Model fusion method and device

Abstract

The invention provides a model fusion method and device, and the method comprises the steps: receiving first indication information, the first indication information is used for indicating the information of a model needing to be fused, and the model needing to be fused corresponds to different parameter ranges of at least one first channel feature; and performing model fusion based on the first indication information to determine a fused model. By adopting the method and the device, the generalization performance of the model can be ensured.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a model fusion method and apparatus. Background Technology

[0002] Currently, with the development and application of artificial intelligence (AI) technology, AI models have shown broad application prospects in many fields, covering industrial automation, gaming, agriculture, energy, environmental protection, and many others. They have also been widely used in wireless communication, achieving superior performance. However, when the model is actually deployed for inference, if the training channel data does not match the actual channel characteristics, the model will fail to achieve ideal performance. Specifically, this manifests as insufficient generalization performance on unknown data distributions, leading to significant performance degradation. Summary of the Invention

[0003] This application proposes a model fusion method and apparatus that can guarantee the generalization performance of the model.

[0004] In a first aspect, embodiments of this application provide a model fusion method, which can be applied to a terminal-side device. The terminal-side device can be a terminal device, a component in the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software that can implement all or part of the functions of the terminal device. The method includes: receiving first indication information, the first indication information being used to indicate information about a model to be fused, the model to be fused corresponding to at least one different parameter range of a first channel feature; and performing model fusion based on the first indication information to determine the fused model.

[0005] In the above method, by employing the aforementioned approach, even when the model supported by the terminal side lacks certain channel features, the mismatch between the adopted single model (one of the models supported by the terminal side) and the channel features of the actual environment (i.e., the missing channel features) is avoided. This means that when the training channel data of a single model does not match the actual channel features, the single model cannot achieve ideal performance. By using existing models (the models that need to be fused among the models supported by the terminal side) to perform model fusion, a fused model is obtained. Through the generalization ability of model fusion, good performance can be achieved even when channel features are missing. In summary, the generalization performance of the model is guaranteed, thereby ensuring the model's performance. Furthermore, using the fused model for operation results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0006] In one possible implementation, the information of the model to be fused is related to the information of the second channel feature and the information of the model supported by the terminal side. The information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model.

[0007] The above method ensures the generalization performance of the model, thereby guaranteeing the model's overall performance.

[0008] Optionally, the second channel feature includes: at least one second channel feature, and different parameter ranges of at least one second channel feature.

[0009] In another possible implementation, the information of the models to be fused is also related to at least one of the following, which includes: terminal-side model fusion capability information, or upper limit information on the number of models to be fused transmitted.

[0010] The above method takes into account the constraints of actual transmission, namely the upper limit of the number of models that need to be fused for transmission, and the constraints of terminal capabilities, namely the terminal-side model fusion capability information, thereby ensuring the generalization performance of the model.

[0011] In another possible implementation, at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0012] In another possible implementation, the second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0013] In another possible implementation, the first indication information includes the identification information of the model that needs to be fused.

[0014] Optionally, the identification information of the models to be fused includes the index information of the models to be fused.

[0015] The above method helps to identify and determine the models that need to be fused.

[0016] Secondly, embodiments of this application provide a model fusion method, which can be applied to a network-side device. This network-side device can be a network device, a component within the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the network device's functions. The method includes: determining a second channel feature; determining information about the model to be fused based on the second channel feature and information about models supported by the terminal side, wherein the information about each model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature among the at least one first channel feature corresponding to each model; and sending first indication information, which is used to indicate the information about the model to be fused, and the different parameter ranges of the at least one first channel feature corresponding to the model to be fused.

[0017] Optionally, the second channel feature includes: the channel feature corresponding to the terminal device, the current channel feature, the predicted channel feature, or the historical channel feature.

[0018] In the above method, even when the model supported by the terminal side lacks certain channel features, it avoids the problem of a mismatch between the adopted single model (one of the models supported by the terminal side) and the channel features of the actual environment (i.e., the missing channel features). This means that when the training channel data of a single model does not match the actual channel features, the single model cannot achieve ideal performance. By indicating the information of the model to be fused by the network device, the terminal device performs fusion based on the information of the model to be fused. That is, it uses the existing model (the model to be fused among the models supported by the terminal side) to perform model fusion to obtain the fused model. Through the generalization ability of model fusion, good performance can be achieved even when channel features are missing. In summary, it can guarantee the generalization performance of the model, thereby guaranteeing the model's performance. Furthermore, using the fused model for operation results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0019] In one possible implementation, the information on determining the models to be fused based on the second channel characteristics and the information on the models supported by the terminal side includes: determining the information on the models to be fused based on the second channel characteristics and the information on the models supported by the terminal side, and at least one of the following, wherein at least one of the following includes: terminal-side model fusion capability information, or upper limit information on the number of transmitted models to be fused.

[0020] In the above method, when determining the information of the models that need to be fused, the network device considers the constraints of actual transmission, namely the upper limit information of the number of models that need to be fused, and the constraints of terminal capabilities, namely the terminal-side model fusion capability information, thereby ensuring the generalization performance of the models.

[0021] In another possible implementation, at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0022] In another possible implementation, the second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0023] In another possible implementation, the second channel feature includes: a first feature, a first parameter range corresponding to the first feature, a second feature, and a second parameter range corresponding to the second feature. The information of the model supported by the terminal side includes information of a first model. The information of the first model includes: the first feature, a third parameter range corresponding to the first feature, the second feature, and a fourth parameter range corresponding to the second feature. The first feature and the second feature are at least one first channel feature corresponding to the first model, and the third parameter range and the fourth parameter range are the parameter ranges of each channel feature in the at least one first channel feature corresponding to the first model. The step of determining the information of the model to be fused based on the second channel feature and the information of the model supported by the terminal side includes: the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than a third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than a fourth threshold. The information of the model to be fused includes the identification information of the first model.

[0024] In the above method, even when the model supported by the terminal side lacks certain channel features, the generalization performance of the model can be guaranteed, thereby ensuring the performance of the model.

[0025] In another possible implementation, the method further includes receiving information about the model supported by the terminal side.

[0026] Optionally, the network device sends information about the models supported by the network side to the terminal device.

[0027] In the above method, by having the network device and the terminal device exchange information about their respective supported models, the models on both sides can be aligned for subsequent use.

[0028] Thirdly, embodiments of this application provide a model fusion method, which can be applied to a terminal-side device. This terminal-side device can be a terminal device, a component within the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the terminal device's functions. The method includes: receiving first information, the first information indicating at least one second channel feature, different parameter ranges corresponding to the at least one second channel feature, and weights corresponding to the at least one second channel feature; determining a model to be fused based on the at least one second channel feature, the different parameter ranges corresponding to the at least one second channel feature, and information about models supported by the terminal side, wherein the information about each model in the information about models supported by the terminal side includes: at least one first channel feature corresponding to each model, or the parameter range of each first channel feature among the at least one first channel feature corresponding to each model; and performing model fusion based on the weights corresponding to the at least one second channel feature and the model to be fused to determine the fused model.

[0029] In the above method, by employing the aforementioned approach, even when the model supported by the terminal side lacks certain channel features, the mismatch between the adopted single model (one of the models supported by the terminal side) and the channel features of the actual environment (i.e., the missing channel features) is avoided. This means that when the training channel data of a single model does not match the actual channel features, the single model cannot achieve ideal performance. The terminal device can use the first information indicated by the network device to perform model fusion using existing models (the models that need to be fused among the models supported by the terminal side) to obtain a fused model. Through the generalization capability of model fusion, good performance can be achieved even when channel features are missing. In summary, the generalization performance of the model is guaranteed, thereby ensuring the model's performance. Furthermore, using the fused model results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0030] In one possible implementation, the at least one second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0031] In another possible implementation, at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0032] In another possible implementation, the first information includes: a first feature, a first parameter range corresponding to the first feature, a second feature, a second parameter range corresponding to the second feature, and a first weight, wherein the first feature and the second feature are one of the at least one second channel features, the first parameter range and the second parameter range are different parameter ranges corresponding to the at least one second channel feature, and the first weight is the weight corresponding to the at least one second channel feature; the information of the model supported by the terminal side includes information of a first model, which includes: the first feature, a third parameter range corresponding to the first feature, the second feature, and a fourth parameter range corresponding to the second feature, wherein the first feature and the second feature are the at least one first channel feature. The first channel feature is defined as a channel feature, and the third parameter range and the fourth parameter range are parameter ranges for each of the at least one first channel feature. The determination of the model to be fused based on the at least one second channel feature, the different parameter ranges corresponding to the at least one second channel feature, and the information of the models supported by the terminal side includes: if the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than a third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than a fourth threshold, then the model to be fused includes the first model.

[0033] The above method ensures the generalization performance of the model, thereby guaranteeing the model's overall performance.

[0034] Fourthly, embodiments of this application provide a model fusion method, which can be applied to a network-side device. The network-side device can be a network device, a component in the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software that can implement all or part of the functions of the network device. The method includes: determining first information, the first information being used to indicate at least one second channel feature, different parameter ranges corresponding to the at least one second channel feature, and weights corresponding to the at least one second channel feature; and sending the first information.

[0035] In the above method, the method of sending the first information from the network device to the terminal device can help the terminal device to use the existing model (the model that needs to be fused among the models supported by the terminal side) to perform model fusion to obtain the fused model, thereby ensuring the generalization performance and performance of the model.

[0036] In one possible implementation, the at least one second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0037] Fifthly, embodiments of this application provide a model fusion apparatus, which can be a terminal device, a component in the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software that can implement all or part of the functions of the terminal device.

[0038] In one possible implementation, the model fusion device may include modules, units, or means that correspond one-to-one with the methods / operations / steps / actions described in the first aspect. These modules, units, or means may be hardware circuits, software, or a combination of hardware circuits and software.

[0039] In one possible implementation, the model fusion apparatus includes a processing unit and a transceiver unit. The transceiver unit is configured to receive first indication information, which indicates information about the models to be fused, wherein the models to be fused correspond to different parameter ranges of at least one first channel feature. The processing unit is configured to perform model fusion based on the first indication information to determine the fused model.

[0040] In one possible implementation, the information of the model to be fused is related to the information of the second channel feature and the information of the model supported by the terminal side. The information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model.

[0041] In another possible implementation, the information of the models to be fused is also related to at least one of the following, which includes: terminal-side model fusion capability information, or upper limit information on the number of models to be fused transmitted.

[0042] In another possible implementation, at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0043] In another possible implementation, the second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0044] In another possible implementation, the first indication information includes the identification information of the model that needs to be fused.

[0045] For the technical effects of the fifth aspect or possible implementation, please refer to the introduction of the technical effects of the first aspect or corresponding implementation.

[0046] Sixthly, embodiments of this application provide a model fusion apparatus, which can be a network device, a component in the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the functions of the network device.

[0047] In one possible implementation, the model fusion device may include modules, units, or means that correspond one-to-one with the methods / operations / steps / actions described in the second aspect. These modules, units, or means may be hardware circuits, software, or a combination of hardware circuits and software.

[0048] In one possible implementation, the model fusion apparatus includes: a processing unit and a transceiver unit. The processing unit is configured to determine a second channel feature; the processing unit is configured to determine information about the models to be fused based on the second channel feature and information about models supported by the terminal side, wherein the information about each model supported by the terminal side includes: at least one first channel feature corresponding to each model, and a parameter range for each first channel feature; the transceiver unit is configured to send first indication information, the first indication information indicating the information about the models to be fused, and different parameter ranges for at least one first channel feature corresponding to the models to be fused.

[0049] In one possible implementation, the processing unit is configured to determine information about the models that need to be fused based on the second channel characteristics and information about the models supported by the terminal side, as well as at least one of the following, which includes: terminal-side model fusion capability information, or upper limit information on the number of transmitted models that need to be fused.

[0050] In another possible implementation, at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0051] In another possible implementation, the second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0052] In another possible implementation, the second channel feature includes: a first feature, a first parameter range corresponding to the first feature, a second feature, and a second parameter range corresponding to the second feature. The information of the model supported by the terminal side includes information of a first model. The information of the first model includes: the first feature, a third parameter range corresponding to the first feature, the second feature, and a fourth parameter range corresponding to the second feature. The first feature and the second feature are at least one first channel feature corresponding to the first model, and the third parameter range and the fourth parameter range are the parameter ranges of each channel feature in the at least one first channel feature corresponding to the first model. The processing unit is configured to determine that the information of the model to be fused includes the identification information of the first model when the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than a third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than a fourth threshold.

[0053] In another possible implementation, the transceiver unit is further configured to receive information about the model supported by the terminal side.

[0054] For the technical effects of the sixth aspect or possible implementation, please refer to the introduction of the technical effects of the second aspect or corresponding implementation.

[0055] In a seventh aspect, embodiments of this application provide a model fusion apparatus, which can be a terminal device, a component in the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the functions of the terminal device.

[0056] In one possible implementation, the model fusion device may include modules, units, or means that correspond one-to-one with the methods / operations / steps / actions described in the first aspect. These modules, units, or means may be hardware circuits, software, or a combination of hardware circuits and software.

[0057] In one possible implementation, the model fusion apparatus includes: a processing unit and a transceiver unit. The transceiver unit is configured to receive first information, which indicates at least one second channel feature, different parameter ranges corresponding to the at least one second channel feature, and weights corresponding to the at least one second channel feature. The processing unit is configured to determine the model to be fused based on the at least one second channel feature, the different parameter ranges corresponding to the at least one second channel feature, and information of models supported by the terminal side. The information of each model supported by the terminal side includes: at least one first channel feature corresponding to each model, or the parameter range of each first channel feature among the at least one first channel feature corresponding to each model. The processing unit is configured to perform model fusion based on the weights corresponding to the at least one second channel feature and the model to be fused to determine the fused model.

[0058] In one possible implementation, the at least one second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0059] In another possible implementation, at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0060] In another possible implementation, the first information includes: a first feature, a first parameter range corresponding to the first feature, a second feature, a second parameter range corresponding to the second feature, and a first weight, wherein the first feature and the second feature are one of the at least one second channel features, the first parameter range and the second parameter range are different parameter ranges corresponding to the at least one second channel feature, and the first weight is the weight corresponding to the at least one second channel feature; the information of the model supported by the terminal side includes information of a first model, the information of the first model including: the first feature, a third parameter range corresponding to the first feature, the second feature, and a fourth parameter range corresponding to the second feature, wherein... The first feature and the second feature are channel features in the at least one first channel feature, and the third parameter range and the fourth parameter range are parameter ranges for each first channel feature in the at least one first channel feature; the processing unit is configured to determine that the model to be fused includes the first model when the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than a third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than a fourth threshold.

[0061] For the technical effects of the seventh aspect or possible implementation, please refer to the introduction of the technical effects of the third aspect or corresponding implementation.

[0062] Eighthly, embodiments of this application provide a model fusion apparatus, which can be a network device, a component in the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the functions of the network device.

[0063] In one possible implementation, the model fusion device may include modules, units, or means that correspond one-to-one with the methods / operations / steps / actions described in the second aspect. These modules, units, or means may be hardware circuits, software, or a combination of hardware circuits and software.

[0064] In one possible implementation, the model fusion apparatus includes a processing unit and a transceiver unit. The processing unit is configured to determine first information, which indicates at least one second channel feature, different parameter ranges corresponding to the at least one second channel feature, and weights corresponding to the at least one second channel feature. The transceiver unit is configured to transmit the first information.

[0065] In one possible implementation, the at least one second channel feature includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

[0066] For the technical effects of the eighth aspect or possible implementation, please refer to the introduction of the technical effects of the fourth aspect or corresponding implementation.

[0067] Ninthly, embodiments of this application provide a model fusion apparatus, which includes at least one processor that invokes a computer program or instructions to execute the method described in the first aspect or a possible implementation thereof.

[0068] In one possible implementation, the model fusion device also includes a memory and a communication interface. Optionally, the memory and processor are integrated together.

[0069] In one possible implementation, the memory is located outside the model fusion device.

[0070] In a tenth aspect, embodiments of this application provide a model fusion apparatus, which includes at least one processor that invokes a computer program or instructions to execute the method described in the second aspect or a possible implementation thereof.

[0071] In one possible implementation, the model fusion device also includes a memory and a communication interface. Optionally, the memory and processor are integrated together.

[0072] In one possible implementation, the memory is located outside the model fusion device.

[0073] Eleventhly, embodiments of this application provide a model fusion apparatus, which includes at least one processor that invokes computer programs or instructions to execute the methods described in the third aspect or possible implementations of the third aspect.

[0074] In one possible implementation, the model fusion device also includes a memory and a communication interface. Optionally, the memory and processor are integrated together.

[0075] In one possible implementation, the memory is located outside the model fusion device.

[0076] In a twelfth aspect, embodiments of this application provide a model fusion apparatus, which includes at least one processor that invokes a computer program or instructions to perform the method described in the fourth aspect or a possible implementation thereof.

[0077] In one possible implementation, the model fusion device also includes a memory and a communication interface. Optionally, the memory and processor are integrated together.

[0078] In one possible implementation, the memory is located outside the model fusion device.

[0079] In a thirteenth aspect, embodiments of this application provide a chip system including at least one processor for executing computer programs or instructions to implement any of the above aspects or possible implementations of any of the above aspects.

[0080] In one possible implementation, the input of the chip system corresponds to the receiving operation in any of the above aspects or possible implementations, and the output of the chip system corresponds to the sending operation in any of the above aspects or possible implementations.

[0081] Optionally, the processor is coupled to the memory via an interface.

[0082] Optionally, the chip system may also include a memory in which computer program instructions are stored.

[0083] In a fourteenth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program or instructions that, when executed on a processor, implement the methods described above.

[0084] In a fifteenth aspect, embodiments of this application provide a computer program product that includes a computer program or instructions that, when executed on a processor, implement the methods described above.

[0085] In a sixteenth aspect, embodiments of this application provide a model fusion system, which includes: the apparatus as described in the ninth aspect and the apparatus as described in the tenth aspect, or the apparatus as described in the eleventh aspect and the apparatus as described in the twelfth aspect. Attached Figure Description

[0086] Figure 1A This is a schematic diagram of the architecture of a communication system provided in an embodiment of this application;

[0087] Figure 1B This is a schematic diagram of the architecture of another communication system provided in the embodiments of this application;

[0088] Figure 1C This is a schematic diagram of a possible application framework in the communication system provided in the embodiments of this application;

[0089] Figure 1D This is a schematic diagram of another possible application framework in the communication system provided in the embodiments of this application;

[0090] Figure 2 This is a schematic diagram of a neuron structure;

[0091] Figure 3 This is a schematic diagram of a neural network structure;

[0092] Figure 4 This is a schematic diagram of a model fusion method provided in an embodiment of this application;

[0093] Figure 5 This is a schematic diagram of a second channel feature and a model supported by the terminal device side, provided in an embodiment of this application;

[0094] Figure 6 This is a schematic diagram of the channel feature levels supported by the respective models of a network device and a terminal device provided in this application embodiment;

[0095] Figure 7 This is a schematic diagram of yet another model fusion method provided in the embodiments of this application;

[0096] Figure 8 This is a schematic diagram of a first indication information provided in an embodiment of this application;

[0097] Figure 9 This is a schematic diagram illustrating how to determine the models that need to be fused, as provided in an embodiment of this application.

[0098] Figures 10-12 This is a schematic diagram of yet another model fusion method provided in the embodiments of this application;

[0099] Figure 13 This is a schematic diagram of the structure of a model fusion device provided in an embodiment of this application;

[0100] Figure 14 This is a schematic diagram of the structure of another model fusion device provided in the embodiments of this application;

[0101] Figure 15 This is a schematic diagram of the structure of a chip system provided in an embodiment of this application. Detailed Implementation

[0102] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0103] References to "one embodiment" or "some embodiments" as described in this application mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0104] In the description of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. "And / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can represent: a, b, c; a and b; a and c; b and c; or a and b and c. Where a, b, and c can be single or multiple.

[0105] It is understood that in this application, "instruction" can include direct instruction, indirect instruction, explicit instruction, and implicit instruction. When describing a certain instruction information to indicate A, it can be understood that the instruction information carries A, directly indicates A, or indirectly indicates A.

[0106] In this application, the information indicated by the instruction information is called the information to be instructed. In specific implementations, there are many ways to instruct the information to be instructed, such as, but not limited to, directly instructing the information to be instructed, such as the information to be instructed itself or its index; indirectly instructing the information to be instructed by instructing other information, where there is a relationship between the other information and the information to be instructed; or instructing only a part of the information to be instructed, while the other parts are known or pre-agreed upon. For example, the instruction of specific information can be achieved by using a pre-agreed (e.g., protocol-defined) arrangement of various pieces of information, thereby reducing instruction overhead to some extent.

[0107] The information to be instructed can be sent as a whole or divided into multiple sub-information messages, and the sending period and / or timing of these sub-information messages can be the same or different. This application does not limit the specific sending method. The sending period and / or timing of these sub-information messages can be predefined, for example, according to a protocol, or configured by the transmitting device by sending configuration information to the receiving device.

[0108] It is understood that "send" and "receive" in this application refer to the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information being XX, which can include direct transmission via the air interface or indirect transmission via the air interface from other units or modules. "Receive information from YY" can be understood as the source of the information being YY, which can include direct reception from YY via the air interface or indirect reception from YY via the air interface from other units or modules. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface.

[0109] In other words, sending and receiving can occur between devices, such as between network devices and terminal devices, or within a device, such as between components, modules, chips, software modules, or hardware modules within the device via buses, wiring, or interfaces.

[0110] It is understandable that information may undergo necessary processing, such as encoding and modulation, between the source and destination, but the destination can understand the valid information from the source. Similar statements in this application can be interpreted in a similar way and will not be elaborated further.

[0111] The communication method provided in this application can be applied to cellular communication systems related to the 3rd Generation Partnership Project (3GPP), such as 4th generation (4G) communication systems, such as Long Term Evolution (LTE) communication systems. The LTE communication system may include LTE frequency division duplex (FDD) systems and LTE time division duplex (TDD) systems. It can also be applied to 5th generation (5G) communication systems, such as 5G new radio (NR) communication systems, or to various future communication systems and future communication networks. The method provided in this application embodiment can also be applied to Bluetooth systems, wireless local area network (WLAN) systems, wireless fidelity (WiFi) systems, LoRa systems, or vehicle-to-everything (V2X) systems, communication systems supporting the integration of multiple wireless technologies, device-to-device (D2D) systems, vehicle-to-everything (V2X) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), and Internet of Things (IoT) communication systems. The method provided in this application embodiment can also be applied to satellite communication systems, wherein the satellite communication system can be integrated with the above-mentioned communication systems.The wireless communication systems involved in this application also include, but are not limited to: narrowband Internet of Things (NB-IoT) systems, global system for mobile communications (GSM), enhanced data rate for GSM evolution (EDGE), wideband code division multiple access (WCDMA) systems, code division multiple access 2000 (CDMA2000) systems, or time division-synchronization code division multiple access (TD-SCDMA) systems.

[0112] In a communication system, one network element can send signals to or receive signals from another network element. These signals can include information, signaling, or data. The term "network element" can also be replaced by an entity, network entity, device, communication equipment, communication module, node, communication node, etc. This application uses a network element as an example for description. For instance, a communication system can include at least one terminal device and at least one network device. The network device can send downlink signals to the terminal device, and / or the terminal device can send uplink signals to the network device. It is understood that the terminal device in this application can be replaced by a first network element, and the network device can be replaced by a second network element, both performing the corresponding method embodiments described in this application.

[0113] Please see Figure 1A , Figure 1A This is a schematic diagram of the architecture of a communication system provided in an embodiment of this application, to Figure 1A The application scenario used in this application is illustrated using the communication system 100 architecture shown as an example. The communication system 100 includes at least one network device, such as... Figure 1A The network device 110 and communication system 100 shown may also include at least one terminal device, such as Figure 1A The terminal devices 120 and 130 are shown. Network device 110 can communicate with the terminal devices (such as terminal devices 120 and 130) via a wireless link. Communication devices in this communication system, for example, network device 110 and terminal device 120, can communicate via multi-antenna technology.

[0114] In wireless communication networks, such as mobile communication networks, the services supported by the networks are becoming increasingly diverse, thus requiring increasingly diverse demands. For example, networks need to support ultra-high speeds, ultra-low latency, and / or massive connectivity. This characteristic makes network planning, network configuration, and / or resource scheduling increasingly complex. Furthermore, as network functions become more powerful, such as supporting higher spectrum levels, supporting higher-order multiple-input multiple-output (MIMO) technologies, supporting beamforming, and / or supporting beam management, network energy efficiency has become a hot research topic. These new demands, new scenarios, and new characteristics bring unprecedented challenges to network planning, operation, and efficient operation. To meet these challenges, artificial intelligence (AI) technology can be introduced into wireless communication networks to achieve network intelligence. To support AI technology in wireless networks, AI nodes, also known as AI network elements, may be introduced.

[0115] Please see Figure 1B , Figure 1B This is a schematic diagram of the architecture of another communication system provided in the embodiments of this application. The communication system 200 includes at least one network device, such as... Figure 1B The network device 110 and communication system 200 shown may also include at least one terminal device, such as Figure 1B The terminal devices 120 and 130 are shown. Compared to Figure 1A Regarding the communication system 100 shown, Figure 1B The communication system 200 shown also includes an AI network element 140. The AI ​​network element 140 is used to perform AI-related operations, such as building training datasets or training AI models.

[0116] In one possible implementation, network device 110 can send data related to the training of the AI ​​model to AI network element 140, which then constructs a training dataset and trains the AI ​​model. For example, the data related to the training of the AI ​​model may include data reported by the terminal device. AI network element 140 can send the results of operations related to the AI ​​model to network device 110, which then forwards them to the terminal device. For example, the results of operations related to the AI ​​model may include at least one of the following: a trained AI model, model evaluation results, or test results. Exemplarily, a portion of the trained AI model may be deployed on network device 110, and another portion on the terminal device. Alternatively, the trained AI model may be deployed on network device 110. Or, the trained AI model may be deployed on the terminal device.

[0117] It should be understood that Figure 1B Taking the direct connection between AI network element 140 and network device 110 as an example, in other scenarios, AI network element 140 can also be connected to a terminal device. Alternatively, AI network element 140 can be connected to both network device 110 and a terminal device simultaneously. Alternatively, AI network element 140 can also be connected to network device 110 through a third-party network element. This application embodiment does not limit the connection relationship between AI network element and other network elements. Figure 1B Taking AI network element 140 as a standalone network element as an example, AI network element 140 can also be configured as a module in network devices and / or terminal devices, for example, configured in Figure 1B This application does not limit the network device 110 or terminal device shown.

[0118] It should be noted that, Figure 1A and Figure 1B This is a simplified illustration for ease of understanding only. For example, the communication system may also include other devices, such as wireless relay devices and / or wireless backhaul devices. Figure 1A and Figure 1B The figures are not shown. In practical applications, this communication system may include multiple network devices or multiple terminal devices. This application does not limit the number of network devices and terminal devices included in the communication system.

[0119] It should be understood that Figure 1A and Figure 1B The network devices and terminal devices mentioned can be hardware, software based on functional distinctions, or a combination of both. The network devices described above can be any of the network devices described below, and the terminal devices can be any of the terminal devices described below. It should be noted that the methods described in the embodiments of this application can be applied to... Figure 1A and Figure 1B The communication system shown.

[0120] (1) Terminal equipment, also known as user equipment (UE), user unit, user station, mobile station (MS), remote station, mobile device, mobile terminal (MT), terminal, wireless communication equipment, etc., is a device that provides voice or data connectivity to a user. Specifically, it includes devices that provide voice connectivity to a user, devices that provide data connectivity to a user, or devices that provide both voice and data connectivity to a user. For example, it may include handheld devices with wireless connectivity or processing devices connected to a wireless modem. This terminal equipment can communicate with the core network via a radio access network (RAN), exchanging voice or data with the RAN, or interacting with the RAN to exchange voice and data. Currently, terminal devices can be: mobile phones, tablets, laptops, PDAs, mobile internet devices (MIDs), wearable devices (such as smartwatches, smart bracelets, pedometers, etc.), in-vehicle devices (such as cars, bicycles, electric vehicles, airplanes, ships, trains, high-speed trains, etc.), virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, smart home devices (such as refrigerators, televisions, air conditioners, electricity meters, etc.), intelligent robots, workshop equipment, wireless terminals in self-driving vehicles, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, or flying devices (such as intelligent robots, hot air balloons, drones, airplanes), etc. Terminal devices can also be other devices with terminal functions; for example, a terminal device can also be a device that performs terminal functions in D2D communication.Terminal devices can also include vehicle-to-everything (V2X) terminal devices, machine-to-machine / machine-type communications (M2M / MTC) terminal devices, Internet of Things (IoT) terminal devices, light UEs, reduced capability UEs (REDCAP UEs), subscriber units, subscriber stations, mobile stations, remote stations, access points (APs), remote terminals, access terminals, user terminals, user agents, or user devices, and drone equipment. For example, this can include mobile phones (or "cellular" phones), computers with mobile terminal devices, portable, pocket-sized, handheld, and computer-embedded mobile devices, etc. Examples include personal communication service (PCS) telephones, cordless telephones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication capabilities, computing devices, or other processing devices connected to a wireless modem. It also includes limited devices, such as devices with low power consumption, limited storage capacity, or limited computing power. Examples include information sensing devices such as barcode scanners, radio frequency identification (RFID), sensors, global positioning systems (GPS), and laser scanners. In this application, terminal devices with wireless transceiver capabilities and chips that can be installed in the aforementioned terminal devices are collectively referred to as terminal devices.

[0121] As an example and not a limitation, in the embodiments of this application, when the terminal device can be a wearable device, wearable devices can also be called wearable smart devices. Wearable devices are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, accessories, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly defined, wearable smart devices include those with comprehensive functions, large size, and the ability to achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those focused on a specific application function that require cooperation with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0122] It should be noted that, in the embodiments of this application, the device used to implement the functions of the terminal device can be the terminal device itself, or it can be any device capable of supporting the terminal device in implementing the functions, such as a chip system, hardware circuit, software module, or a hardware circuit plus a software module. This device can be installed in the terminal device or used in conjunction with the terminal device. In the embodiments of this application, the chip system can be composed of chips, or it can include chips and other discrete devices. In this embodiment, the terminal device is used as an example to illustrate the device used to implement the functions of the terminal device, and this does not constitute a limitation on the solutions of the embodiments of this application.

[0123] (2) A network device is a device deployed in a wireless access network to provide wireless communication functions for terminal devices. A network device may also be called a wireless access network (RAN) entity, access network equipment, wireless access network device, access node, wireless node, network node, or communication device, etc.

[0124] For example, the network device can be an access network device for a cellular system related to the 3GPP (3rd Generation Partnership Project). For instance, a fourth-generation (4G) mobile communication system or a 5G mobile communication system. The network device can also be an access network device in an open RAN (O-RAN or ORAN) or cloud radio access network (CRAN). Alternatively, the network device can also be an access network device in a communication system resulting from the convergence of two or more of the above communication systems.

[0125] Network equipment includes, but is not limited to: evolved Node B (eNB), radio network controller (RNC), Node B (NB), base station controller (BSC), base transceiver station (BTS), home base station (e.g., home-evolved Node B, or home Node B, HNB), baseband unit (BBU), access point (AP), relay station, macro base station, micro base station, wireless relay node, donor node or similar, or combinations thereof, in Wi-Fi systems; radio controller, wireless backhaul node, transmitting and receiving point (TRP), transmitting point (TP), master station, slave station, motorslide retainer (MSR) node, transmission node, or transceiver node in CRAN scenarios. Network equipment can also be access network equipment in 5G mobile communication systems. For example, a next-generation NodeB (gNB), TRP, TP in a New Radio (NR) system, or one or a group of antenna panels (including multiple antenna panels) of a base station in a 5G mobile communication system. Alternatively, network equipment can also be a network node constituting a gNB or transmission point. For example, a central unit (CU), a distributed unit (DU), or a radio unit (RU). Network equipment can also be a baseband unit (BBU), a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH). Network equipment can also refer to communication modules, modems, or chips used in the aforementioned equipment or devices. Network equipment can also be a mobile switching center and equipment that performs base station functions in D2D, V2X, and M2M communications, network-side equipment in next-generation communication networks, and equipment that performs base station functions in future communication systems. Network equipment can support networks with the same or different access technologies. Network devices can also be servers, wearable devices, vehicles, or in-vehicle equipment. For example, in V2X technology, network devices can be roadside units (RSUs).

[0126] Network equipment can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move depending on the location of that mobile base station. In other examples, a helicopter or drone can be configured as a device to communicate with another base station.

[0127] In some deployments, the network devices mentioned in the embodiments of this application may be devices including CU, DU, or CU and DU, or devices with control plane CU nodes (central unit-control plane (CU-CP)) and user plane CU nodes (central unit-user plane (CU-UP)) and DU nodes. For example, the network devices may include gNB-CU-CP, gNB-CU-UP, and gNB-DU.

[0128] In some deployments, multiple RAN nodes collaborate to assist terminals in achieving wireless access, with different RAN nodes each implementing some of the base station's functions. For example, RAN nodes can be CUs, DUs, CU-CPs, CU-UPs, or RUs. CUs and DUs can be configured separately or included in the same network element, such as a BBU. RUs can be included in radio frequency equipment or radio frequency units, such as RRUs, AAUs, or RRHs.

[0129] RAN nodes can support one or more types of fronthaul interfaces, each corresponding to a DU and RU with different functions. If the fronthaul interface between the DU and RU is a common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and RU is another type of interface, relative to CPRI, some downlink and / or uplink baseband functions, such as, for downlink, precoding, digital beamforming (BF), or one or more of inverse fast Fourier transform (IFFT) / cyclic prefix addition (CP), are moved from the DU to the RU; and for uplink, digital beamforming (BF), or one or more of fast Fourier transform (FFT) / cyclic prefix removal (CP), are moved from the DU to the RU. In one possible implementation, the interface can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the segmentation between DU and RU differs, corresponding to different categories (Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, F.

[0130] Taking eCPRI Cat A as an example, for downlink transmission, the DU is configured to implement one or more functions preceding and following layer mapping (i.e., coding, rate matching, scrambling, modulation, and layer mapping), while other functions following layer mapping (e.g., resource element (RE) mapping, digital beamforming (BF), or one or more functions following inverse fast Fourier transform (IFFT) / cyclic prefix (CP) addition) are moved to the RU. For uplink transmission, the DU is configured to implement one or more functions preceding and following de-mapping (i.e., decoding, rate matching de-mapping, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and de-RE mapping), while other functions following de-mapping (e.g., digital BF or fast Fourier transform (FFT) / CP removal) are moved to the RU. It is understandable that the functional descriptions of the DU and RU corresponding to various types of eCPRI can be found in the eCPRI protocol, and will not be elaborated here.

[0131] In one possible implementation, the processing unit in the BBU used to implement baseband functions is called the baseband high (BBH) unit, and the processing unit in the RRU / AAU / RRH used to implement baseband functions is called the baseband low (BBL) unit.

[0132] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called O-CU (open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0133] It should be noted that, in the embodiments of this application, the device used to implement the functions of the network device can be a network device itself; it can also be a device capable of supporting the network device in implementing the functions, such as a chip system, hardware circuit, software module, or hardware circuit plus software module. This device can be installed in the network device or used in conjunction with the network device. In the embodiments of this application, the chip system can be composed of chips or may include chips and other discrete devices. In this embodiment, the device used to implement the functions of the network device is described as a network device, and this does not constitute a limitation on the solutions of the embodiments of this application.

[0134] Network devices and / or terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can also be deployed in the air on airplanes, balloons, and satellites. This application does not limit the scenario in which the network devices and terminal devices are located. Furthermore, terminal devices and network devices can be hardware devices, or software functions running on dedicated hardware or general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities that include dedicated or general-purpose hardware devices and software functions. This application does not limit the specific form of the terminal devices and network devices.

[0135] Please see Figure 1C , Figure 1C This is a schematic diagram of a possible application framework in a communication system provided in an embodiment of this application. For example... Figure 1C As shown, network elements in a communication system are connected via interfaces (e.g., NG, Xn) or air interfaces. These network element nodes, such as core network equipment, access network nodes (RAN nodes), terminals, or one or more devices in the operation administration and maintenance (OAM) system, are equipped with one or more AI modules (for clarity, ...). Figure 1C (Only one is shown in the text). The access network node can be a single RAN node or can include multiple RAN nodes, such as CU and DU. The CU and / or DU can also be configured with one or more AI modules. Optionally, the CU can also be split into CU-CP and CU-UP. One or more AI models are configured in CU-CP and / or CU-UP. The method described in the embodiments of this application can be applied to... Figure 1C The communication system shown.

[0136] Figure 1CThe AI ​​module in the network element is used to implement corresponding AI functions. The AI ​​modules deployed in different network elements can be the same or different. Depending on the parameter configuration, the AI ​​module can implement different functions. The AI ​​module model can be configured based on one or more of the following parameters: structural parameters (e.g., at least one of the following: number of neural network layers, neural network width, inter-layer connections, neuron weights, neuron activation function, or bias in the activation function), input parameters (e.g., type and / or dimension of input parameters), or output parameters (e.g., type and / or dimension of output parameters). The bias in the activation function can also be referred to as the neural network bias.

[0137] An AI module can have one or more models. A model can infer an output, which includes one or more parameters. The learning, training, or inference processes of different models can be deployed on different nodes or devices, or they can be deployed on the same node or device.

[0138] Optionally, the AI ​​module can also be called an AI network element or an AI node.

[0139] Optionally, the AI ​​node can be deployed in one or more of the following locations within the communication system: access network devices, terminal devices, or core network devices, etc. Alternatively, the AI ​​node can be deployed independently, for example, in a location other than any of the aforementioned devices, such as in the host or cloud server of an over-the-top (OTT) system. The AI ​​node can communicate with other devices in the communication system, which can be one or more of the following: network devices, terminal devices, or core network elements, etc.

[0140] It is understood that this application does not limit the number of AI nodes. For example, when there are multiple AI nodes, these nodes can be divided based on function, such as different AI nodes being responsible for different functions.

[0141] It can also be understood that AI nodes can be independent devices, integrated into the same device to implement different functions, or they can be network components in hardware devices, software functions running on dedicated hardware, or virtualization functions instantiated on a platform (e.g., a cloud platform). This application does not limit the specific form of the AI ​​nodes described above. It should be noted that the methods described in the embodiments of this application can be applied to... Figure 1C The aforementioned communication system.

[0142] Please see Figure 1D , Figure 1D This is a schematic diagram illustrating another possible application framework in the communication system provided in an embodiment of this application. For example... Figure 1DAs shown, the communication system includes a RAN intelligent controller (RIC). For example, the RIC can be... Figure 1C The AI ​​module shown is used to implement AI-related functions. The RIC includes near-real-time RIC (near-RT RIC) and non-real-time RIC (Non-RT RIC). Non-real-time RIC primarily processes non-real-time information, such as data that is not sensitive to latency, with latency in the order of seconds. Real-time RIC primarily processes near-real-time information, such as data that is relatively sensitive to latency, with latency in the order of tens of milliseconds. The method described in the embodiments of this application can be applied to... Figure 1D The communication system shown.

[0143] The near real-time RIC is used for model training and inference. For example, it can be used to train an AI model and then use that AI model for inference. The near real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data. Optionally, the near real-time RIC can deliver inference results to RAN nodes and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the near real-time RIC delivers the inference result to the DU, and the DU sends it to the RU.

[0144] The non-real-time RIC is also used for model training and inference. For example, it can be used to train an AI model and then use that model for inference. The non-real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data, and the inference results can be delivered to the RAN nodes and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the non-real-time RIC delivers the inference results to the DU, which then forwards them to the RU.

[0145] The near real-time RIC and non-real-time RIC can also be set up as separate network elements. Optionally, the near real-time RIC and non-real-time RIC can also be part of other devices. For example, the near real-time RIC can be set in the RAN node (e.g., in CU, DU), while the non-real-time RIC can be set in the OAM, cloud server, core network device, or other network device.

[0146] In another possible implementation, the network device can be a network device equipped with one or more AI modules. The network device can be... Figure 1C The core network equipment, access network node (RAN node), or one or more devices in the OAM are shown. For example, the AI ​​module can be... Figure 1D The RICs shown can be near real-time RICs or non-real-time RICs. For example, near real-time RICs are set up in RAN nodes (e.g., CUs, DUs), while non-real-time RICs are set up in OAMs, cloud servers, core network devices, or other network devices. The RIC can be trained by obtaining subsets from multiple terminal devices from RAN nodes (e.g., CUs, CU-CPs, CU-UPs, DUs, and / or RUs), recombining them into a training dataset #2, and then training based on training dataset #2. Exemplarily, near real-time RICs and non-real-time RICs can also be set up separately as network elements; the network device can be either a near real-time RIC or a non-real-time RIC.

[0147] To better understand the solutions provided in the embodiments of this application, some terms, concepts or processes involved in the embodiments of this application will be introduced below.

[0148] I. Terminology Explanation

[0149] (1) Artificial intelligence: to give machines human intelligence, using computer hardware and software to simulate certain intelligent behaviors of humans, including machine learning and many other methods.

[0150] (2) Machine learning (ML): Learning models or rules from raw data. There are many different machine learning methods, such as neural networks, decision trees, support vector machines, etc.

[0151] (3) AI model: Here it refers to a function model that maps an input of a certain dimension to an output of a certain dimension, and its model parameters are obtained through machine learning training. For example, f(x)=ax^2+b is a quadratic function model, which can be regarded as an AI model. a and b are the parameters of the model, which can be obtained through machine learning training.

[0152] (4) Neural network: Here it refers to artificial neural network, which is a mathematical model that imitates the behavior characteristics of animal neural networks to perform distributed parallel information processing. It is a special form of AI model.

[0153] (5) Deep neural network (DNN): A neural network with multiple hidden layers.

[0154] (6) Deep learning (DL): Machine learning using deep neural networks.

[0155] (7) Dataset: Data used for model training, validation and testing in machine learning. The quantity and quality of the data will affect the effect of machine learning.

[0156] (8) Model training: By selecting a suitable loss function, the model parameters are trained using an optimization algorithm to minimize the value of the loss function.

[0157] (9) Loss function: used to measure the difference between the model’s predicted value and the true value.

[0158] (10) Model testing: Evaluate model performance using test data after training.

[0159] (11) Model application: Use the trained model to solve practical problems.

[0160] II. Artificial Intelligence and Machine Learning

[0161] (1) Machine Learning:

[0162] Machine learning is an important technological approach to achieving artificial intelligence. Machine learning can be divided into supervised learning, unsupervised learning, and reinforcement learning.

[0163] (2) Supervised learning:

[0164] Supervised learning, based on collected sample values ​​and labels, uses machine learning algorithms to learn the mapping relationship between sample values ​​and labels, and expresses this learned mapping relationship using a machine learning model. The process of training the machine learning model is the process of learning this mapping relationship. For example, in signal detection, the noisy received signal is the sample, and the corresponding real constellation point is the label. Machine learning aims to learn the mapping relationship between samples and labels through training, that is, to enable the machine learning model to learn a signal detector. During training, the model parameters are optimized by calculating the error between the model's predicted values ​​and the real labels. Once the mapping relationship is learned, it can be used to predict the sample label of each new sample. The mapping relationship learned in supervised learning can include linear mappings and nonlinear mappings. Based on the type of label, the learning task can be divided into classification tasks and regression tasks.

[0165] (3) Unsupervised learning:

[0166] Unsupervised learning relies solely on collected sample values, using algorithms to discover inherent patterns within the samples. One type of unsupervised learning algorithm uses the samples themselves as supervisory signals; that is, the model learns the mapping relationship from sample to sample, which is called self-supervised learning. During training, model parameters are optimized by calculating the error between the model's predictions and the samples themselves. Self-supervised learning can be used for signal compression and decompression recovery applications; common algorithms include autoencoders and generative adversarial networks.

[0167] (4) Reinforcement learning:

[0168] Reinforcement learning, unlike supervised learning, is a type of algorithm that learns problem-solving strategies through interaction with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have explicit "correct" action labels. The algorithm needs to interact with the environment to obtain reward signals from the environment, and then adjust its decision actions to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user based on the total system throughput feedback from the wireless network, aiming to achieve a higher system throughput. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and the optimal decision action. However, because the label of the "correct action" cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action." Reinforcement learning training is achieved through iterative interaction with the environment.

[0169] (5) DNN

[0170] Deep Neural Networks (DNNs) are a specific implementation of machine learning. According to the general approximation theorem, neural networks can theoretically approximate any continuous function, thus enabling them to learn arbitrary mappings. Traditional communication systems rely on extensive expert knowledge to design communication modules, while DNN-based deep learning communication systems can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.

[0171] The idea behind DNNs (Dual Neural Networks) originates from the neuronal structure of the brain. Each neuron performs a weighted summation of its input values, and the result is passed through a non-linear function to produce the output. (See also...) Figure 2 , Figure 2 This is a schematic diagram of a neuron structure. Specifically, assume the input to the neuron is x = [x0, ..., x...]. n The weights corresponding to the inputs are w = [w0, ..., w0]. n The bias of the weighted summation is b. The nonlinear function can take many forms; one example is the max{0,x} maximum value function. The effect of a neuron's execution can be... DNNs typically have a multi-layered structure, with each layer containing multiple neurons. The input layer processes the received values ​​through neurons and then passes them to the hidden layers. Similarly, the hidden layers then pass the computation results to the final output layer, producing the final output of the DNN. (See also...) Figure 3 , Figure 3 This is a schematic diagram of a neural network structure.

[0172] DNNs typically have more than one hidden layer, and these hidden layers often directly affect the ability to extract information and fit functions. Increasing the number of hidden layers or widening the width of each layer can improve the function fitting ability of a DNN. The weights in each neuron are the parameters of the DNN network model. The model parameters are optimized through the training process, enabling the DNN network to extract data features and express mapping relationships. DNNs generally use supervised or unsupervised learning strategies to optimize model parameters.

[0173] Based on the way the network is constructed, DNNs can be divided into feedforward neural networks (FNN), convolutional neural networks (CNN), and recurrent neural networks (RNN). Figure 3 The image shows an FNN network, characterized by complete pairwise connections between neurons in adjacent layers. This makes FNNs typically require a large amount of storage space and result in high computational complexity.

[0174] CNNs are neural networks specifically designed to process data with a grid-like structure. For example, time-series data (discrete sampling along the time axis) and image data (two-dimensional discrete sampling) can both be considered grid-like data. CNNs do not use all the input information at once for computation; instead, they use a fixed-size window to extract a portion of the information for convolution operations, which significantly reduces the computational cost of model parameters. Furthermore, depending on the type of information extracted by the window (such as people and objects in an image representing different types of information), each window can use different convolution kernels, allowing CNNs to better extract features from the input data.

[0175] Recurrent Neural Networks (RNNs) are a type of distributed neural network (DNN) that utilizes feedback time-series information. Their input includes the current input value and their own output value from the previous time step. RNNs are well-suited for acquiring temporally correlated sequence features, and are particularly applicable to applications such as speech recognition and channel coding / decoding.

[0176] The FNN, CNN, and RNN mentioned above are common neural network structures, all built upon neurons. As introduced above, each neuron performs a weighted summation operation on its input values, and the result is passed through a nonlinear function to produce the output. We call the weights of the weighted summation operation and the nonlinear function in the neural network the parameters of the neural network. Taking a neuron with max{0,x} as the nonlinear function as an example, we perform... The parameters of the operated neuron are weights w = [w0, ..., w nThe weighted summation bias is b, and the nonlinear function is max{0,x}. The parameters of all neurons in a neural network constitute the parameters of that neural network.

[0177] III. Dataset

[0178] A training dataset is a collection of training samples, each serving as an input to the neural network. It's used for model training. The training dataset is one of the most crucial parts of machine learning; the training process essentially involves learning certain features from the training dataset to minimize the difference between the neural network's output and the ideal target value. Typically, even with the same network structure, neural networks trained on different training datasets will have different weights and outputs. Therefore, the composition and selection of the training dataset, to a certain extent, determine the performance of the trained neural network.

[0179] When AI models are deployed in an over-the-air (OTA) architecture, whether for offline or online model updates / training, data from the actual network deployment is required to form the dataset needed for model updates / training. A high-quality training dataset helps wireless communication AI algorithm design achieve greater performance gains and improves the generalization ability and robustness of the final algorithm across various scenarios. Conversely, a flawed training dataset can easily lead to inaccurate gain evaluation, model overfitting, weak generalization ability, and poor scenario adaptability, among other problems.

[0180] IV. Model Fusion

[0181] Model merging refers to combining multiple supervised fine-tuning (SFT) models at the parameter granularity to obtain a fused model. By aggregating multiple models, the fused model can achieve robust performance, thus avoiding the errors of individual models. Therefore, model merging can achieve effects similar to multi-task learning, meaning the fused model can simultaneously "learn" multiple tasks and may also achieve better in-domain performance and better out-of-distribution generalization.

[0182] Model fusion merges models at the parameter granularity, a method that may seem direct and even crude. However, it is supported by the redundancy of SFT parameter updates and the orthogonality of task vectors, giving it a certain degree of rationality. Even with current, relatively simple fusion methods, it has already achieved good results. Therefore, model fusion can help solve the aforementioned problem of weak generalization, thereby achieving adaptability to multiple scenarios.

[0183] V. Model Monitoring

[0184] Model monitoring refers to monitoring the performance of an AI model to determine if it is functioning correctly. If the model's performance is poor, it may be necessary to switch to non-AI mode, replace the model, or update it. Model monitoring can be achieved by monitoring the accuracy of the AI ​​model's output (also known as intermediate key performance indicators, KPIs) or by monitoring system performance (also known as eventual KPIs). Monitoring the accuracy of the AI ​​model's output involves comparing its output with the corresponding label or ground truth to determine if the model's performance meets requirements. Monitoring system performance involves monitoring whether the communication system's performance meets requirements after using the AI ​​model. Intermediate KPIs typically include generalized cosine similarity (GCS), square generalized cosine similarity (SGCS), and normalized mean squared error (NMSE). Final KPIs typically include throughput, spectral efficiency, transmission rate, block error rate (BLER), hypothetical BLER, and hybrid automatic repeat request (HARQ) feedback. Model monitoring can be performed by the UE or the base station.

[0185] The specific calculation method for intermediate KPIs is shown below.

[0186] When the KPI is GCS, then It is the predicted channel state information (CSI) for each resource element, w i The truth value CSI for each resource unit.

[0187] When the KPI is SGCS, then It is the predicted CSI for each resource unit, w i The truth value CSI for each resource unit.

[0188] When the KPI is the mean squared error (MSE), then It is the predicted CSI for each resource unit, w i The truth value CSI for each resource unit.

[0189] When the KPI is NMSE, then It is the predicted CSI for each resource unit, w i The truth value CSI for each resource unit.

[0190] Currently, with the development and application of AI technology, AI models have shown broad application prospects in many fields, covering industrial automation, gaming, agriculture, energy, environmental protection, and many others. However, when a model lacking certain channel characteristics is used for model monitoring, that is, when the model lacks certain channel characteristics, it can be understood that the model used does not match the channel characteristics of the actual environment, which can lead to performance fluctuations and insufficient generalization performance, resulting in significant performance degradation. To solve the above problems, the embodiments of this application propose the following solutions.

[0191] Please see Figure 4 , Figure 4 This is a schematic diagram of a model fusion method provided in an embodiment of this application. Figure 4 The method shown can be applied to both terminal-side devices and network-side devices. The terminal-side device can be a terminal device, a component within the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the terminal device's functions. The network-side device can be a network device, a component within the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the network device's functions. The following... Figure 4 The embodiments shown are described using terminal-side devices as terminal equipment and network-side devices as network equipment as examples. The method includes, but is not limited to, the following steps:

[0192] Step S401: The network device determines the second channel characteristics.

[0193] The second channel feature may include: the channel feature corresponding to the terminal device, the current channel feature, the predicted channel feature, or the historical channel feature. The process by which the network device determines the second channel feature may include: the network device acquiring the second channel feature by sending and receiving reference signals, including a sounding reference signal (SRS) and a channel state information reference signal (CSI-RS); alternatively, the network device receiving the second channel feature reported by the terminal device; or the network device determining it through algorithmic calculation. When the second channel feature includes the predicted channel feature, the network device can obtain the predicted channel feature through the output of an AI model, where the AI ​​model may be trained using historical channel features.

[0194] The determination of a second channel feature by the network device includes: the network device determining at least one second channel feature and different parameter ranges among the at least one second channel feature. The second channel feature includes at least one of the following: frequency selection, aging, angle spread (AS), or delay spread (DS). The determination of a second channel feature by the network device also includes: the network device determining at least two second channel features and the parameter range corresponding to each of the at least two second channel features. In one example, the determination of a second channel feature by the network device includes: frequency selection, a range 1 corresponding to frequency selection, angle spread, and a range 2 corresponding to angle spread, wherein range 1 is [Fmin0, Fmax0], and range 2 is [ASmin0, ASmax0].

[0195] Step S402: The network device determines the information of the model that needs to be fused based on the second channel characteristics and the information of the model supported by the terminal side.

[0196] Optionally, the information of the model supported by the terminal side can also be referred to as the level or category of the channel features supported by the terminal side model. The information of the model to be fused is related to the second channel feature and the information of the model supported by the terminal side. The information of the model to be fused is the information of one or more models from the information of the model supported by the terminal side.

[0197] The information for each model supported by the terminal includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature. This can be understood as the terminal supporting at least one model, where each model corresponds to at least one first channel feature, and each of these first channel features corresponds to a parameter range. The at least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or delay spread. Each model corresponds to at least two first channel features, and each of these at least two first channel features corresponds to a parameter range.

[0198] In one example, the terminal device supports two models: Model 1 and Model 2. The information of Model 1 includes two first channel features corresponding to Model 1 and the parameter range corresponding to each of the two first channel features. For example, the two first channel features corresponding to Model 1 include frequency selection and angle spread. The parameter range corresponding to each of the two first channel features includes a range 3 corresponding to frequency selection and a range 4 corresponding to angle spread. The range 3 is [Fmin1, Fmax1] and the range 4 is [ASmin1, ASmax1]. In summary, the information of Model 1 includes frequency selection, the range 3 corresponding to frequency selection, angle spread, and the range 4 corresponding to angle spread. The range 3 is [Fmin1, Fmax1] and the range 4 is [ASmin1, ASmax1]. The information of Model 2 includes two first channel features corresponding to Model 2, and the parameter range corresponding to each of the two first channel features. For example, the two first channel features corresponding to Model 2 include frequency selection and angle spread. The parameter range corresponding to each of the two first channel features includes: range 5 corresponding to frequency selection and range 6 corresponding to angle spread. Among them, range 5 is [Fmin2, Fmax2] and range 6 is [ASmin2, ASmax2]. In summary, the information of Model 2 includes frequency selection, the range 5 corresponding to frequency selection, angle spread and the range 6 corresponding to angle spread. Among them, range 5 is [Fmin2, Fmax2] and range 6 is [ASmin2, ASmax2].

[0199] In one possible implementation, the second channel feature includes: a first feature, a first parameter range corresponding to the first feature, a second feature, and a second parameter range corresponding to the second feature. The information of the model supported by the terminal side includes information of a first model. The information of the first model includes: a first feature, a third parameter range corresponding to the first feature, a second feature, and a fourth parameter range corresponding to the second feature. The first feature and the second feature are at least one first channel feature corresponding to the first model, and the third parameter range and the fourth parameter range are the parameter ranges of each channel feature in the at least one first channel feature corresponding to the first model. Based on the second channel feature and the information of the model supported by the terminal side, the information of the model to be fused is determined, including: the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than a third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than a fourth threshold. The information of the model to be fused includes the identification information of the first model.

[0200] Optionally, the aforementioned distances can be determined by the absolute value of the difference. For example, the distance between the lower limit of the third parameter range and the lower limit of the first parameter range being less than a first threshold includes cases where the absolute value of the difference between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold. Optionally, the distance between the lower limit of the third parameter range and the lower limit of the first parameter range being less than a first threshold includes cases where the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than or equal to a first threshold; the distance between the upper limit of the third parameter range and the upper limit of the first parameter range being less than a second threshold includes cases where the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than or equal to a second threshold; the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range being less than a third threshold includes cases where the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than or equal to a third threshold; and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range being less than a fourth threshold includes cases where the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than or equal to a fourth threshold. Optionally, the first threshold and the second threshold can be the same or different. The third threshold and the fourth threshold can be the same or different, and this application embodiment does not limit this. Optionally, the first threshold, the second threshold, the third threshold, and the fourth threshold may be determined by the network device, or predefined by the protocol, or determined through negotiation between the network device and the terminal device. This application embodiment does not impose any limitations on these thresholds.

[0201] The second channel feature includes: a first feature, a first parameter range corresponding to the first feature, a second feature, and a second parameter range corresponding to the second feature. At least one second channel feature includes the first feature and the second feature. The parameter range of each second channel feature in the at least one second channel feature includes the first parameter range corresponding to the first feature and the second parameter range corresponding to the second feature.

[0202] The above process can be understood as follows: when one or more of the following conditions are met: the range of the third parameter is an adjacent range of the range of the first parameter; the range of the third parameter is a subset of the range of the first parameter; the range of the third parameter overlaps with the range of the first parameter; the range of the fourth parameter is an adjacent range of the range of the second parameter; the range of the fourth parameter is a subset of the range of the second parameter; or the range of the fourth parameter overlaps with the range of the second parameter, the information of the model to be fused includes the identification information of the first model. In one example, when the range of the third parameter is [0,1], the range of the first parameter is [1,2], and the range of the third parameter is an adjacent range of the first parameter. In another example, when the range of the third parameter is [0,2], the range of the first parameter is [1,3], and the range of the third parameter overlaps with the range of the first parameter. In yet another example, when the range of the third parameter is [0,1], the range of the first parameter is [0,2], and the range of the third parameter is a subset of the first parameter.

[0203] In one example, the second channel feature includes: frequency selection, a frequency selection corresponding range 1, angle spread, and an angle spread corresponding range 2, where range 1 is [Fmin0, Fmax0] and range 2 is [ASmin0, ASmax0]. The terminal device supports models 1 and 2. The information of the models supported by the terminal device includes information from model 1 and model 2. Model 1 information includes: frequency selection, a frequency selection corresponding range 3, angle spread, and an angle spread corresponding range 4, where range 3 is [Fmin1, Fmax1] and range 4 is [ASmin1, ASmax1]. Model 2 information includes: frequency selection, a frequency selection corresponding range 5, angle spread, and an angle spread corresponding range 6, where range 5 is [Fmin2, Fmax2] and range 6 is [ASmin2, ASmax2]. Model 3 information includes: frequency selection, a frequency selection corresponding range 7, angle spread, and an angle spread corresponding range 8, where range 7 is [Fmin3, Fmax3] and range 6 is [ASmin3, ASmax3].

[0204] For example, [Fmin0, Fmax0] = [0, 1], [ASmin0, ASmax0] = [1, 3], [Fmin1, Fmax1] = [0, 1], [ASmin1, ASmax1] = [2, 3], [Fmin2, Fmax2] = [3, 4], [ASmin2, ASmax2] = [4, 5], [Fmin3, Fmax3] = [0, 1], [ASmin3, ASmax3] = [1, 2]

[0205] For Model 1, Fmin1 (lower limit of the third parameter range) - Fmin0 (lower limit of the first parameter range) = 0 - 0 = 0, |0| = 0 < 1 (first threshold), Fmax1 (upper limit of the third parameter range) - Fmax0 (upper limit of the first parameter range) = 1 - 1 = 0, |0| = 0 < 1 (second threshold), ASmin1 (lower limit of the fourth parameter range) - ASmin0 (lower limit of the second parameter range) = 2 - 1 = 1, |1| = 1 < 2 (third threshold), ASmax1 (upper limit of the fourth parameter range) - ASmax0 (upper limit of the second parameter range) = 3 - 3 = 0, |0| = 0 < 1 (fourth threshold). Therefore, the information of the model to be fused includes the information of Model 1.

[0206] For Model 2, Fmin2 (lower limit of the third parameter range) - Fmin0 (lower limit of the first parameter range) = 3 - 0 = 3, |3| = 3 > 1 (first threshold), Fmax1 (upper limit of the third parameter range) - Fmax0 (upper limit of the first parameter range) = 4 - 1 = 3, |3| = 3 > 1 (second threshold), ASmin1 (lower limit of the fourth parameter range) - ASmin0 (lower limit of the second parameter range) = 4 - 1 = 3, |3| = 3 > 2 (third threshold), ASmax1 (upper limit of the fourth parameter range) - ASmax0 (upper limit of the second parameter range) = 5 - 3 = 2, |2| = 2 > 1 (fourth threshold). Therefore, the information of the model to be fused does not include the information of Model 2.

[0207] For Model 3, Fmin3 (lower limit of the third parameter range) - Fmin0 (lower limit of the first parameter range) = 0 - 0 = 0, |0| = 0 < 1 (first threshold), Fmax3 (upper limit of the third parameter range) - Fmax0 (upper limit of the first parameter range) = 1 - 1 = 0, |0| = 0 < 1 (second threshold), ASmin3 (lower limit of the fourth parameter range) - ASmin0 (lower limit of the second parameter range) = 1 - 1 = 0, |0| = 0 < 2 (third threshold), ASmax3 (upper limit of the fourth parameter range) - ASmax0 (upper limit of the second parameter range) = 2 - 3 = -1, |-1| = 1 = 1 (fourth threshold). Therefore, the information of the model to be fused includes the information of Model 3.

[0208] Therefore, in summary, the network device determines that the models that need to be merged include Model 1 and Model 3. The network device sends a first instruction information to the terminal device, which includes the identification information of Model 1 and the identification information of Model 3.

[0209] For example, [Fmin0, Fmax0] = [1,2], [ASmin0, ASmax0] = [1,5], [Fmin1, Fmax1] = [0,1], [ASmin1, ASmax1] = [2,3], [Fmin2, Fmax2] = [3,4], [ASmin2, ASmax2] = [4,5], [Fmin3, Fmax3] = [0,1], [ASmin3, ASmax3] = [1,2]. The network device determines that the information of the models that need to be merged includes the information of model 1, model 2 and model 3. The models that need to be merged include model 1, model 2 and model 3. The network device sends a first indication information to the terminal device. The first indication information includes the identification information of model 1, model 2 and model 3.

[0210] In yet another example, see Figure 5 , Figure 5This is a schematic diagram of a second channel feature and a model supported by the terminal device, provided in an embodiment of this application. The second channel feature includes: angle spread, with the parameter range corresponding to the angle spread being [1,2], and delay spread, with the parameter range corresponding to the delay spread being [1,2]. The information of the model supported by the terminal device includes: information of model 1, model 2, model 3, model 4, model 5, model 6, model 7, and model 8. Among them, the information of model 1 includes: angle spread, with the parameter range corresponding to the angle spread being [0,1], delay spread, and delay spread... The corresponding parameter range is [0,1]. Model 2 information includes: angle spread, with a parameter range of [1,2], delay spread, and a parameter range of [0,1]. Model 3 information includes: angle spread, with a parameter range of [2,3], delay spread, and a parameter range of [0,1]. Model 4 information includes: angle spread, with a parameter range of [0,1], delay spread, and a parameter range of [1,2]. Model 5 information includes: angle spread, with a parameter range of [0,1]. [2,3], delay spread, the parameter range corresponding to delay spread is [1,2]. The information of model 6 includes: angle spread, the parameter range corresponding to angle spread is [0,1], delay spread, the parameter range corresponding to delay spread is [2,3]. The information of model 7 includes: angle spread, the parameter range corresponding to angle spread is [1,2], delay spread, the parameter range corresponding to delay spread is [2,3]. The information of model 8 includes: angle spread, the parameter range corresponding to angle spread is [2,3], delay spread, the parameter range corresponding to delay spread is [2,3]. The network device is based on the second channel. The information of the features and terminal-side supported models determines the information of the models that need to be fused, including information of model 1, model 2, model 3, model 4, model 5, model 6, model 7, and model 8. The models that need to be fused include models 1 to 8. The network device sends first indication information to the terminal device. The first indication information includes the identification information of model 1, model 2, model 3, model 4, model 5, model 6, model 7, and model 8.

[0211] In another possible implementation, the network device determines the information of the models that need to be fused based on the second channel characteristics and the information of the models supported by the terminal side, as well as at least one of the following, which includes: terminal side model fusion capability information, or upper limit information on the number of models that need to be fused transmitted.

[0212] In addition to being related to the second channel characteristics and the information of the models supported by the terminal side, the information of the models that need to be fused is also related to at least one of the following, which includes: terminal side model fusion capability information, or upper limit information of the number of models that need to be fused transmitted.

[0213] Among them, the terminal-side model fusion capability information includes the maximum number of models that the terminal can fuse, and the upper limit information on the number of models that need to be fused for transmission includes the upper limit of the number of models that need to be fused for transmission supported by the network side.

[0214] Optionally, before the network device determines the model information that needs to be fused, the terminal device sends terminal-side model fusion capability information to the network device.

[0215] Specifically, when the network device determines the information of the models to be fused based on the second channel characteristics, the information of the models supported by the terminal side, the terminal side's model fusion capability information, and the upper limit information of the number of models to be fused in transmission, the minimum value between the terminal side's model fusion capability information and the upper limit information of the number of models to be fused in transmission can be taken. In one example, the terminal side's model fusion capability information includes that the terminal side can fuse a maximum of 4 models, and the upper limit information of the number of models to be fused in transmission includes that the number of models to be fused in transmission is a maximum of 10. Therefore, min{10,4} = 4. Optionally, the network device also needs to send indication information to the terminal device. This indication information is used to indicate the 4 models with higher priority among the 10 models to be fused, for example, the indices of these 4 models are i1, i2, i3, and i4, respectively.

[0216] In the above method, when determining the information of the models that need to be fused, the network device considers the constraints of actual transmission, namely the upper limit information of the number of models that need to be fused, and the constraints of terminal capabilities, namely the terminal-side model fusion capability information, thereby ensuring the generalization performance of the models.

[0217] In another possible implementation, the network device receives information about the model supported by the terminal side from the terminal device.

[0218] Optionally, the information of the model supported by the terminal side can be determined by the terminal device based on the channel characteristics of the training data. Optionally, the terminal device can also receive information of the model supported by the network side from the network device, which is determined by the network device based on the channel characteristics of the training data. Optionally, the information of the model supported by the network side can also be referred to as the level or category of the channel characteristics supported by the network-side model. For specific information about the model supported by the network side, please refer to the information of the model supported by the terminal side, which will not be repeated here. It should be noted that the information of the model supported by the terminal side and the information of the model supported by the network side can be the same or different, and this application embodiment does not limit this. In one example, please refer to... Figure 6 , Figure 6 (a) is a schematic diagram provided in this application embodiment of a terminal device sending the channel feature levels supported by the terminal-side model to a network device. The channel feature levels supported by the terminal-side model include level 1, level 2, ... level N, and may include: information of the models supported by the terminal side, including information of model 1, information of model 2, ..., information of model N, wherein level 1 corresponds to the information of model 1, and the information of model 1 includes at least one first channel feature corresponding to model 1, and the parameter range of each first channel feature in the at least one first channel feature corresponding to model 1; level 2 corresponds to the information of model 2, and the information of model 2 includes at least one first channel feature corresponding to model 2, and the parameter range of each first channel feature in the at least one first channel feature corresponding to model 2, and so on, until the information of model N corresponding to level N can be referred to the above description, and will not be repeated here. Please refer to Figure 6 (b) is a schematic diagram of a network device sending channel feature levels supported by a network-side model to a terminal device according to an embodiment of this application. The channel feature levels supported by the network-side model include level 1, level 2, ..., level N. For details, please refer to the relevant description in the information of the network-side supported model, which will not be repeated here.

[0219] In the above method, by having the network device and the terminal device exchange information about the models they support, the model capabilities of both sides can be aligned. Furthermore, this facilitates subsequent use, that is, the network device determines the information of the model to be fused based on the information of the model supported by the terminal side, and then instructs the terminal device accordingly.

[0220] Step S403: The network device sends the first instruction information to the terminal device.

[0221] Correspondingly, the terminal device receives the first instruction information from the network device.

[0222] The first indication information is used to indicate information about the models to be fused, wherein the models to be fused correspond to different parameter ranges of at least one first channel feature. The different parameter ranges of the models to be fused corresponding to at least one first channel feature include: each model to be fused corresponds to at least one first channel feature, and each first channel feature corresponds to a parameter range. In one possible implementation, the first indication information includes identification information of the models to be fused, which may be index information. In another possible implementation, the first indication information includes information about the models to be fused. In yet another possible implementation, the first indication information includes both identification information and information about the models to be fused.

[0223] In one example, the first indication information includes the identification information of model 1 and the identification information of model 3, wherein the identification information of model 1 is ID1 and the identification information of model 3 is ID3.

[0224] In another example, the first indication information includes information about model 1 and information about model 3. The information about model 1 includes: frequency selection, the range 3 corresponding to the frequency selection, angle extension, and the range 4 corresponding to the angle extension. The range 3 is [Fmin1, Fmax1] and the range 4 is [ASmin1, ASmax1]. The information about model 3 includes: frequency selection, the range 7 corresponding to the frequency selection, angle extension, and the range 8 corresponding to the angle extension. The range 7 is [Fmin3, Fmax3] and the range 6 is [ASmin3, ASmax3].

[0225] In another example, the first indication information includes the identification information of model 1, the information of model 1, the identification information of model 3, and the information of model 3. The identification information of model 1 is identifier 1. The information of model 1 includes: frequency selection, the range 3 corresponding to frequency selection, angle extension, and the range 4 corresponding to angle extension. Among them, the range 3 is [Fmin1, Fmax1], and the range 4 is [ASmin1, ASmax1]. The identification information of model 3 is ID3. The information of model 3 includes frequency selection, the range 7 corresponding to frequency selection, angle extension, and the range 8 corresponding to angle extension. Among them, the range 7 is [Fmin3, Fmax3], and the range 6 is [ASmin3, ASmax3].

[0226] Step S404: The terminal device performs model fusion based on the first indication information to determine the fused model.

[0227] In this process, the terminal device performs model fusion on the models indicated by the first indication information to determine the fused model. In one example, the first indication information includes the identification information of model 1 and model 3. The terminal device performs model fusion on model 1 and model 3 to determine the fused model. Model fusion involves merging multiple supervised fine-tuning models at the parameter granularity to obtain a fused model. Using this method, robustness can be maintained even when the models supported by the terminal side lack certain channel features, ensuring performance on the missing channel features of the models.

[0228] The fused model can be used for channel estimation, channel coding and decoding, beam management, and channel compression feedback.

[0229] exist Figure 4 The described method, by employing the aforementioned approach, avoids the problem of a mismatch between the single model (one of the models supported by the terminal) and the actual channel characteristics (i.e., the missing channel characteristics) in situations where the model supported by the terminal lacks certain channel features. This mismatch prevents the single model from achieving ideal performance. By fusing existing models (the models that need to be fused among the models supported by the terminal) to obtain a fused model, the generalization ability of model fusion allows for good performance even with missing channel features. In summary, it ensures the generalization performance of the model, thereby guaranteeing its overall performance. Furthermore, using the fused model results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0230] Please see Figure 7 , Figure 7 This is a schematic diagram of yet another model fusion method provided in the embodiments of this application. Figure 7 The method shown can be applied to both terminal-side devices and network-side devices. The terminal-side device can be a terminal device, a component within the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the terminal device's functions. The network-side device can be a network device, a component within the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the network device's functions. The following... Figure 7 The embodiments shown are described using terminal-side devices as terminal equipment and network-side devices as network equipment as examples. The method includes, but is not limited to, the following steps:

[0231] Step S701: The network device determines the first information.

[0232] The first information is used to indicate at least one second channel feature, the at least one second channel feature corresponding to different parameter ranges, and the weights corresponding to the at least one second channel feature. The second channel feature includes at least one of the following: frequency selection, aging, angle spread (AS), or delay spread (DS).

[0233] The first information is used to indicate at least two second channel features, each of the at least two second channel features corresponding to a parameter range, and the weights corresponding to the at least two second channel features. In one possible implementation, if the weights corresponding to the at least two second channel features are not included in the second channel features, it is assumed that the weights of each of the at least two second channel features are equal.

[0234] In one example, the first information includes: angle expansion, the parameter range corresponding to the angle expansion [0,1], delay expansion, the parameter range corresponding to the delay expansion [0,1], and a weight of 0.6; angle expansion, the parameter range corresponding to the angle expansion [0,1], delay expansion, the parameter range corresponding to the delay expansion [1,2], and a weight of 0.4.

[0235] In yet another example, see Figure 8 , Figure 8 This is a schematic diagram of a first indication information provided in an embodiment of this application. The numbers in the squares represent the weights corresponding to at least one second feature. The sum of all the numbers in the squares is 1. A square with a number represents the time delay spread, the parameter range corresponding to the time delay spread, the angle spread, the parameter range corresponding to the angle spread, and the corresponding weight. Taking a square with a number of 0.03 as an example, the information presented by this square includes: time delay spread, the parameter range [0,1] corresponding to the time delay spread, the angle spread, the parameter range [0,1] corresponding to the angle spread, and the corresponding weight of 0.03. The first indication information includes the information presented by all the squares with numbers.

[0236] Step S702: The network device sends the first information to the terminal device.

[0237] Correspondingly, the terminal device receives the first information from the network device.

[0238] Step S703: The terminal device determines the model to be fused based on at least one second channel feature, the different parameter ranges corresponding to at least one second channel feature, and the information of the model supported by the terminal side.

[0239] The information of each model in the information of the models supported by the terminal side includes: at least one first channel feature corresponding to each model, or the parameter range of each first channel feature in the at least one first channel feature corresponding to each model. For details, please refer to the relevant description in the information of the models supported by the terminal side in step S402.

[0240] In one possible implementation, the first information includes: a first feature, a first parameter range corresponding to the first feature, a second feature, a second parameter range corresponding to the second feature, and a first weight, wherein the first feature and the second feature are one of the at least one second channel features, the first parameter range and the second parameter range are different parameter ranges corresponding to the at least one second channel feature, and the first weight is the weight corresponding to the at least one second channel feature; the information of the model supported by the terminal side includes information of a first model, the information of the first model including: the first feature, a third parameter range corresponding to the first feature, the second feature, and a fourth parameter range corresponding to the second feature, wherein the first feature and the second feature are one of the at least one first channel features, and the third parameter range and the fourth parameter range are the parameter ranges of each first channel feature in the at least one first channel feature; determining the model to be fused based on the at least one second channel feature, the different parameter ranges corresponding to the at least one second channel feature, and the information of the model supported by the terminal side includes:

[0241] If the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than a third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than a fourth threshold, then the model to be fused includes the first model.

[0242] Optionally, the aforementioned distances can be determined by the absolute value of the difference. For example, the distance between the lower limit of the third parameter range and the lower limit of the first parameter range being less than a first threshold includes cases where the absolute value of the difference between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold. Optionally, the distance between the lower limit of the third parameter range and the lower limit of the first parameter range being less than a first threshold includes cases where the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than or equal to a first threshold; the distance between the upper limit of the third parameter range and the upper limit of the first parameter range being less than a second threshold includes cases where the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than or equal to a second threshold; the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range being less than a third threshold includes cases where the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than or equal to a third threshold; and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range being less than a fourth threshold includes cases where the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than or equal to a fourth threshold. Optionally, the first threshold and the second threshold can be the same or different. The third threshold and the fourth threshold can be the same or different, and this application embodiment does not limit this. Optionally, the first threshold, the second threshold, the third threshold, and the fourth threshold may be determined by the network device and sent to the terminal device, or predefined by the protocol, or determined by the terminal device, or determined through negotiation between the network device and the terminal device. This application embodiment does not limit the specific threshold.

[0243] In one example, the first information includes:

[0244] Angle spread, the parameter range corresponding to angle spread [0,1], delay spread, the parameter range corresponding to delay spread [0,1], weight is 0.6;

[0245] Angle spread, the parameter range corresponding to angle spread is [0,1], delay spread, the parameter range corresponding to delay spread is [1,2], and the weight is 0.4.

[0246] The terminal-side supports information about models including Model 1, Model 2, and Model 3.

[0247] The information for Model 1 includes: angle spread, the parameter range [0,1] corresponding to the angle spread, delay spread, and the parameter range [0,1] corresponding to the delay spread.

[0248] The information for Model 2 includes angle spread, the parameter range corresponding to the angle spread [0,1], delay spread, and the parameter range corresponding to the delay spread [1,2].

[0249] The information for Model 3 includes angle spread, the parameter range corresponding to the angle spread [5,6], delay spread, and the parameter range corresponding to the delay spread [7,8].

[0250] The terminal device determines the models that need to be fused, including Model 1 and Model 2, based on the first information and the information of the models supported by the terminal side.

[0251] In yet another example, see Figure 9 , Figure 9 This is a schematic diagram illustrating how to determine the models that need to be fused, as provided in an embodiment of this application. The first information includes:

[0252] Angle spread, the parameter range corresponding to angle spread [0,1], delay spread, the parameter range corresponding to delay spread [0,1], weight is 0.6;

[0253] Angle spread, the parameter range for angle spread is [1,2], delay spread, the parameter range for delay spread is [1,2], and the weight is 0.4. Specifically, as follows... Figure 9 The squares marked with a ☆ are shown in the image.

[0254] The terminal-side supports information on models including Model 1, Model 2, Model 3, Model 4, and Model 5, specifically as follows: Figure 9 The filled squares in the diagram are shown below, where...

[0255] The information for Model 1 includes: angle spread, the parameter range [0,1] corresponding to the angle spread, delay spread, and the parameter range [0,1] corresponding to the delay spread.

[0256] The information for Model 2 includes angle spread, the parameter range corresponding to the angle spread [1,2], delay spread, and the parameter range corresponding to the delay spread [0,1].

[0257] The information for Model 3 includes: angle spread, the parameter range corresponding to the angle spread [0,1], delay spread, and the parameter range corresponding to the delay spread [1,2].

[0258] The information for Model 4 includes angle spread, the parameter range corresponding to the angle spread [2,3], delay spread, and the parameter range corresponding to the delay spread [1,2].

[0259] The information for Model 5 includes angle spread, the parameter range corresponding to the angle spread [1,2], delay spread, and the parameter range corresponding to the delay spread [2,3].

[0260] The terminal device determines the models that need to be fused based on the first information and the information of the models supported by the terminal side, including Model 1, Model 2, Model 3, Model 4 and Model 5.

[0261] Step S704: The terminal device performs model fusion based on the weights corresponding to at least one second channel feature and the model to be fused to determine the fused model.

[0262] The process of determining the fused model by model fusion based on the weights corresponding to at least one second channel feature and the model to be fused may include: the terminal device determining the weight values ​​corresponding to the model to be fused using an algorithm based on the weights corresponding to at least one second channel feature and the model to be fused, and then performing model fusion based on the model to be fused and the weight values ​​corresponding to the model to be fused to determine the fused model. Optionally, the fused model can be used in channel estimation, channel encoding / decoding, beam management, and channel compression feedback.

[0263] In one possible implementation, the first information includes: a first feature, a first parameter range corresponding to the first feature, a second feature, a second parameter range corresponding to the second feature, and a first weight. The model to be fused includes a first model, and the information of the first model includes: the first feature, a third parameter range corresponding to the first feature, the second feature, and a fourth parameter range corresponding to the second feature. When the third parameter range is the same as the first parameter range, and the fourth parameter range is the same as the second parameter range, the first model is fused according to the first weight during model fusion. When the third parameter range is different from the first parameter range, and / or the fourth parameter range is different from the second parameter range, the weight value corresponding to the first model is calculated according to the algorithm during model fusion.

[0264] In one example, see Figure 9 The terminal device determines the models to be fused, including Model 1, Model 2, Model 3, Model 4, and Model 5, based on the first information and the information of the models supported by the terminal side. In one possible implementation, when fusing models, the weight corresponding to Model 1 is 0.6. Then, the weight allocation ratio of Model 2, Model 3, Model 4, and Model 5 is calculated according to the algorithm, with the total weight sum being 0.4. For example, the weight corresponding to Model 2 is 0.1, the weight corresponding to Model 3 is 0.05, the weight corresponding to Model 4 is 0.2, and the weight corresponding to Model 5 is 0.05. Then, based on the models to be fused and the weight corresponding to each model, the fused model is determined. In another possible implementation, when fusing models, the weight corresponding to Model 1 is 0.6. The proportion of Model 1 with a weight of 0.6 can be increased according to the algorithm calculation, while the proportion of the total weight of Model 2, Model 3, Model 4, and Model 5 (0.4) is relatively reduced, and then the model fusion is performed.

[0265] exist Figure 7The described method, by employing the aforementioned approach, avoids the problem of a mismatch between the single model (one of the models supported by the terminal) and the actual channel characteristics (i.e., the missing channel characteristics) when the model supported by the terminal lacks certain channel features. This means that when the training channel data of a single model does not match the actual channel characteristics, the single model cannot achieve ideal performance. The terminal device can use the first information indicated by the network device to perform model fusion with existing models (the models that need to be fused among the models supported by the terminal) to obtain a fused model. Through the generalization capability of model fusion, good performance can be achieved even when channel features are missing. In summary, the generalization performance of the model is guaranteed, thereby ensuring the model's performance. Furthermore, using the fused model results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0266] It should be noted that the embodiments of this application can be applied to model inference.

[0267] Please see Figure 10 , Figure 10 This is a schematic diagram of yet another model fusion method provided in the embodiments of this application. Figure 10 The method shown can be applied to both terminal-side devices and network-side devices. The terminal-side device can be a terminal device, a component within the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the terminal device's functions. The network-side device can be a network device, a component within the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the network device's functions. The following... Figure 10 The embodiments shown are described using terminal-side devices as terminal equipment and network-side devices as network equipment as examples. The method includes, but is not limited to, the following steps:

[0268] Step S1001: The network device obtains channel characteristics based on the channel.

[0269] This step may include: the network device determining the second channel characteristics. See the relevant description in step S401 for details.

[0270] Step S1002: The network device informs the terminal device to perform model fusion and provides the identification information of the models to be fused.

[0271] This step may include: the network device determining the information of the model to be fused based on the second channel characteristics and the information of the model supported by the terminal side; the network device sending first indication information to the terminal device, the first indication information being used to indicate the information of the model to be fused. For details, please refer to the relevant descriptions in steps S402 and S403.

[0272] Step S1003: The terminal device performs model fusion based on the identification information of the models to be fused to determine the fused model.

[0273] This step may include: the terminal device performing model fusion based on the first indication information to determine the fused model. For details, please refer to the relevant description in step S404.

[0274] exist Figure 10 The described method, by employing the above approach, avoids the problem of a mismatch between the single model (one of the models supported by the terminal) and the channel characteristics of the actual environment (i.e., the missing channel characteristics) when the model supported by the terminal lacks certain channel features. This means that when the training channel data of a single model does not match the actual channel characteristics, the single model cannot achieve ideal performance. By using existing models (the models that need to be fused among the models supported by the terminal) to perform model fusion, a fused model is obtained. Through the generalization ability of model fusion, good performance can be achieved even when channel features are missing. In summary, the generalization performance of the model can be guaranteed, thereby guaranteeing the model's performance. Furthermore, using the fused model results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0275] Please see Figure 11 , Figure 11 This is a schematic diagram of yet another model fusion method provided in the embodiments of this application. Figure 11 The method shown can be applied to both terminal-side devices and network-side devices. The terminal-side device can be a terminal device, a component within the terminal device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the terminal device's functions. The network-side device can be a network device, a component within the network device (e.g., a processor, chip, circuit, or chip system), or a logic module or software capable of implementing all or part of the network device's functions. The following... Figure 11 The embodiments shown are described using terminal-side devices as terminal equipment and network-side devices as network equipment as examples. The method includes, but is not limited to, the following steps:

[0276] Step S1101: The network device obtains channel characteristics based on the channel.

[0277] This step may include: the network device determining first information, the first information indicating at least one second channel feature, the at least one second channel feature corresponding to different parameter ranges, and the weight corresponding to the at least one second channel feature. See the relevant description in step S701 for details.

[0278] Step S1102: The network device sends the channel characteristics and corresponding weights to the terminal device.

[0279] This step may include: the network device sending first information to the terminal device. See the relevant description in step S702 for details.

[0280] Step S1103: The terminal device performs model fusion based on channel characteristics, corresponding weights, and information from the models supported by the terminal side to determine the fused model.

[0281] This step may include: the terminal device determining the model to be fused based on at least one second channel feature, the different parameter ranges corresponding to the at least one second channel feature, and information about the models supported by the terminal side; and the terminal device determining the fused model based on the weights corresponding to the at least one second channel feature and the model to be fused. See the relevant descriptions in steps S703 and S704 for details.

[0282] exist Figure 11 The described method, by employing the aforementioned approach, avoids the problem of a mismatch between the single model (one of the models supported by the terminal) and the actual channel characteristics (i.e., the missing channel characteristics) when the model supported by the terminal lacks certain channel features. This means that when the training channel data of a single model does not match the actual channel characteristics, the single model cannot achieve ideal performance. The terminal device can use the first information indicated by the network device to perform model fusion with existing models (the models that need to be fused among the models supported by the terminal) to obtain a fused model. Through the generalization capability of model fusion, good performance can be achieved even when channel features are missing. In summary, the generalization performance of the model is guaranteed, thereby ensuring the model's performance. Furthermore, using the fused model results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0283] It should be noted that this application takes the case of a terminal device or network device lacking a certain channel feature as an example, that is, the model used lacks certain channel features. This can be understood as the model used not matching the channel features of the actual environment. Model fusion is used to solve this problem. However, this application may also be applicable to other missing scenarios that may occur in the future. The embodiments of this application are not limited.

[0284] It should be noted that when the embodiments of this application are applied to an open RAN architecture, for example, Figure 1D The communication system shown in this application embodiment. The deployment of the AI ​​model on the network side can be in the same physical entity as the network device, or it can be in a different physical entity. The deployment of the AI ​​model on the terminal side can be in the same physical entity as the network device, or it can be in a different physical entity. For example, the information of the model supported by the terminal side and the model fusion on the terminal side can be implemented on the chip inside the terminal device, or it can be located outside the terminal device, such as in the host or motion server of the OTT system. The deployment on the network side can be implemented on the chip inside the network device, or it can be located outside the network device, such as in the intelligent network element. The near real-time RIC (located in the RAN node, such as in the CU / DU) is collectively referred to as the network-side AI model deployment device of the intelligent network element. Taking a communication system including intelligent network elements, network devices, terminal devices, and OTT as an example, and using... Figure 4 For an example of the method implementation, please refer to [link / reference]. Figure 12 , Figure 12 This is a schematic diagram of another model fusion method provided in an embodiment of this application, which includes, but is not limited to, the following steps:

[0285] Step S1201: The network device determines the second channel characteristics.

[0286] Please refer to the relevant description in step S401 for details.

[0287] Step S1202: The network device sends the second channel feature to the intelligent network element.

[0288] Step S1203: The intelligent network element determines the information of the model that needs to be fused based on the second channel characteristics and the information of the model supported by the terminal side.

[0289] For details, please refer to the relevant description in step S402.

[0290] Step S1204: The intelligent network element sends the first instruction information to the terminal device through the network device.

[0291] Accordingly, the terminal device receives the first instruction information.

[0292] The first indication information is used to indicate the information of the model to be fused, and the model to be fused corresponds to at least one different parameter range of a first channel feature. See the relevant description in step S403 for details.

[0293] Step S1205: The terminal device sends the first instruction information to the OTT.

[0294] Step S1206: OTT performs model fusion based on the first indication information to determine the fused model.

[0295] For details, please refer to the relevant description in step S404.

[0296] exist Figure 12 The described method, by employing the aforementioned approach, avoids the problem of a mismatch between the single model (one of the models supported by the terminal) and the actual channel characteristics (i.e., the missing channel characteristics) in situations where the model supported by the terminal lacks certain channel features. This mismatch prevents the single model from achieving ideal performance. By fusing existing models (the models that need to be fused among the models supported by the terminal) to obtain a fused model, the generalization ability of model fusion allows for good performance even with missing channel features. In summary, it ensures the generalization performance of the model, thereby guaranteeing its overall performance. Furthermore, using the fused model results in a more robust model compared to a single model, and the error of the fused model is smaller than that of a single model.

[0297] The methods of the embodiments of this application have been described in detail above, and the apparatus of the embodiments of this application is provided below.

[0298] Please see Figure 13 , Figure 13 This is a schematic diagram of the structure of a model fusion device 1300 provided in an embodiment of this application. The model fusion device 1300 may include modules, units or means corresponding to the methods / operations / steps / actions executed by the terminal-side device or network-side device in the above method embodiments. The modules, units or means may be hardware circuits, software, or a combination of hardware circuits and software.

[0299] In one possible implementation, the model fusion device 1300 may include a processing unit 1301 and a transceiver unit 1302, the specific details of which are as follows:

[0300] The processing unit 1301 is used for data processing. The transceiver unit 1302 can implement corresponding communication functions. The transceiver unit 1302 can also be called a communication interface or a communication module.

[0301] Optionally, the model fusion apparatus 1300 may further include a storage unit, which can be used to store instructions and / or data. The processing unit 1301 can read the instructions and / or data in the storage module to enable the implementation of the aforementioned method embodiments.

[0302] Optionally, the transceiver unit 1302 may include a sending unit and a receiving unit. The sending unit is used to perform the sending operation in the above method embodiments. The receiving unit is used to perform the receiving operation in the above method embodiments.

[0303] It should be noted that the model fusion device 1300 may include a transmitting unit but not a receiving unit. Alternatively, the model fusion device 1300 may include a receiving unit but not a transmitting unit. Specifically, it depends on whether the above-described scheme executed by the model fusion device 1300 includes both transmitting and receiving actions.

[0304] Optionally, the model fusion device 1300 is used to perform the above. Figure 4 and Figure 10 The actions performed by the terminal device in the illustrated embodiment are shown above. For details, please refer to the above. Figure 4 and Figure 10 The relevant descriptions in the illustrated embodiments will not be elaborated here. For example, the model fusion apparatus 1300 is used to execute the following scheme: a transceiver unit 1302 is used to receive first indication information, which indicates the information of the model to be fused, and the model to be fused corresponds to at least one different parameter range of a first channel feature; a processing unit 1301 is used to perform model fusion based on the first indication information to determine the fused model.

[0305] It should be noted that the implementation and beneficial effects of each module can be found by referring to [the relevant documentation / reference]. Figure 4 and Figure 10 The corresponding description of the method embodiments shown.

[0306] Optionally, the model fusion device 1300 is used to perform the above. Figure 4 and Figure 10 The actions performed by the network device in the illustrated embodiment are shown above. For details, please refer to the above. Figure 4 and Figure 10 The relevant descriptions in the illustrated embodiments will not be elaborated here. For example, the model fusion apparatus 1300 is used to execute the following scheme: a processing unit 1301 is used to determine a second channel feature; the processing unit 1301 is also used to determine the information of the model to be fused based on the second channel feature and the information of the model supported by the terminal side, wherein the information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model; a transceiver unit 1302 is used to send first indication information, the first indication information being used to indicate the information of the model to be fused, and the different parameter ranges of the at least one first channel feature corresponding to the model to be fused.

[0307] It should be noted that the implementation and beneficial effects of each module can be found by referring to [the relevant documentation / reference]. Figure 4 and Figure 10 The corresponding description of the method embodiments shown.

[0308] Optionally, the model fusion device 1300 is used to perform the above. Figure 7 and Figure 11 The actions performed by the terminal device in the illustrated embodiment are shown above. For details, please refer to the above. Figure 7 and Figure 11 The relevant descriptions in the illustrated embodiments are not elaborated here. For example, the model fusion device 1300 is used to execute the following scheme: a transceiver unit 1302 is used to receive first information, the first information being used to indicate at least one second channel feature, different parameter ranges corresponding to the at least one second channel feature, and weights corresponding to the at least one second channel feature; a processing unit 1301 is used to determine the model to be fused based on the at least one second channel feature, the different parameter ranges corresponding to the at least one second channel feature, and information of the models supported by the terminal side, wherein the information of each model in the information of the models supported by the terminal side includes: at least one first channel feature corresponding to each model, or the parameter range of each first channel feature in the at least one first channel feature corresponding to each model; the processing unit 1301 is used to perform model fusion based on the weights corresponding to the at least one second channel feature and the model to be fused to determine the fused model.

[0309] It should be noted that the implementation and beneficial effects of each module can be found by referring to [the relevant documentation / reference]. Figure 7 and Figure 11 The corresponding description of the method embodiments shown.

[0310] Optionally, the model fusion device 1300 is used to perform the above. Figure 7 and Figure 11 The actions performed by the network device in the illustrated embodiment are shown above. For details, please refer to the above. Figure 7 and Figure 11 The relevant descriptions in the illustrated embodiments will not be elaborated here. For example, the model fusion device 1300 is used to execute the following scheme: a processing unit 1301 is used to determine first information, the first information being used to indicate at least one second channel feature, different parameter ranges corresponding to the at least one second channel feature, and weights corresponding to the at least one second channel feature; a transceiver unit 1302 is used to transmit the first information.

[0311] It should be noted that the implementation and beneficial effects of each module can be found by referring to [the relevant documentation / reference]. Figure 4 and Figure 10 The corresponding description of the method embodiments shown.

[0312] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, there may be other division methods.

[0313] The processing unit 1301 in the above embodiments can be implemented by at least one processor or processor-related circuitry. The transceiver unit 1302 can be implemented by a transceiver or transceiver-related circuitry. The transceiver unit 1302 can also be referred to as a communication module or communication interface. The storage module can be implemented by at least one memory.

[0314] Please see Figure 14 , Figure 14 This is a schematic diagram of another model fusion device 1400 provided in the embodiments of this application. The model fusion device 1400 may include modules, units or means corresponding to the methods / operations / steps / actions executed by the terminal-side device or network-side device in the above method embodiments. The modules, units or means may be hardware circuits, software, or a combination of hardware circuits and software.

[0315] The model fusion device 1400 includes at least one processor 1401. Optionally, it also includes a communication interface 1403 and a memory 1402. The processor 1401, memory 1402, and communication interface 1403 are interconnected via a bus 1404. Optionally, the processor 1401 and memory 1402 can be integrated together.

[0316] The memory 1402 includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or compact disc read-only memory (CD-ROM), and is used for related computer programs and data. The communication interface 1403 is used for receiving and sending data.

[0317] Processor 1401 can be one or more central processing units (CPUs). When processor 1401 is a CPU, the CPU can be a single-core CPU or a multi-core CPU.

[0318] The processor 1401 in the model fusion device 1400 is used to read computer programs or instructions stored in the memory 1402 to implement the functions of the above-mentioned processing unit, and the communication interface 1403 in the model fusion device 1400 is used to implement the functions of the above-mentioned transceiver unit.

[0319] Please see Figure 15 , Figure 15 This is a schematic diagram of a chip system architecture provided in an embodiment of this application. This chip system architecture can be used in network-side devices or devices or components applied in network devices, and can also be applied to terminal devices or devices or components applied in terminal devices. Input / output control is used to manage the input and output signals of the device; for example, input / output control can be represented as a modem, keyboard, mouse, touchscreen, etc. Input / output control may also be part of the processor. A receiver / transmitter is used to communicate with other devices. The receiver / transmitter may include a modem for modulating information or demodulating modulated information. An antenna is used to transmit or receive signals. Storage may include random access memory (RAM) or read-only memory (ROM). The storage may be used to store code that can be executed by the processor to implement corresponding functions. Processors may include intelligent hardware devices such as general-purpose processors, digital signal processors (DSPs), central processing units (CPUs), field-programmable gate arrays (FPGAs), graphics processing units (GPUs), and neural processing units (NNs).

[0320] This application also provides a computer-readable storage medium storing a computer program or instructions that, when executed on a processor, implement the method performed by a terminal device or network device in the above method embodiments.

[0321] This application also provides a computer program product, which includes a computer program or instructions that, when run on a processor, implement the method executed by a terminal device or network device in the above method embodiments.

[0322] This application also provides a communication system, which includes the terminal device and the network device described in the above embodiments. The terminal device is used to perform some or all of the operations performed by the terminal device in the above method embodiments, and the network device is used to perform some or all of the operations performed by the network device in the above method embodiments.

[0323] It is understood that the processor in the embodiments of this application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor may be a microprocessor or any conventional processor.

[0324] The method steps in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Alternatively, the ASIC can reside in a base station or terminal. Of course, the processor and storage medium can also exist as discrete components in the base station or terminal.

[0325] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are performed entirely or partially. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, or other programmable device. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video optical disc; or it can be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both types of storage media.

[0326] In the various embodiments of this application, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0327] In the description of this application, terms such as “first,” “second,” “S401,” or “S402” are used only for the purpose of distinguishing descriptions and for the convenience of context. The different sequence numbers themselves do not have specific technical meanings and should not be construed as indicating or implying relative importance, nor should they be construed as indicating or implying the order of execution of operations. The order of execution of each process should be determined by its function and internal logic.

Claims

1. A model fusion method, characterized in that, include: Receive first indication information, the first indication information being used to indicate information about the model to be fused, the model to be fused corresponding to at least one different parameter range of a first channel feature; Based on the first indication information, model fusion is performed to determine the fused model.

2. The method according to claim 1, characterized in that, The information of the model to be fused is related to the second channel feature and the information of the model supported by the terminal side. The information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model.

3. The method according to claim 2, characterized in that, The information of the model to be fused is also related to at least one of the following, which includes: Information on the terminal-side model fusion capability, or the upper limit of the number of models that need to be fused during transmission.

4. The method according to claim 2 or 3, characterized in that, At least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

5. The method according to any one of claims 1-4, characterized in that, The first indication information includes the identification information of the models that need to be fused.

6. A model fusion method, characterized in that, include: Determine the characteristics of the second channel; Based on the information of the second channel feature and the model supported by the terminal side, the information of the model that needs to be fused is determined. The information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model. Send a first indication message, which is used to indicate the information of the model that needs to be fused, and the different parameter ranges of at least one first channel feature corresponding to the model that needs to be fused.

7. The method according to claim 6, characterized in that, The information for determining the model to be fused based on the second channel features and the information of the model supported by the terminal side includes: Based on the second channel characteristics and information on the models supported by the terminal side, and at least one of the following, information on the models that need to be fused is determined, wherein at least one of the following includes: terminal side model fusion capability information, or upper limit information on the number of transmitted models that need to be fused.

8. The method according to claim 6 or 7, characterized in that, At least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

9. The method according to any one of claims 6-8, characterized in that, The second channel feature includes: a first feature, a first parameter range corresponding to the first feature, a second feature, and a second parameter range corresponding to the second feature. The information of the models supported by the terminal side includes information of the first model. The information of the first model includes: the first feature, the range of the third parameter corresponding to the first feature, the second feature, and the range of the fourth parameter corresponding to the second feature. Wherein, the first feature and the second feature are at least one first channel feature corresponding to the first model, and the third parameter range and the fourth parameter range are the parameter ranges of each channel feature in at least one first channel feature corresponding to the first model; The information for determining the model to be fused based on the second channel features and the information of the model supported by the terminal side includes: The distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or The distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than the third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than the fourth threshold. The information of the models that need to be fused includes the identification information of the first model.

10. The method according to any one of claims 6-8, characterized in that, The method further includes: Receive information about the models supported by the terminal side.

11. A model fusion device, characterized in that, The device includes a transceiver unit and a processing unit. The transceiver unit is used to receive first indication information, which indicates information about the model to be fused, and the model to be fused corresponds to at least one different parameter range of a first channel feature. The processing unit is used to perform model fusion based on the first indication information to determine the fused model.

12. The apparatus according to claim 11, characterized in that, The information of the model to be fused is related to the second channel feature and the information of the model supported by the terminal side. The information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model.

13. The apparatus according to claim 12, characterized in that, The information of the model to be fused is also related to at least one of the following, which includes: Information on the terminal-side model fusion capability, or the upper limit of the number of models that need to be fused during transmission.

14. The apparatus according to claim 12 or 13, characterized in that, At least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

15. The apparatus according to any one of claims 11-14, characterized in that, The first indication information includes the identification information of the models that need to be fused.

16. A model fusion device, characterized in that, The device includes a transceiver unit and a processing unit. The processing unit is used to determine the second channel characteristics; The processing unit is used to determine the information of the model to be fused based on the second channel feature and the information of the model supported by the terminal side. The information of each model in the information of the model supported by the terminal side includes: at least one first channel feature corresponding to each model, and the parameter range of each first channel feature in the at least one first channel feature corresponding to each model. The transceiver unit is used to send first indication information, which indicates information about the model to be fused, and the model to be fused corresponds to at least one different parameter range of a first channel feature.

17. The apparatus according to claim 16, characterized in that, The processing unit is configured to determine information about the models that need to be fused based on the second channel characteristics, information about the models supported by the terminal side, and at least one of the following, wherein at least one of the following includes: terminal side model fusion capability information, or upper limit information on the number of models that need to be fused transmitted.

18. The apparatus according to claim 16 or 17, characterized in that, At least one first channel feature corresponding to each model includes at least one of the following: frequency selection, aging, interference intensity, angle spread, or time delay spread.

19. The apparatus according to any one of claims 16-18, characterized in that, The second channel feature includes: a first feature, a first parameter range corresponding to the first feature, a second feature, and a second parameter range corresponding to the second feature. The information of the models supported by the terminal side includes information of the first model. The information of the first model includes: the first feature, the range of the third parameter corresponding to the first feature, the second feature, and the range of the fourth parameter corresponding to the second feature. Wherein, the first feature and the second feature are at least one first channel feature corresponding to the first model, and the third parameter range and the fourth parameter range are the parameter ranges of each channel feature in at least one first channel feature corresponding to the first model; The processing unit is configured to: the distance between the lower limit of the third parameter range and the lower limit of the first parameter range is less than a first threshold, and the distance between the upper limit of the third parameter range and the upper limit of the first parameter range is less than a second threshold, and / or If the distance between the lower limit of the fourth parameter range and the lower limit of the second parameter range is less than the third threshold, and the distance between the upper limit of the fourth parameter range and the upper limit of the second parameter range is less than the fourth threshold, then the information of the model to be fused includes the identification information of the first model.

20. The apparatus according to any one of claims 6-8, characterized in that, The transceiver unit is also used to receive information about the models supported by the terminal side.

21. A model fusion device, characterized in that, The apparatus includes at least one processor, which is configured to invoke a computer program or instructions to perform the method as described in claims 1-5.

22. A model fusion device, characterized in that, The apparatus includes at least one processor, which is configured to invoke a computer program or instructions to perform the method as described in claims 6-10.

23. A communication system, characterized in that, The communication system includes: the apparatus as described in claim 21 and the apparatus as described in claim 22.

24. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions that, when executed on a processor, implement the method as described in any one of claims 1-10.

25. A computer program product, characterized in that, The computer program product includes a computer program or instructions that, when run on a processor, implement the method as described in any one of claims 1-10.

Images