with the target CSI to determine the bilateral AI model's performance and generate a performance testing result.

[0188] In some embodiments, the performance test result is used to indicate the performance of a bilateral AI model consisting of a first AI model and a second AI model. The network device is configured with performance metrics to measure this performance test result. To ensure the performance of the bilateral AI model, the performance test result must meet the performance metrics. If the performance test result is determined to meet the performance metrics, the network device sends an instruction to the terminal, instructing the terminal to perform model inference based on the currently configured second AI model, while the network device simultaneously performs model inference based on the first AI model. If the performance test result is determined not to meet the performance metrics, the network device instructs the terminal to update the currently configured AI model using methods such as model training, model parameter updating, model transfer, or data transfer, so that the updated AI model meets the performance metrics.

[0189] Optionally, in some embodiments, before step S2101 described above, the method further includes:

[0190] The network device identifies the second AI model as the reference AI model;

[0191] The network device trains and generates the first AI model based on the second AI model and the first dataset.

[0192] In some embodiments, the second AI model is a standardized reference AI model. The network device can identify the second AI model configured in the terminal based on information such as the model identifier and model ID of the second AI model. Once the second AI model is determined to be a standardized reference AI model, the network device derives the corresponding decoder model in the network device based on the standardized reference AI model and the first dataset, and trains it to generate the first AI model on the network device side.

[0193] Optionally, in some embodiments, before step S2101 described above, the method further includes:

[0194] The network device determines the first AI model as the reference AI model;

[0195] The network device trains and generates a second AI model based on the first AI model and the second dataset.

[0196] In some embodiments, the first AI model configured in the network device can also be set as a standardized reference AI model. That is, the network device trains and generates a second AI model on the terminal side based on a standardized decoder model and a second dataset. This second AI model is an encoder model. After generating the second AI model, bilateral AI model performance testing is performed based on the first AI model and the second AI model.

[0197] In step S2102, the network device sends the first information to the terminal based on the performance test results.

[0198] In some embodiments, the terminal receives first information.

[0199] In some embodiments, the first information is used to instruct the terminal to determine a third AI model based on the first information.

[0200] In some embodiments, the third AI model is the CSI generation part model that the terminal will use for inference.

[0201] In some embodiments, after generating the performance test results of the bilateral AI model based on the above steps, the network device compares the performance test results with the configured performance indicator requirements to determine whether the currently obtained performance check results meet the performance indicator requirements.

[0202] For example, if the current performance test result meets the performance requirements, the network device instructs the terminal to perform model inference based on the first information and the currently configured second AI model. For instance, if the first AI model is a standardized decoder model, the first information includes the model parameters of the second AI model. The terminal updates the preset AI model according to these model parameters to obtain the second AI model and performs model inference based on it. If the second AI model is a standardized encoder model, the first information is an instruction, and the terminal uses the first AI model to perform model inference based on this instruction.

[0203] Optionally, in some embodiments, step S2102 above includes:

[0204] Based on the performance test results, the network device determines that it meets the set performance requirements.

[0205] The network device sends the first information to the terminal, and the first information is also used to indicate that the third AI model is the second AI model.

[0206] For example, in this embodiment, the second AI model is a standardized encoder model. Based on performance testing results, when the network device determines that the current first AI model and the second AI model meet the set performance requirements, it sends first information to the terminal. This first information is indicative information, and can also be used to instruct the terminal to use the second AI model for model inference.

[0207] Optionally, in some embodiments, step S2102 above includes:

[0208] Based on the performance test results, the network device determines that it meets the set performance requirements;

[0209] The network device sends first information to the terminal based on the second AI model. The first information includes the second model parameters of the second AI model. The second model parameters are used by the terminal to update and set the AI ​​model and generate the third AI model.

[0210] For example, in this embodiment, the first AI model is a standardized decoder model. Based on performance testing results, the network device determines when the current first and second AI models meet the set performance requirements. The network device then sends the second model parameters of the trained second AI model to the terminal. The terminal updates and sets the AI ​​model according to the second model parameters, generating a third AI model. At this time, the third AI model generated in the terminal is the second AI model.

[0211] It should be noted that the AI ​​model configured in the terminal can be a standardized reference AI model or a pre-trained AI model. The parameters of this reference AI model are updated based on the parameters of the second model to obtain the third AI model used for inference.

[0212] Optionally, in some embodiments, step S2102 above includes:

[0213] Based on the performance test results, the network device determines that it does not meet the set performance requirements.

[0214] The network device trains and generates a fourth AI model based on the first AI model and the second dataset.

[0215] The network device sends the first information to the terminal based on the fourth AI model.

[0216] For example, when a network device determines that the set performance requirements are not met based on performance testing results, it continues to train and generate a fourth AI model that meets the performance requirements, based on the first AI model and the second dataset. This fourth AI model is an encoder model. Based on the fourth AI model, the device sends first information to the terminal. This first information may include first model parameters corresponding to the fourth AI model. The terminal updates the set AI model based on these first model parameters to obtain the fourth AI model, and then uses this fourth AI model as the third AI model for model inference. Optionally, the first information may include the fourth AI model itself, instructing the terminal to use the fourth AI model as the third AI model for model inference.

[0217] For example, the first information includes the first model parameters of the fourth AI model. The terminal updates and sets the AI ​​model based on the first model parameters to generate the third AI model. Optionally, the terminal can also train and generate the third AI model based on the first model parameters and the third dataset.

[0218] For example, (1) set the standardized reference AI model as encoder (defined as encoder#1).

[0219] Step S1: During the training phase, the NW side trains the decoder#1 model based on the collected dataset and the standardized encoder#1. The trained decoder#1 model can meet the predefined performance requirements.

[0220] Step S2: The UE and NW use encoder#1 and decoder#1 for inference respectively. The NW uses a model performance monitoring method to perform bilateral AI model performance testing. For example, the UE sends the target CSI input to encoder#1 and the output of encoder#1 to the NW. The NW performs performance testing based on the target CSI and the output of encoder#1. If the performance of encoder#1 is found to be insufficient to meet the performance requirements, and the UE has the capability to receive model parameters, the NW updates the reference encoder#1 model parameters based on the collected dataset to obtain encoder#2, and sends the model parameters of encoder#2 to the UE.

[0221] Step S3: During the inference phase, the UE performs inference based on encoder #1 and the NW-side decoder #1 (which is based on training). Alternatively, the UE performs inference based on encoder #2 and the NW-side decoder #1 (which is based on training).

[0222] The dataset collected by the NW side can be field data reported by the UE side, or it can be field datasets reported by other users or data generated by the NW based on simulation assumptions, or it can include both the above-mentioned field data measured by the UE and the data generated by simulation assumptions, or field datasets reported by other users.

[0223] (2) Define the standardized reference decoder model as decoder#1, and standardize the encoder model structure as well.

[0224] During the model training phase: The NW side trains encoder #2 based on decoder #1 and the collected dataset to meet the performance requirements. Then, the model parameters of encoder #2 are sent to the UE.

[0225] During the model inference phase: the UE uses encoder#2 for model inference, while the NW side uses decoder#1 for model inference.

[0226] Optionally, in some embodiments, the third dataset includes at least one of the following:

[0227] Data sets sent by network devices;

[0228] Data set of terminal measurements.

[0229] For example, the third dataset configured in the terminal could be a dataset sent from the network device to the terminal, used to train the encoder model. It could also be a dataset collected from measurements within the terminal.

[0230] For example, (1) set a standardized reference encoder, defined as encoder#1.

[0231] Figure 2B is a schematic diagram of the bilateral model collaborative training process according to an embodiment of this disclosure. As shown in Figure 2B, step S1: During the training phase, the NW side trains a decoder#2 model based on the collected dataset and the standardized encoder#1. The trained decoder#2 model can meet the predefined performance requirements. Then, the model parameters of encoder#1 are sent to the UE.

[0232] Step S2: The UE and NW use encoder#1 and decoder#2 for inference respectively. The NW uses a model performance monitoring method to perform bilateral AI model performance testing. For example, the UE sends the target CSI input to encoder#1 and the output of encoder#1 to the NW. The NW performs performance testing based on the target CSI and the output of encoder#1. If the NW detects that the performance of encoder#1 cannot meet the performance index requirements, and the UE has the ability to receive model parameters, then the NW updates the reference encoder#1 model parameters based on the collected dataset to obtain encoder#2, and sends the model parameters of encoder#2 to the UE.

[0233] Step S3: The UE and NW use encoder#2 and decoder#1 for inference respectively. Similar to step S2 above, the NW monitors the performance of the bilateral models. If the performance requirements are met, then the bilateral models on the UE side and the NW side are encoder#2 and decoder#2 respectively. Otherwise, the NW instructs the UE to update encoder#1 or encoder#2 to obtain encoder#3, or to train and generate a new model encoder#3.

[0234] Step S4: During the inference phase, the UE uses one of the encoders (encoder#1, encoder#2, and encoder#3) that meets the performance requirements of the two-sided model for inference, while the NW side performs inference based on the trained decoder#2.

[0235] In some embodiments, if the above steps do not include the encoder#2 step, the UE does not use the NW-transmitted encoder#2 for inference, but instead retrains based on the above method 3a-1 to obtain encoder#3.

[0236] (2) Set a standardized reference decoder, defined as decoder#1, and standardize the encoder model structure of the reference model.

[0237] The training method is the same as in the above embodiment. The NW uses decoder#1 for model inference. The UE performs model inference based on the received encoder#2, or the UE retrains or updates encoder#3 based on the received encoder#2 and performs model inference based on encoder#3.

[0238] For example, the NW side is set to have already trained encoder#1 and decoder#1 models based on the collected dataset and the standardized reference encoder model structure.

[0239] Step S1: The NW side passes the model parameters of the trained encoder#1 to the UE.

[0240] Step S2: The UE and NW respectively use encoder#1 and decoder#1 for inference. The NW can monitor the performance of the bilateral model based on the input and output data of the encoder#1 model sent by the UE. If the monitored bilateral model meets the performance requirements, the NW instructs the UE to use encoder#1 for inference through one or more of RRC (Radio Resource Control), MAC-CE (Media Access Control-Control Element), or DCI signaling. Otherwise, the NW can also instruct the UE to retrain encoder#2 that meets the performance requirements based on the received encoder#1 through one or more of RRC, MAC-CE, or DCI signaling. Optionally, the performance requirement indicators can be predefined or sent to the UE by the NW.

[0241] Step S3: The UE trains encoder #2 based on the collected training dataset and encoder #1. The UE or NW monitors the performance of encoder #2. If encoder #2 meets the performance requirements, the UE and NW use encoder #2 and decoder #1 for inference, respectively.

[0242] Optionally, in some embodiments, step S2102 above includes:

[0243] Based on the performance test results, the network device determines that it does not meet the set performance requirements.

[0244] The network device trains and generates a fourth AI model based on the first AI model and the second dataset. The fourth AI model is a partial model generated by CSI.

[0245] The network device sends the first information to the terminal based on the fourth AI model. The first information includes the fifth dataset of the fourth AI model.

[0246] For example, the fifth dataset includes the target CSI and CSI feedback. When the performance detection results of the bilateral AI model do not meet the set performance requirements, the network device memorizes the first AI model and the second dataset, and trains to generate a fourth AI model. The method for generating the fourth AI model is the same as in the above embodiments, and can be referred to the above embodiments, so it will not be repeated here. The network device sends first information to the terminal based on the fourth AI model, and the first information includes the fifth dataset. The terminal can update the model parameters of the set AI model based on the fifth dataset to generate a third AI model. Optionally, the terminal can also retrain and generate a new AI model based on the fifth dataset to obtain the third AI model.

[0247] In step S2103, the terminal determines the third AI model based on the first information.

[0248] For example, the first information could be instruction information, instructing the terminal to use the second AI model as the third AI model. Optionally, the first information could be model parameter information, which the terminal uses to update and set the model parameters of the AI ​​model, generating the third AI model. The terminal could also use the model parameter information and the third AI model to train a model and generate the third AI model. The first information could also be the target CSI and CSI feedback, which the terminal uses to update and set the model parameters of the AI ​​model, generating the third AI model.

[0249] In some embodiments, the names of information, etc., are not limited to the names described in the embodiments. Terms such as "information", "message", "signal", "signaling", "report", "configuration", "indication", "instruction", "command", "channel", "parameter", "domain", "field", "symbol", "symbol", "codebook", "codeword", "codepoint", "bit", "data", "program", and "chip" can be used interchangeably.

[0250] In some embodiments, the terms "codebook," "codeword," and "precoding matrix" can be used interchangeably. For example, a codebook can be a collection of one or more codewords / precoding matrices.

[0251] In some embodiments, the terms "uplink", "uplink", and "physical uplink" can be used interchangeably, as can the terms "downlink", "downlink", and "physical downlink", as well as the terms "sidelink", "sidelink", "sidelink communication", "sidelink communication", "direct connection", "direct link", "direct communication", and "direct link communication".

[0252] In some embodiments, the terms “downlink control information (DCI),” “downlink (DL) assignment,” “DL DCI,” “uplink (UL) grant,” and “UL DCI” can be used interchangeably.

[0253] In some embodiments, terms such as "physical downlink shared channel (PDSCH)" and "DL data" can be used interchangeably, as can terms such as "physical uplink shared channel (PUSCH)" and "UL data".

[0254] In some embodiments, the terms “radio”, “wireless”, “radio access network (RAN)”, “access network (AN)”, and “RAN-based” can be used interchangeably.

[0255] In some embodiments, the terms "search space", "search space set", "search space configuration", "search space set configuration", "control resource set (CORESET)", and "CORESET configuration" can be used interchangeably.

[0256] In some embodiments, the terms "synchronization signal (SS)," "synchronization signal block (SSB)," "reference signal (RS)," "pilot," and "pilot signal" can be used interchangeably.

[0257] In some embodiments, terms such as “moment,” “point in time,” “time,” and “time location” can be used interchangeably, as can terms such as “duration,” “segment,” “time window,” “window,” and “time.”

[0258] In some embodiments, the terms "component carrier (CC)," "cell," "frequency carrier," and "carrier frequency" can be used interchangeably.

[0259] In some embodiments, the terms “resource block (RB)”, “physical resource block (PRB)”, “sub-carrier group (SCG)”, “resource element group (REG)”, “PRB pair”, “RB pair”, “resource element (RE)”, and “sub-carrier” can be used interchangeably.

[0260] In some embodiments, terms such as wireless access scheme and waveform can be used interchangeably.

[0261] In some embodiments, the terms "precoding", "precoder", "weight", "precoding weight", "quasi-co-location (QCL)", "transmission configuration indication (TCI) status", "spatial relation", "spatial domain filter", "transmission power", "phase rotation", "antenna port", "antenna port group", "layer", "the number of layers", "rank", "resource", "resource set", "resource group", "beam", "beam width", "beam angular degree", "antenna", "antenna element", and "panel" can be used interchangeably.

[0262] In some embodiments, the terms “frame”, “radio frame”, “subframe”, “slot”, “sub-slot”, “mini-slot”, “symbol”, “symbol”, and “transmission time interval (TTI)” can be used interchangeably.

[0263] In some embodiments, "acquire," "get," "obtain," "receive," "transmit," "bidirectional transmission," and "send and / or receive" can be used interchangeably and can be interpreted as receiving from other entities, acquiring from protocols, acquiring from higher layers, obtaining through self-processing, or autonomous implementation. Protocols include, for example, at least one of the 3GPP protocol, Wi-Fi protocol, and audio and / or video protocols.

[0264] In some embodiments, terms such as “send,” “transmit,” “report,” “distribute,” “transfer,” “bidirectional transmission,” “send and / or receive” can be used interchangeably.

[0265] In some embodiments, terms such as "certain," "preset," "default," "set," "indicated," "a certain," "any," and "first" can be used interchangeably. "Certain A," "preset A," "default A," "set A," "indicated A," "a certain A," "any A," and "first A" can be interpreted as A pre-defined in a protocol or the like, or as A obtained through setting, configuration, or indication, or as specific A, a certain A, any A, or first A, but are not limited thereto.

[0266] In some embodiments, the determination or judgment can be made by a value represented by 1 bit (0 or 1), or by a true or false value (boolean), or by a comparison of numerical values ​​(e.g., a comparison with a predetermined value), but is not limited thereto.

[0267] In some embodiments, "not expecting to receive" can be interpreted as not receiving on time domain resources and / or frequency domain resources, or as not performing subsequent processing on the data and / or instructions received; "not expecting to send" can be interpreted as not sending, or as sending but not expecting the receiver to respond to the sent content.

[0268] In some embodiments, if an arrow in the interaction diagram representing the sending of information, signaling, etc. from one subject to another passes through other subjects, it can be interpreted as the information being forwarded from one subject to another via other subjects, or it can be interpreted as the information being sent from one subject to another without passing through other subjects.

[0269] The communication method involved in the embodiments of this disclosure may include at least one of steps S2101-S2103. For example, step S2101 may be implemented as a standalone embodiment, step S2102 may be implemented as a standalone embodiment, step S2101+S2103 may be implemented as a standalone embodiment, step S2101+S2102 may be implemented as a standalone embodiment, and step S2102+S2103 may be implemented as a standalone embodiment, but is not limited thereto.

[0270] In some embodiments, steps S2101, S2102, and S2103 may be performed in an alternate order or simultaneously.

[0271] In some embodiments, step S2101 is optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0272] In some embodiments, step S2102 is optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0273] In some embodiments, the steps and their optional implementations in other embodiments described before or after this embodiment, as well as other related parts in the specification, can be referred to, and will not be repeated here.

[0274] Figure 3A is an interactive schematic diagram of a communication method according to an embodiment of the present disclosure. As shown in Figure 3A, the embodiments of the present disclosure relate to a communication method, which includes:

[0275] Step S3101: The network device trains and generates the first AI model based on the second AI model and the first dataset.

[0276] The optional implementation of step S3101 can be found in the optional implementation of step S2101 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0277] In step S3102, the network device performs bilateral AI model performance testing on the first AI model and the second AI model, and generates performance testing results.

[0278] The optional implementation of step S3102 can be found in the optional implementation of step S2101 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0279] In step S3103, the network device determines that the set performance requirements are met based on the performance test results and sends the first information to the terminal.

[0280] In some embodiments, the first information is instruction information, which can also be used to instruct the terminal to use a second AI model for model inference.

[0281] The optional implementation of step S3103 can be found in the optional implementation of step S2102 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0282] In step S3104, the network device determines that the set performance requirements are not met based on the performance test results, and trains and generates a fourth AI model based on the first AI model and the second dataset.

[0283] The optional implementation of step S3104 can be found in the optional implementation of step S2102 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0284] In step S3105, the network device sends the first information to the terminal based on the fourth AI model.

[0285] In some embodiments, the first model parameters are used by the terminal to update and set the AI ​​model according to the first model parameters to generate a third AI model, or the first model parameters are used by the terminal to train and generate a third AI model according to the first model parameters and a third dataset.

[0286] The optional implementation of step S3105 can be found in the optional implementation of step S2102 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0287] In step S3106, the terminal determines the third AI model based on the first information.

[0288] The optional implementation of step S3106 can be found in the optional implementation of step S2103 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0289] The model training method involved in the embodiments of this disclosure may include at least one of steps S3101 to S3106. For example, step S3101 may be implemented as a separate embodiment, step S3102 may be implemented as a separate embodiment, and steps S3101 and S3102 may be implemented as separate embodiments, but are not limited thereto.

[0290] In some embodiments, the order of any two steps S3101 to S3106 may be interchanged or they may be performed simultaneously. In some embodiments, one or more steps S3101 to S3106 are optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0291] In some embodiments, the steps and their optional implementations in other embodiments described before or after this embodiment, as well as other related parts in the specification, can be referred to, and will not be repeated here.

[0292] Figure 3B is an interactive schematic diagram of a communication method according to an embodiment of the present disclosure. As shown in Figure 3B, the present disclosure relates to a communication method, which includes:

[0293] Step S3201: The network device trains and generates a second AI model based on the first AI model and the second dataset.

[0294] The optional implementation of step S3201 can be found in the optional implementation of step S2101 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0295] In step S3202, the network device performs bilateral AI model performance testing on the first AI model and the second AI model, and generates performance testing results.

[0296] The optional implementation of step S3202 can be found in the optional implementation of step S2101 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0297] In step S3203, the network device determines that the set performance requirements are met based on the performance test results, and sends the second model parameters of the second AI model to the terminal.

[0298] The optional implementation of step S3203 can be found in the optional implementation of step S2102 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0299] In step S3204, the network device determines that the set performance requirements are not met based on the performance test results, and trains and generates a fourth AI model based on the first AI model and the second dataset.

[0300] The optional implementation of step S3204 can be found in the optional implementation of step S2102 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0301] In step S3205, the network device sends the first information to the terminal based on the fourth AI model.

[0302] In some embodiments, the first model parameters are used by the terminal to update and set the AI ​​model according to the first model parameters to generate a third AI model, or the first model parameters are used by the terminal to train and generate a third AI model according to the first model parameters and a third dataset.

[0303] In step S3206, the terminal determines the third AI model based on the first information.

[0304] The optional implementation of step S3206 can be found in the optional implementation of step S2103 in Figure 2A, as well as other related parts in the embodiments involved in Figure 2A, which will not be repeated here.

[0305] The model training method involved in the embodiments of this disclosure may include at least one of steps S3201 to S3206. For example, step S3201 may be implemented as a separate embodiment, step S3202 may be implemented as a separate embodiment, and steps S3201 and S3202 may be implemented as separate embodiments, but are not limited thereto.

[0306] In some embodiments, the order of any two steps S3201 to S3206 may be interchanged or they may be performed simultaneously. In some embodiments, one or more steps S3201 to S3206 are optional, and one or more of these steps may be omitted or substituted in different embodiments.

[0307] In some embodiments, the steps and their optional implementations in other embodiments described before or after this embodiment, as well as other related parts in the specification, can be referred to, and will not be repeated here.

[0308] In some embodiments, the bilateral AI model collaborative training method includes those shown in Embodiments 1 to 5.

[0309] Example 1: Collaborative training of a bilateral AI model based on a standardized reference AI model (method 1 above) and encoder model parameter passing (method 3b above).

[0310] (1) Set up a standardized encoder model (method 1-1 above).

[0311] Step S1: The decoder is trained on the NW side based on the standardized encoder and the dataset collected on the NW side.

[0312] Step S2: The NW monitors the performance of the bilateral AI model. If the monitored performance meets the requirements, the NW sends an indication message to the UE side to directly use the standardized encoder for inference, and does not execute Step S3. Otherwise, the NW side trains an encoder that meets the performance requirements based on the dataset collected by the NW and the trained decoder, and sends an indication message to instruct the UE to use the model obtained in Step S3 for inference.

[0313] Step S3: The NW side transmits the model parameters corresponding to the encoder trained in step S2 to the UE side. The UE side performs inference based on the received encoder model parameters. The model structure of the trained encoder is the same as the standardized encoder model structure.

[0314] (2) Set up a standardized decoder model (methods 1-2 above).

[0315] The encoder adopts a standardized model structure. The NW side trains the encoder based on the standardized reference decoder and the dataset collected by the NW side, and sends the model parameters to the UE. The UE and NW perform inference based on the trained encoder and the standardized decoder, respectively.

[0316] Example 2: Collaborative training of bilateral AI models is completed by using a standardized AI / ML model (method 1 above) and training the encoder model after passing encoder model parameters (method 3a-1 above).

[0317] (1) Set up a standardized reference encoder model (method 1-1 above).

[0318] Step S1: The decoder is trained on the NW side based on the standardized encoder and the dataset collected on the NW side.

[0319] Step S2: The NW monitors the performance of the bilateral AI model. If the monitored performance meets the requirements, the NW sends an indication message to the UE side to directly use the encoder for inference, and does not execute step S3. Otherwise, the NW side trains the encoder based on the dataset collected by the NW and the trained decoder, and sends an indication message to the UE to retrain the encoder based on the above method 3a-1.

[0320] Step S3: The NW side passes the model parameters corresponding to the encoder trained in step S2 to the UE side. The UE side then trains an encoder that meets the performance requirements based on the dataset collected by the UE side.

[0321] Step S4: NW and UE use the trained encoder and decoder respectively for inference.

[0322] (2) Set up a standardized decoder model (methods 1-2 above). The encoder model structure has been standardized.

[0323] Step S1: The NW side trains an encoder based on the standardized decoder and the dataset collected by the NW side, and then passes the model parameters of the encoder model to the UE.

[0324] Step S2: Following steps S2-S3 of the standardized encoder described above, complete the collaborative training of the bilateral model. The difference from step S2 is that the encoder model update or retraining is completed by the UE side.

[0325] Example 3: Collaborative training of bilateral AI models is completed based on a standardized reference AI model (method 1 above), model parameter transfer (method 3b above), and training the model after model parameter transfer (method 3a-1 above).

[0326] (1) Set up a standardized encoder model (1-1 above).

[0327] Steps S1-S2: Steps S1 and S2 are the same as those described in Example 1 above when standardizing the encoder model.

[0328] Step S3: The NW further monitors the UE's model performance based on the model parameter passing. If the performance meets the requirements, the UE uses the trained encoder obtained based on the model parameter passing for inference. Otherwise, the NW sends an indication message to the UE, instructing the UE to retrain the received encoder.

[0329] Step S4: The UE trains an encoder based on the dataset transmitted from the network and / or the dataset measured by the UE, and then performs inference based on the retrained encoder.

[0330] (2) Set up a standardized decoder model (methods 1-2 above), where the Encoder model structure has been standardized.

[0331] Step S1: The NW side trains an encoder based on the standardized decoder and the dataset collected by the NW side, and then passes the model parameters of the encoder model to the UE.

[0332] Step S2: The NW monitors the performance of the bilateral AI model. If the monitored performance meets the requirements, the NW sends an indication message to the UE side to directly use the encoder for inference, and does not execute Step S3. Otherwise, the NW side retrains an encoder that meets the performance requirements based on the dataset collected by the NW and the trained decoder, and sends an indication message to instruct the UE to use the model obtained in Step S3 for inference.

[0333] Steps S3-S4: Steps S3 and S4 are the same as steps S3 and S4 of the standardized encoder model described in Embodiment 3 above.

[0334] Example 4: Collaborative training of bilateral AI models is completed based on model parameter transfer (method 3b above) and training the model after model parameter transfer (method 3a-1 above).

[0335] Step S1: The NW side passes the encoder's model parameters to the UE.

[0336] Step S2: The UE uses the received encoder for inference. Then, the NW monitors the performance of the bilateral AI model. If the monitored performance meets the requirements, the NW sends an indication message to the UE to directly use the encoder for inference. Otherwise, the NW sends an indication message to the UE to execute step S3 to retrain the encoder.

[0337] Step S3: The UE trains the received encoder and the dataset collected by the UE to obtain an encoder that meets the performance requirements.

[0338] In some embodiments, the dataset collected by the UE may be a dataset measured by the UE or a dataset sent by the NW.

[0339] Example 5: Collaborative training of bilateral AI models based on standardized AI / ML models (method 1 above) and transmitted datasets for target CSI and feedback CSI (method 4-1 above).

[0340] (1) Set up a standardized reference encoder model (method 1-1 above).

[0341] Step S1: The decoder is trained on the NW side based on the standardized encoder and the dataset collected on the NW side.

[0342] Step S2: The NW monitors the performance of the bilateral AI model. If the monitored performance meets the requirements, the NW sends an indication message to the UE side to directly use the encoder for inference, and does not execute Step S3. Otherwise, the NW side trains the encoder based on the dataset collected by the NW and the trained decoder, and sends the dataset (target CSI and feedback CSI) of the trained encoder to the UE.

[0343] Step S3: The UE updates the standardized encoder's model parameters based on the received target CSI and feedback CSI to obtain a new encoder, or retrains a new encoder. The UE trains an encoder that meets the performance requirements based on the collected dataset.

[0344] Step S4: The NW and UE use the trained encoder and decoder respectively for inference, or the UE uses the standardized encoder and the decoder trained by the NW for inference.

[0345] (2) Set up a standardized decoder model (methods 1-2 above).

[0346] Step S1: The NW side trains an encoder based on the standardized decoder and the dataset collected by the NW side, and then passes the dataset (target CSI and feedback CSI) used to train the encoder to the UE.

[0347] Step S2: The UE side trains the encoder based on the received dataset.

[0348] Step S3: The UE and NW use the trained encoder and decoder respectively for inference.

[0349] This disclosure also proposes an apparatus (also referred to as a communication device, etc.) for implementing any of the above methods. For example, an apparatus is proposed that includes units or modules for implementing the steps performed by the terminal in any of the above methods. Furthermore, another apparatus is proposed that includes units or modules for implementing the steps performed by a network device (e.g., an access network device, a core network functional node, a core network device, etc.) in any of the above methods.

[0350] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC). The functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD). Taking a field-programmable gate array (FPGA) as an example, it can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0351] In this embodiment, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), or a Deep Learning Processing Unit (DPU).

[0352] Figure 4 is a schematic diagram of the structure of a terminal according to an embodiment of the present disclosure. Terminal 4100 is used to execute any of the above methods. In some embodiments, as shown in Figure 4, terminal 4100 may include at least one of a transceiver module 4101, a processing module 4102, etc. In some embodiments, the transceiver module 4101 is used to receive first information sent by a network device, the first information being generated by the network device based on performance detection results, the performance detection results being the bilateral AI model performance detection results of a first AI model and a second AI model, the first AI model being the CSI recovery partial model of the network device, and the second AI model being the CSI generation partial model of the terminal; the processing module 4102 is used to determine a third AI model based on the first information, the third AI model being the CSI generation partial model used for inference in the terminal. Optionally, the transceiver module is used to execute at least one of the communication steps such as sending and / or receiving performed by terminal 101 in any of the above methods, which will not be elaborated here. Optionally, the processing module is used to execute at least one of the other steps performed by terminal 101 in any of the above methods, which will not be elaborated here.

[0353] In some embodiments, the transceiver module may include a transmitting module and / or a receiving module, which may be separate or integrated. Optionally, the transceiver module may be interchangeable with a transceiver.

[0354] In some embodiments, the processing module may be a single module or may include multiple sub-modules. Optionally, the multiple sub-modules may each perform all or part of the steps required by the processing module.

[0355] In some embodiments, the processing module can be replaced by the processor, and the transceiver module can be replaced by the transceiver.

[0356] Figure 5 is a schematic diagram of the structure of a network device according to an embodiment of this disclosure. The network device 5100 is used to perform any of the above methods. In some embodiments, as shown in Figure 5, the network device 5100 may include at least one of a processing module 5101, a transceiver module 5102, etc. In some embodiments, the processing module 5101 is used to perform bilateral AI model performance testing on a first AI model and a second AI model, generating a performance testing result. The first AI model is the Channel State Information (CSI) recovery part model of the network device, and the second AI model is the CSI generation part model of the terminal. The transceiver module 5102 is used to send first information to the terminal based on the performance testing result. The first information is used by the terminal to determine a third AI model based on the first information. The third AI model is the CSI generation part model used for inference in the terminal. Optionally, the transceiver module is used to perform at least one of the communication steps such as sending and / or receiving performed by the network device 102 in any of the above methods, which will not be elaborated here. Optionally, the processing module is used to perform at least one of the other steps performed by the network device 102 in any of the above methods, which will not be elaborated here.

[0357] In some embodiments, the transceiver module may include a transmitting module and / or a receiving module, which may be separate or integrated. Optionally, the transceiver module may be interchangeable with a transceiver.

[0358] In some embodiments, the processing module may be a single module or may include multiple sub-modules. Optionally, the multiple sub-modules may each perform all or part of the steps required by the processing module.

[0359] In some embodiments, the processing module can be replaced by the processor, and the transceiver module can be replaced by the transceiver.

[0360] Figure 6A is a schematic diagram of the structure of the communication device 6100 proposed in an embodiment of this disclosure. The communication device 6100 can be a network device (e.g., access network device, core network device, etc.), a terminal (e.g., user equipment, etc.), a chip, chip system, or processor that supports the network device in implementing any of the above methods, or a chip, chip system, or processor that supports the terminal in implementing any of the above methods. The communication device 6100 can be used to implement the methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0361] As shown in Figure 6A, the communication device 6100 is used to execute any of the above methods. In some embodiments, the communication device 6100 includes one or more processors 6101. The processor 6101 may be a general-purpose processor or a special-purpose processor, such as a baseband processor or a central processing unit. The baseband processor may be used to process communication protocols and communication data, and the central processing unit may be used to control communication devices (e.g., base stations, baseband chips, terminal devices, terminal device chips, DUs or CUs, etc.), execute programs, and process program data. Optionally, the communication device 6100 is used to execute any of the above methods. Optionally, one or more processors 6101 are used to invoke instructions to cause the communication device 6100 to execute any of the above methods.

[0362] In some embodiments, the communication device 6100 further includes one or more transceivers 6102. When the communication device 6100 includes one or more transceivers 6102, the transceiver 6102 performs at least one of the communication steps such as sending and / or receiving in the above-described method, and the processor 6101 performs at least one of the other steps. In optional embodiments, the transceiver may include a receiver and / or a transmitter, which may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, transceiver circuit, interface circuit, interface, etc., can be used interchangeably; the terms transmitter, transmitting unit, transmitter, transmitting circuit, etc., can be used interchangeably; the terms receiver, receiving unit, receiver, receiving circuit, etc., can be used interchangeably.

[0363] In some embodiments, the communication device 6100 further includes one or more memories 6103 for storing data and / or instructions. Optionally, one or more processors 6101 are used to invoke instructions stored in the memory 6103 to cause the communication device 6100 to perform any of the above methods. Optionally, all or part of the memory 6103 may also be located outside the communication device 6100. In an optional embodiment, the communication device 6100 may include one or more interface circuits 6104. Optionally, the interface circuit 6104 is connected to the memory 6102 and can be used to receive data and / or instructions from the memory 6102 or other devices, and can be used to send data and / or instructions to the memory 6102 or other devices. For example, the interface circuit 6104 can read data and / or instructions stored in the memory 6102 and send the data and / or instructions to the processor 6101.

[0364] The communication device 6100 described in the above embodiments may be a network device or a terminal, but the scope of the communication device 6100 described in this disclosure is not limited thereto, and the structure of the communication device 6100 may not be limited by FIG. 6A. The communication device may be a standalone device or a part of a larger device. For example, the communication device may be: (1) a standalone integrated circuit IC, or chip, or chip system or subsystem; (2) a collection of one or more ICs, optionally, the IC collection may also include storage components for storing data, programs and / or instructions; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, smart terminal device, cellular phone, wireless device, handheld device, mobile unit, vehicle device, network device, cloud device, artificial intelligence device, etc.; (6) others, etc.

[0365] Figure 6B is a schematic diagram of the structure of chip 6200 according to an embodiment of this disclosure. For cases where the communication device 6100 can be a chip or a chip system, please refer to the schematic diagram of chip 6200 shown in Figure 6B, but it is not limited thereto.

[0366] Chip 6200 includes one or more processors 6201. Chip 6200 is used to perform any of the methods described above.

[0367] In some embodiments, chip 6200 further includes one or more interface circuits 6202. Optionally, terms such as interface circuit, interface, and transceiver pin can be used interchangeably. In some embodiments, chip 6200 further includes one or more memories 6203 for storing data and / or instructions. Optionally, all or part of the memories 6203 may be located outside of chip 6200. Optionally, interface circuit 6202 is connected to memory 6203, and interface circuit 6202 can be used to receive data and / or instructions from memory 6203 or other devices, and interface circuit 6202 can be used to send data and / or instructions to memory 6203 or other devices. For example, interface circuit 6202 can read data and / or instructions stored in memory 6203 and send the data and / or instructions to processor 6201.

[0368] In some embodiments, the interface circuit 6202 performs at least one of the communication steps, such as sending and / or receiving, in the above-described method. For example, the interface circuit 6202 performing the communication steps, such as sending and / or receiving, in the above-described method means that the interface circuit 6202 performs data and / or instruction interaction between the processor 6201, the chip 6200, the memory 6203, or the transceiver device. In some embodiments, the processor 6201 performs at least one of the other steps.

[0369] The modules and / or devices described in the various embodiments, such as virtual devices, physical devices, and chips, can be combined or separated arbitrarily as needed. Optionally, some or all steps can also be performed collaboratively by multiple modules and / or devices, which is not limited here.

[0370] This disclosure also proposes a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform any of the above methods. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but not limited thereto; it may also be a storage medium readable by other devices. Optionally, the storage medium may be a non-transitory storage medium, but not limited thereto; it may also be a temporary storage medium.

[0371] This disclosure also proposes a program product, including a program and / or instructions, which, when executed by a communication device, cause the communication device to perform any of the above methods. Optionally, the program product is a computer program product. Optionally, the program product is stored on the storage medium.

[0372] This disclosure also proposes a computer program that, when run on a computer, causes the computer to perform any of the above methods.

Claims

1. A communication method, executed by a network device, characterized in that, The method includes: A bilateral AI model performance test is performed on the first AI model and the second AI model to generate a performance test result. The first AI model is the Channel State Information (CSI) recovery part model of the network device, and the second AI model is the CSI generation part model of the terminal. Based on the performance test results, a first message is sent to the terminal. The first message is used by the terminal to determine a third AI model, which is the CSI generation part model used for inference in the terminal.

2. The method according to claim 1, characterized in that, The method further includes: The second AI model was determined to be the reference AI model; The first AI model is generated by training based on the second AI model and the first dataset.

3. The method according to claim 1 or 2, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are met; The first information is sent to the terminal, and the first information is also used to indicate that the third AI model is the second AI model.

4. The method according to claim 1 or 2, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are not met; Based on the first AI model and the second dataset, a fourth AI model is trained and generated, which is a CSI-generated partial model. According to the fourth AI model, the first information is sent to the terminal. The first information includes the first model parameters of the fourth AI model. The first model parameters are used by the terminal to update and set the AI ​​model to generate the third AI model, or the first model parameters are used by the terminal to train and generate the third AI model based on the first model parameters and the third dataset.

5. The method according to claim 1 or 3, characterized in that, The method further includes: The second model parameters of the second AI model are sent to the terminal, and the second model parameters are used by the terminal to generate the second AI model.

6. The method according to claim 5, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are not met; The first information is sent to the terminal, and the first information is also used to instruct the terminal to train and generate the third AI model based on the second model parameters and the third dataset.

7. The method according to claim 1 or 2, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are not met; Based on the first AI model and the second dataset, a fourth AI model is trained and generated, which is a CSI-generated partial model. According to the fourth AI model, the first information is sent to the terminal. The first information includes the fifth dataset of the fourth AI model. The fifth dataset is used by the terminal to update and set the AI ​​model to generate the third AI model, or the fifth dataset is used by the terminal to train the model to generate the third AI model. The fifth dataset includes target CSI and CSI feedback.

8. The method according to claim 1, characterized in that, The method further includes: The first AI model is determined to be the reference AI model; The second AI model is trained and generated based on the first AI model and the second dataset.

9. The method according to claim 8, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are met; Based on the second AI model, the first information is sent to the terminal. The first information includes the second model parameters of the second AI model. The second model parameters are used by the terminal to update and set the AI ​​model to generate the third AI model.

10. The method according to claim 8, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are not met; According to the second AI model, the first information is sent to the terminal. The first information includes the second model parameters of the second AI model. The second model parameters are used by the terminal to train and generate the third AI model based on the second model parameters and the third dataset.

11. The method according to claim 8, characterized in that, The step of sending first information to the terminal based on the performance detection result includes: Based on the performance test results, it is determined that the set performance requirements are not met; According to the second AI model, the first information is sent to the terminal. The first information includes a sixth dataset of the second AI model. The sixth dataset is used by the terminal to update and set the AI ​​model to generate the third AI model. Alternatively, the sixth dataset is used by the terminal to train the model to generate the third AI model. The sixth dataset includes target CSI and CSI feedback.

12. The method according to claim 4 or 10, characterized in that, The third dataset includes at least one of the following: The dataset sent by the network device; The dataset measured by the terminal.

13. A communication method, executed by a terminal, characterized in that, The method includes: The network device receives first information sent by the network device, the first information being generated by the network device based on performance testing results, the performance testing results being generated by the network device after performing bilateral AI model performance testing on a first AI model and a second AI model, the first AI model being the CSI recovery part model of the network device, and the second AI model being the CSI generation part model of the terminal. Based on the first information, a third AI model is determined, which is the CSI generation part model used for inference in the terminal.

14. The method according to claim 13, characterized in that, The step of determining the third AI model based on the first information includes: The second AI model was determined to be the reference AI model; Based on the first information, the third AI model is determined to be the second AI model.

15. The method according to claim 13, characterized in that, The first information includes the first model parameters of the fourth AI model, which is generated by the network device based on the first AI model and the second dataset. Determining the third AI model based on the first information includes: The AI ​​model is updated and configured based on the parameters of the first model to generate the third AI model; or... The third AI model is trained and generated based on the parameters of the first model and the third dataset.

16. The method according to claim 13, characterized in that, The method further includes: Receive the second model parameters sent by the network device, wherein the second model parameters are the model parameters of the second AI model; The AI ​​model is updated and set according to the parameters of the second model, and the second AI model is generated.

17. The method according to claim 16, characterized in that, The step of determining the third AI model based on the first information includes: Based on the first information, determine the third dataset; The third AI model is trained and generated based on the second model parameters and the third dataset.

18. The method according to claim 13, characterized in that, The first information includes a fifth dataset of the fourth AI model, the fifth dataset including target CSI and CSI feedback, the fourth AI model being trained and generated by the network device based on the first AI model and the second dataset, and determining the third AI model based on the first information includes: The AI ​​model is updated and configured based on the fifth dataset to generate the third AI model; or... The third AI model is generated by training the model based on the fifth dataset.

19. The method according to claim 13, characterized in that, The first information includes the second model parameters of the second AI model, and the step of determining the third AI model based on the first information includes: The first AI model is determined to be the reference AI model; Based on the second model parameters, the AI ​​model is updated and set to generate the third AI model.

20. The method according to claim 19, characterized in that, The first information includes the second model parameters of the second AI model, and the step of determining the third AI model based on the first information includes: The first AI model is determined to be the reference AI model; The third AI model is trained and generated based on the second model parameters and the third dataset.

21. The method according to claim 13, characterized in that, The first information includes a sixth dataset of the second AI model, the sixth dataset including target CSI and CSI feedback, and the step of determining the third AI model based on the first information includes: Based on the sixth dataset, update the set AI model to generate the third AI model; or... The third AI model is generated by training the model based on the sixth dataset.

22. The method according to claim 15 or 20, characterized in that, The third dataset includes at least one of the following: The dataset sent by the network device; The dataset measured by the terminal.

23. A communication device, characterized in that, The communication device is used to perform the communication method according to any one of claims 1-12 and 13-22.

24. A storage medium storing instructions, characterized in that, When the instructions are executed on the communication device, the communication device performs the communication method as described in any one of claims 1-12 or 13-22.

25. A program product comprising at least one of a program and instructions, characterized in that, When at least one of the programs or instructions is executed by the communication device, it implements the steps of the communication method according to claims 1-12, or when at least one of the programs or instructions is executed by the communication device, it implements the steps of the communication method according to claims 13-22.

Images