Communication method and apparatus

Abstract

A communication method and apparatus, which can be applied to the field of communications. In the method, a network device can send, to a terminal, information of an LLM simulator and information used for indicating the structure of an adaptation layer. The terminal performs training on the basis of the information of the LLM simulator, the information used for indicating the structure of the adaptation layer, and training data, so as to obtain the adaptation layer, and indicates the adaptation layer (for example, indicating parameters of the adaptation layer) to the network device. Upon receiving the information that indicates the adaptation layer, the network device performs data inference on the basis of the adaptation layer and an LLM. The adaptation layer comprises an input adaptation layer, and an output of the input adaptation layer can be used as an input of the LLM (or the LLM simulator). A network side does not directly send the LLM to the terminal, but sends the information of the LLM simulator, and in general, the LLM simulator can be obtained by compressing the LLM. Therefore, privacy of an LLM model would not be leaked. In addition, the terminal does not need to continuously report the training data, so that the air interface overhead can be reduced, and data privacy leakage in the terminal can be avoided.

Description

Communication method and apparatus

[0001] The present application claims priority to the Chinese Patent Application No. 202411244750.3, filed on September 5, 2024, and entitled “Communication method and apparatus”, the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] Embodiments of the present application relate to the field of communication, in particular to a communication method and apparatus. BACKGROUND

[0003] Transfer learning is a kind of machine learning (ML), which can be understood as fine-tuning a pre-trained model for a new related task. Through transfer learning, an existing model can be adjusted for a new related task using new data, thereby reducing the training overhead.

[0004] In a mobile communication system, common applications of transfer learning include: using wireless corpus to perform transfer learning on a large language model (LLM) trained by a base station for a certain communication task, such as beam management, to obtain a new model for the communication task.

[0005] However, wireless corpus usually includes data on the terminal side. If transfer learning is performed on the terminal side, the LLM model needs to be downloaded to the terminal by the base station, which will destroy the privacy of the LLM model. SUMMARY

[0006] The present application provides a communication method and apparatus, which can protect model privacy during transfer learning.

[0007] In a first aspect, a communication method is provided. The method can be performed by a terminal, or by a component of the terminal, such as a processor, a chip, or a chip system of the terminal, or by a logic module or software that can realize all or part of the functions of the terminal. Hereinafter, the terminal is taken as an example for description. The method includes: receiving first information from a network device, the first information including information of a large language model (LLM) simulator; receiving second information from the network device, the second information being used to indicate a structure of an adaptation layer, the adaptation layer including an input adaptation layer, an output of the input adaptation layer being an input of the LLM simulator; performing training according to the first information, the second information, and training data to obtain the adaptation layer; and sending third information to the network device, the third information being used to indicate the adaptation layer. Exemplarily, the third information used to indicate the adaptation layer can also be understood as: the third information is used to feedback the adaptation layer. For example, the third information can include parameters of the adaptation layer.

[0008] Based on the scheme, the network device can send the information of the LLM simulator and the information for indicating the adaptation layer structure to the terminal. The terminal trains the adaptation layer (such as the parameters of the adaptation layer) according to the information of the LLM simulator, the information for indicating the adaptation layer structure, and the training data, and indicates the adaptation layer to the network device, such as indicating the parameters of the adaptation layer to the network device. In the scheme, since the network side does not directly send the LLM to the terminal, but sends the information of the LLM simulator, the LLM simulator can usually be obtained after the LLM is compressed, so the privacy of the LLM model is not disclosed. In addition, the terminal does not need to continuously report the training data, which can reduce the air interface overhead and avoid disclosure of the data privacy of the terminal. Furthermore, the terminal only needs to train the adaptation layer, which can reduce the computational complexity and the requirement for the capability of the terminal compared with retraining the complete model for the communication task.

[0009] In a possible design, the method further includes: sending capability information to the network device, the capability information being used to determine a compression ratio, the compression ratio being a compression ratio of the LLM simulator and the LLM. The capability information includes at least one of the following: memory, computing capability, or power. For example, the power can be the remaining power, the available power, or the current power of the terminal.

[0010] Based on the possible design, the terminal reports the capability information to the network device, so that the network device can reasonably set the compression ratio based on the capability of the terminal, thereby providing a suitable LLM simulator, avoiding problems in training the adaptation layer due to unreasonable setting of the compression ratio, and improving the reliability of the adaptation layer training.

[0011] In a possible design, the method further includes: sending fourth information to the network device, the fourth information being used to indicate a dimension of an input parameter of the expected input adaptation layer, and / or a structure of the expected input adaptation layer.

[0012] Based on the possible design, the terminal reports the expected adaptation layer structure and / or parameter to the network device, so that the network device can set a reasonable adaptation layer structure and / or dimension of the input / output parameter based on the expected structure and / or dimension of the input / output parameter of the terminal, thereby providing a suitable adaptation layer structure, avoiding problems in training the adaptation layer due to unreasonable setting of the adaptation layer structure, and improving the efficiency of the adaptation layer training.

[0013] In a possible design, the method further includes: receiving fifth information from the network device, the fifth information being used to indicate to update the adaptation layer.

[0014] Based on the possible design, the network device can timely instruct the terminal to update the adaptation layer, so as to optimize the adaptation layer, thereby improving the performance of the communication task based on the joint model and improving the communication quality.

[0015] In a second aspect, a communication method is provided. The method can be performed by a RAN node, or a component of the RAN node, e.g., a processor, a chip, or a chip system, etc., of the RAN node, or a logic module or software that can implement all or part of the functions of the RAN node. The method is described below with reference to the RAN node, and includes: sending, to a terminal, first information, the first information comprising information of a large language model (LLM) simulator; sending, to the terminal, second information, the second information being used to indicate a structure of an adaptation layer, the adaptation layer comprising an input adaptation layer; the first information and the second information being used to train the adaptation layer; receiving, from the terminal, third information, the third information being used to indicate the adaptation layer; and performing data inference according to the adaptation layer and the LLM. The technical effects brought by the second aspect can refer to the technical effects brought by the first aspect, which will not be described herein again.

[0016] In a possible design, the performing data inference according to the adaptation layer and the LLM comprises: inputting data obtained according to a measurement quantity of a first reference signal into a joint model to obtain a first wireless channel parameter and / or a first wireless transmission parameter. The joint model is composed of the adaptation layer and the LLM, an input of the input adaptation layer is an input of the joint model, and an output of the input adaptation layer is an input of the LLM.

[0017] In a possible design, the method further comprises: receiving, from the terminal, capability information, the capability information being used to determine a compression ratio, the compression ratio being a compression ratio of the LLM simulator and the LLM. The capability information comprises at least one of the following: memory, computing capability, or power.

[0018] In a possible design, the method further comprises: receiving, from the terminal, fourth information, the fourth information being used to indicate a dimension of an expected input parameter of the input adaptation layer, and / or an expected structure of the input adaptation layer.

[0019] In a possible design, the method further comprises: sending, to the terminal, fifth information, the fifth information being used to indicate an update of the adaptation layer.

[0020] The technical effects brought by any possible design of the second aspect can refer to the technical effects brought by the corresponding design of the first aspect, which will not be described herein again.

[0021] In a possible design in combination with the first aspect or the second aspect, the training data comprises a measurement quantity of a reference signal and a sample label obtained based on the measurement quantity, and the input of the input adaptation layer is determined by the measurement quantity of the reference signal. Exemplarily, the reference signal comprises a synchronization signal and a physical broadcast channel block (SSB), and / or a channel state information reference signal (CSI-RS).

[0022] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the adaptation layer further includes an output adaptation layer, and an output of the LLM simulator is an input of the output adaptation layer.

[0023] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the information of the LLM simulator includes an application program interface (API) file of the LLM simulator, or includes an open neural network exchange (ONNX) file of the LLM simulator.

[0024] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the second information includes structure information and / or initial parameters of the input adaptation layer.

[0025] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the second information further includes structure information and / or initial parameters of the output adaptation layer.

[0026] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the LLM simulator is obtained by compressing the LLM.

[0027] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the fourth information is further used to indicate a dimension of an output parameter of the expected output adaptation layer, and / or a structure of the expected output adaptation layer.

[0028] With reference to the first aspect or the second aspect, in a possible design of the first aspect or the second aspect, the fifth information includes at least one of the following: structure information and / or initial parameters of the updated input adaptation layer, structure information and / or initial parameters of the updated output adaptation layer, and information of the updated LLM simulator. Alternatively, the fifth information includes an update command, and the update command is used to indicate updating of the adaptation layer.

[0029] In a third aspect, a communication apparatus is provided, configured to implement various methods. The communication apparatus includes modules, units, or means corresponding to the modules, units, or means for implementing the methods, and the modules, units, or means can be implemented by hardware, software, or by a combination of hardware and software. The hardware or software includes one or more modules or units corresponding to the functions.

[0030] In some possible designs, the communication apparatus can include a processing module and a transceiver module. The processing module can be configured to implement the processing functions in any of the aspects and any possible implementation manners thereof. The transceiver module can include a receiving module and a sending module, and be configured to implement the receiving functions and the sending functions in any of the aspects and any possible implementation manners thereof.

[0031] In some possible designs, the transceiver module can be composed of a transceiver circuit, a transceiver, a transceiver, or a communication interface.

[0032] In a fourth aspect, a communication apparatus is provided, which comprises: a processor and a memory; the memory is configured to store computer instructions, and when the processor executes the instructions, the communication apparatus performs the method described in any of the above aspects and any possible design thereof.

[0033] In a fifth aspect, a communication apparatus is provided, which comprises: a processor and a communication interface; the communication interface is configured to communicate with modules outside the communication apparatus; and the processor is configured to execute computer programs or instructions, so that the communication apparatus performs the method described in any of the above aspects and any possible design thereof.

[0034] In a sixth aspect, a communication apparatus is provided, which comprises: at least one processor; the processor is configured to execute computer programs or instructions stored in a memory, so that the communication apparatus performs the method described in any of the above aspects and any possible design thereof. The memory can be coupled with the processor, or can be independent of the processor.

[0035] In a seventh aspect, a communication apparatus (for example, the communication apparatus can be a chip or a chip system) is provided, which comprises a processor configured to implement the functions involved in any of the above aspects and any possible design thereof.

[0036] In some possible designs, the communication apparatus comprises a memory configured to store necessary program instructions and data.

[0037] In some possible designs, when the apparatus is a chip system, the apparatus can be composed of a chip, or can comprise a chip and other discrete devices.

[0038] The communication apparatus described in the third aspect to the seventh aspect can be the terminal in the first aspect, or an apparatus (such as a chip or a chip system) included in the terminal; or the communication apparatus can be the RAN node in the second aspect, or an apparatus (such as a chip or a chip system) included in the RAN node.

[0039] In an eighth aspect, a communication apparatus is provided, which can be a terminal, or a module or unit (for example, a chip, or a chip system, or a circuit) corresponding to the terminal in executing the method / operation / step / action described in the first aspect, or a module or unit capable of being used in matching with the terminal; or the communication apparatus can be a RAN node, or a module or unit (for example, a chip, or a chip system, or a circuit) corresponding to the RAN node in executing the method / operation / step / action described in the second aspect, or a module or unit capable of being used in matching with the RAN node.

[0040] It can be understood that, when the communication apparatus provided in any one of the third aspect to the eighth aspect is a chip, the sending action / function of the communication apparatus can be understood as outputting information, and the receiving action / function of the communication apparatus can be understood as inputting information.

[0041] In a ninth aspect, a computer-readable storage medium is provided, which stores a computer program or instructions, when the computer program or instructions are executed on a communication apparatus, the communication apparatus is enabled to perform the method in any one of the aspects above and any possible design thereof.

[0042] In a tenth aspect, a computer program product is provided, which contains instructions, when the computer program product is executed on a communication apparatus, the communication apparatus is enabled to perform the method in any one of the aspects above and any possible design thereof.

[0043] In an eleventh aspect, a communication system is provided, which includes a terminal and a RAN node. The terminal is configured to implement the method in the first aspect above and any possible design thereof, and the RAN node is configured to implement the method in the second aspect above and any possible design thereof.

[0044] The technical effects brought by any one of the designs in the third aspect to the eleventh aspect can be referred to the technical effects brought by different designs in the first aspect or the second aspect, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0045] FIG. 1 is a flow diagram of a beam management provided in the present application;

[0046] FIG. 2 is a schematic diagram of a wide beam and a narrow beam provided in the present application;

[0047] FIG. 3 is a schematic diagram of a structure of a neuron provided in the present application;

[0048] FIG. 4 is a schematic diagram of a structure of a DNN provided in the present application;

[0049] FIG. 5 is a schematic diagram of transfer learning provided in the present application;

[0050] FIG. 6 is another schematic diagram of transfer learning provided in the present application;

[0051] FIG. 7 is a schematic diagram of a structure of a communication system provided in the present application;

[0052] FIG. 8 is a schematic diagram of a structure of another communication system provided in the present application;

[0053] FIG. 9 is a schematic diagram of a structure of yet another communication system provided in the present application;

[0054] FIG. 10 is a schematic diagram of a structure of an O-RAN system provided in the present application;

[0055] FIG. 11 is a flow diagram of a communication method provided by the present application;

[0056] FIG. 12 is a structural diagram of a joint model provided by the present application;

[0057] FIG. 13 is a flow diagram of an application of the joint model provided by the present application;

[0058] FIG. 14 is a flow diagram of another communication method provided by the present application;

[0059] FIGS. 15-17 are structural diagrams of communication apparatuses provided by the present application. DETAILED DESCRIPTION

[0060] In the description of the present application, unless otherwise specified, “ / ” represents an “or” relationship between the objects associated in front and back, for example, A / B can represent A or B; “and / or” in the present application is only a description of the association between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent: A alone, A and B exist at the same time, and B alone, where A and B can be singular or plural.

[0061] In the description of the present application, unless otherwise specified, “multiple” means two or more than two. “At least one of the following” or the like means any combination of the items, including any combination of single item or multiple items. For example, at least one of a, b, or c can represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0062] In addition, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the same items or similar items with basically the same function and role are distinguished by using “first”, “second”, etc. The skilled in the art can understand that “first”, “second”, etc. do not limit the quantity and execution order, and “first”, “second”, etc. also do not necessarily mean different.

[0063] In the embodiments of the present application, the words “exemplary” or “for example” are used to mean serving as an example, instance, or illustration. Any embodiment or design presented as “exemplary” or “for example” in the embodiments of the present application should not be interpreted as being more preferred or advantageous than other embodiments or design solutions. Rather, the use of “exemplary” or “for example” is intended to present relevant concepts in a specific manner, facilitating understanding.

[0064] It can be understood that the "embodiments" mentioned throughout the specification mean that the specific features, structures or characteristics related to the embodiments are included in at least one embodiment of the present application. Therefore, the various embodiments throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It can be understood that in various embodiments of the present application, the size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0065] It can be understood that in the present application, "…", "if" and "when" all refer to the corresponding processing under certain objective conditions, not the time limit, and also do not require judgment action when implementing, nor does it mean that there are other limitations.

[0066] It can be understood that some optional features in the embodiments of the present application can be implemented independently in some scenarios without relying on other features, such as the scheme currently based on, to solve the corresponding technical problems and achieve the corresponding effects. In some scenarios, it can also be combined with other features according to demand. Correspondingly, the devices given in the embodiments of the present application can also realize these features or functions, which will not be described here.

[0067] In the present application, the same or similar parts of each embodiment can be mutually referred to, unless otherwise specified. In various embodiments of the present application, if there is no special description and no logical conflict, the terms and / or descriptions of different embodiments are consistent and can be mutually referred to. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship. The implementation modes of the present application described below do not constitute a limitation on the protection scope of the present application.

[0068] In order to facilitate the understanding of the technical solutions of the embodiments of the present application, first, a brief introduction of the related technologies of the present application is given as follows.

[0069] 1. Beam management:

[0070] Beam management means establishing and maintaining a set of suitable beam pairs between the network and the terminal. For downlink transmission, the network side needs to select a suitable transmit beam, and the terminal side needs to select a suitable receive beam. The transmit beam and the receive beam jointly form a set of beam pairs to maintain good wireless connection. Usually, as shown in FIG. 1, the beam management process can include beam training, beam maintenance, beam failure recovery, etc.

[0071] Exemplarily, the beam training can also be referred to as beam selection, which is mainly completed through beam sweeping and corresponding beam measurement. The beam sweeping is implemented by sending a reference signal. The reference signal mainly includes a synchronization signal block (SSB) and a channel state information reference signal (CSI-RS). The SSB is configured in a cell granularity, and a beam (also referred to as an SSB beam) used for sending the SSB can be understood as a wide beam. The CSI-RS is a user-level reference signal, and a beam (also referred to as a CSI-RS beam) used for sending the CSI-RS can be understood as a narrow beam. As shown in FIG. 2, a schematic diagram of a wide beam and a narrow beam is given.

[0072] Irrespective of the wide beam or the narrow beam, the network and the terminal need to traverse the candidate beams, and then compare the measurement results of all the candidate beams, such as the reference signal received power (RSRP) of the beam, to determine the respective optimal sending / receiving beam.

[0073] However, due to the base station hardware limitation and the fact that both the SSB beam and the CSI-RS beam are analog beams (full-band weighting beams), the base station can only send one beam at the same time, so that in the beam selection process, different beams need to be sent through multiple time points to cover the service range, resulting in a large beam sweeping overhead.

[0074] In order to reduce the beam sweeping overhead, hierarchical scanning can be performed, that is, wide beams are scanned first, and then a small part of narrow beams are scanned under the wide beams, so as to achieve the goal of reducing the overhead.

[0075] 2. Artificial intelligence (AI) and machine learning (ML):

[0076] Machine learning can be understood as an important technical approach to realize artificial intelligence. Machine learning can be generally divided into supervised learning, unsupervised learning, reinforcement learning, etc.

[0077] Supervised learning learns the mapping relationship between sample values and sample labels according to the collected sample values and sample labels, and uses a machine learning model to express the learned mapping relationship. The process of training the machine learning model is the process of learning the mapping relationship. In signal detection, the noisy received signal is the sample value, and the true constellation point corresponding to the signal is the sample label. Machine learning is expected to learn the mapping relationship between sample values and sample labels through training, that is, to learn a signal detector. During training, the model parameters are optimized by calculating the error between the predicted value of the model and the true label. Once the mapping relationship is learned, the learned mapping relationship can be used to predict the label of each new sample. The learned mapping relationship of supervised learning can include linear mapping and nonlinear mapping. According to the type of sample label, the learned task can be divided into classification task and regression task.

[0078] Unsupervised learning uses algorithms to discover the internal patterns of samples according to the collected sample values. In unsupervised learning, a class of algorithms uses the sample itself as a supervision signal, that is, the model learns the mapping relationship from the sample to the sample, which is called self-supervised learning. During training, the model parameters are optimized by calculating the error between the predicted value of the model and the sample itself. Unsupervised learning can be used for signal compression and decompression recovery applications. Common algorithms include autoencoders and generative adversarial networks.

[0079] Reinforcement learning is different from supervised learning and is a class of algorithms that learn strategies to solve problems by interacting with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have clear "correct" action label data. The algorithm needs to interact with the environment to obtain the reward signal of the environment feedback, and then adjust the decision action to obtain a larger reward signal value. In the following power control, the reinforcement learning model adjusts the downlink transmission power of each user according to the system total throughput rate feedback by the wireless network, and then expects to obtain a higher system throughput rate. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and the optimal decision action. However, because the "correct action" label 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.

[0080] 3、Deep neural network (DNN):

[0081] DNN is a specific implementation form of machine learning. According to the universal approximation theorem, a neural network can theoretically approximate any continuous function, thereby enabling the neural network to learn any mapping. A traditional communication system needs to rely on rich expert knowledge to design a communication module, while a deep learning communication system based on DNN can automatically discover the implicit pattern structure from a large amount of data set, establish the mapping relationship between data, and obtain better performance than the traditional modeling method.

[0082] The idea of DNN comes from the neuron structure of the brain tissue. Each neuron performs a weighted sum operation on its input value and uses a nonlinear function to operate on the weighted sum result to generate an output. For example, as shown in FIG. 3, the input of a neuron is x = [x0, …, xn], the weight corresponding to the input is d = [d0, …, dn], the nonlinear function is f, and the bias of the weighted sum is b. The form of the nonlinear function can be diversified, for example, it can be a maximum value function max{0, x}, and the output of the neuron can be represented as

[0083] The weighted value corresponding to each neuron and the nonlinear function can be understood as the parameter of the DNN network model. The model parameter is optimized through the training process, thereby enabling the DNN network to have the ability to extract data features and express mapping relationships. DNN generally uses a supervised learning or unsupervised learning strategy to optimize the model parameter.

[0084] DNN generally has a multi-layer structure, and each layer can contain multiple neurons. For example, as shown in FIG. 4, a DNN can include an input layer, a hidden layer, and an output layer. The input layer transmits the received values to the hidden layer after processing by the neurons. Similarly, the hidden layer transmits the calculation result to the output layer after processing, thereby generating the final output of the DNN. DNN generally has multiple hidden layers, and the hidden layer often directly affects the ability to extract information and fit functions. Increasing the number of hidden layers of DNN or expanding the width of each layer can improve the function fitting ability of DNN.

[0085] According to the construction mode of the network, DNN can be divided into a feedforward neural network (FNN), a convolutional neural network (CNN), and a recurrent neural network (RNN). FIG. 4 can be understood as a kind of FNN network, and its characteristic is that the neurons of adjacent layers are completely connected two by two, which makes FNN usually need a large amount of storage space, resulting in high computational complexity.

[0086] CNN can be understood as a DNN network for processing data with a similar grid structure. For example, time series data (time axis discrete sampling) and image data (two-dimensional discrete sampling) can be considered as similar grid structure data. CNN does not use all input information at once for operation, but uses a fixed size window to extract part of the information for convolution operation, which greatly reduces the calculation of model parameters. In addition, different convolution kernels can be used for each window according to the type of information extracted by the window (such as people and objects in the same picture as different types of information), which enables CNN to better extract the features of the input data.

[0087] RNN is a class of DNN network that uses feedback time series information. Its input includes the new input value at the current time and the output value of itself at the previous time. RNN is suitable for obtaining sequence features with temporal correlation, such as speech recognition, channel coding and decoding applications.

[0088] 4. Training and inference of AI / ML model:

[0089] In simple terms, model training can be understood as learning certain capabilities from a large number of known data samples, while inference is using the capabilities to quickly and efficiently perform the same / in similar inference process as the training process on data samples not seen in training to obtain the desired inference result.

[0090] Generally, the main difference between training and inference is that the input of the training process is known data, and the input of the inference process is newly collected data, or data not seen in training. The dimensions of the input and output are consistent in training and inference.

[0091] 5. Large language model (LLM):

[0092] LLM is a class of basic models that are trained on a super large amount of data to enable them to understand and generate natural language and other types of content to perform various tasks. These models are usually based on a transformer architecture, such as a generative pre-trained transformer, which is good at processing sequential data such as text input. LLM is composed of multiple layers of neural networks, and the parameters of each layer of neural network can be fine-tuned during the training process. The numerous neural network layers known as attention mechanisms further enhance the functionality of these neural networks, which can adjust specific parts of the data set.

[0093] During training, these models learn to predict the next word in a sentence given the context provided by the preceding words. The model does this by assigning a probability score to a repeating tokenized word (broken down into smaller sequences of characters). These tokens are then converted into embeddings, which are numerical representations of that context.

[0094] To ensure accuracy, this process involves training the LLM on a large corpus of text (billions of pages) that enables the LLM to learn grammar, semantics, and conceptual relationships through zero-shot and self-supervised learning. After being trained on this training data, the LLM can automatically predict the next word based on the input they receive and use the patterns and knowledge they have gained to generate text. The result is coherent and contextually relevant language that can be used for a wide range of natural language understanding (NLU) and content generation tasks.

[0095] 6. Transfer learning:

[0096] Transfer learning is a type of machine learning that can be understood as fine-tuning a pre-trained model for a new related task. Training a new machine learning model is a time-consuming and complex process that requires a large amount of data, high computational power, and multiple iterations before it can be put into production. With transfer learning, an existing model can be adjusted for a new related task using new data, reducing the training overhead.

[0097] In a mobile communication system, a common application of transfer learning includes using a wireless corpus to transfer learn a base station trained LLM for a certain communication task, such as beam management, to obtain a new model for the communication task. The wireless corpus usually includes terminal-side data, such as RSRP of all scanned beams measured by the terminal.

[0098] To implement transfer learning, as shown in FIG. 5, one possible implementation is to perform transfer learning on the model owner side, in which case the data owner needs to constantly upload training data to the LLM model owner. For example, the terminal constantly reports the RSRP of all scanned beams measured by the terminal to the base station, and the base station uses the RSRP reported by the terminal to perform transfer learning to transfer the LLM to a model for completing the communication task.

[0099] As shown in FIG. 6, another possible implementation is to perform transfer learning on the data owner side, in which case the model owner needs to pass the LLM to the data owner. For example, the base station downlink the LLM model to the terminal, and the terminal performs transfer learning on the LLM combined with the RSRP of all scanned beams measured by the terminal to transfer the LLM to a model for completing the communication task, and feeds back the parameters of the model to the base station. The parameters of the model can also be understood as the parameters of the updated LLM model.

[0100] However, in the case of transfer learning at the base station side, the terminal continuously uploading training data may bring serious air interface overhead, and may also involve the data privacy of the terminal. In the case of transfer learning at the terminal side, since the training capability of the LLM model is one of the core competences of the company, the base station issuing the LLM model to the terminal may damage the privacy of the model.

[0101] Based on this, the present application provides a communication method. The network device can send the information of the LLM simulator and the information indicating the adaptation layer structure to the terminal. The terminal trains according to the information of the LLM simulator, the information indicating the adaptation layer structure, and the training data, obtains the adaptation layer (such as the parameters of the adaptation layer), and indicates the adaptation layer (such as the parameters of the adaptation layer) to the network device. After receiving the information indicating the adaptation layer, the network device performs data inference according to the adaptation layer and the LLM. The adaptation layer includes an input adaptation layer, and the output of the input adaptation layer can be used as the input of the LLM (or the LLM simulator). In this scheme, since the network side does not directly send the LLM to the terminal, but sends the information of the LLM simulator, the LLM simulator is usually obtained by compressing the LLM, so the privacy of the LLM model will not be leaked. In addition, there is no need for the terminal to continuously report the training data, which can reduce the air interface overhead and avoid the leakage of the data privacy of the terminal. Furthermore, the terminal only needs to train the adaptation layer, which can reduce the computational complexity and the requirement for the capability of the terminal compared with retraining the complete model for the communication task.

[0102] The technical solutions of the embodiments of the present application can be applied to various communication systems, which can be a third generation partnership project (3GPP) communication system, for example, a long term evolution (LTE) system, a fourth generation (4G) system, a new radio (NR) system, a fifth generation (5G) system, a system of mixed networking of LTE and 5G, a perception system, a communication-perception integrated system, a non-terrestrial network (NTN), a satellite communication system, a device-to-device (D2D) communication system, a vehicle-to-everything (V2X) communication system, a machine-type communication (MTC) system, an internet of things (IoT) system, or other future communication systems, or a fusion system of multiple systems. The communication system can also be a non-3GPP communication system, such as a wireless local area network (WLAN) system, a wireless fidelity (WiFi) system, etc., without limitation.

[0103] It should be noted that the above-mentioned communication system to which the present application is applicable is only an example, and the communication system to which the present application is applicable is not limited thereto. The communication system provided by the present application does not cause any limitation on the solutions of the present application. Here, it is uniformly stated that the following will not be described in detail.

[0104] FIG. 7 is a schematic diagram of a communication system applicable to the communication method according to the embodiments of the present application. As shown in FIG. 7, the communication system 700 can include at least one network device, for example, the network device 710 shown in FIG. 7; the communication system 700 can also include at least one terminal, for example, the terminal 720 and the terminal 730 shown in FIG. 7. The network device 710 and the terminal (such as the terminal 720 and the terminal 730) can communicate through a wireless link. The communication devices in the communication system, for example, the network device 710 and the terminal 720, can communicate through multi-antenna technology.

[0105] FIG. 8 is a schematic diagram of another communication system applicable to the communication method according to the embodiments of the present application. As shown in FIG. 8, compared with the communication system 700 shown in FIG. 7, the communication system 800 shown in FIG. 8 further includes an AI network element 740. The AI network element 740 is configured to perform AI-related operations, for example, constructing a training data set or training an AI model, etc.

[0106] It should be understood that FIG. 8 only illustrates the example that the AI network element 740 is directly connected with the network device 710, and in other scenarios, the AI network element 740 can also be connected with a terminal. Alternatively, the AI network element 740 can be connected with both the network device 710 and the terminal. Alternatively, the AI network element 740 can also be connected with the network device 710 and / or the terminal through a third-party network element. The embodiments of the present application do not limit the connection relationship between the AI network element and other devices. In addition, the AI network element 740 can also be arranged as a module in the network device and / or the terminal, for example, arranged in the network device 710 or the terminal (such as the terminal 720 and the terminal 730) shown in FIG. 7; or the AI network element can also be deployed separately, for example, deployed in a location outside the network device and the terminal, such as a host of an over the top (OTT) system or a cloud server.

[0107] It should be noted that FIG. 7 and FIG. 8 are only simplified schematic diagrams for illustration and understanding, for example, the communication system can also include other devices, such as wireless relay devices and / or wireless backhaul devices, which are not shown in FIG. 7 and FIG. 8. In actual application, the communication system can include multiple network devices, and can also include multiple terminals. The number of network devices and terminals included in the communication system is not limited in the embodiments of the present application.

[0108] In a possible implementation, the terminal is a user-side device with wireless transceiving function, and the terminal can also be referred to as a user equipment (UE), an access terminal, a user unit, a user station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal device, a wireless communication device, a user agent or a user apparatus.

[0109] The terminal can be a fixed device, a mobile device, a handheld device (such as a mobile phone), a wearable device, a vehicle-mounted device, or a wireless device (such as a communication module, a modem, a chip or a chip system, etc.) built-in in the above devices. The terminal is used to connect people, things, machines, etc., and can be widely used in various scenarios, such as: cellular communication, D2D communication, V2X communication, MTC communication, IoT, virtual reality (VR), augmented reality (AR), industrial control, self driving, remote medical, smart grid, transportation safety, smart home, smart office, smart wear, smart transportation, smart city, unmanned aerial vehicle, robot, etc.

[0110] For example, the terminal can be a handheld terminal in cellular communication, a communication device in D2D, an Internet of Things device in MTC, a camera in smart transportation and smart city, or a communication device on a drone, etc. Alternatively, the terminal can be a mobile phone, a tablet computer, a computer with wireless transceiver function, a wearable device, a vehicle, a drone, a helicopter, an airplane, a ship, a robot, a mechanical arm, a smart home device, etc. Embodiments of the present application do not limit the device form of the terminal.

[0111] For example, but not limitation, in embodiments of the present application, the wearable device can also be referred to as a wearable smart device, which is a general term for devices that can be worn and are designed and developed by applying wearable technology to daily wear, such as glasses, gloves, watches, clothing, and shoes, etc. The wearable device is a portable device that can be directly worn on the body or integrated into the clothes or accessories of the user.

[0112] In embodiments of the present application, the device for implementing the function of the terminal can be a terminal or a device capable of supporting the terminal to implement the function, such as a chip or a chip system, which can be installed in the terminal or used in conjunction with the terminal. For example, the chip system can be composed of a chip or include a chip and other discrete devices. In embodiments of the present application, only the device for implementing the function of the terminal is taken as an example for illustration, and the scheme of the present application is not limited.

[0113] In a possible implementation, in embodiments of the present application, the network device can be understood as a network side device with wireless transceiver function, which helps the terminal to implement wireless access.

[0114] As a possible implementation, the network device can be an access network device, such as a base station, an evolved NodeB (eNodeB), an access point (AP), a transmission reception point (TRP), a transmitting point (TP), a next generation NodeB (gNB) in a 5G mobile communication system, a base station in a subsequent evolution of 3GPP, a base station in a future mobile communication system, a master station, a secondary station, a motor slide retainer (MSR) node, a home base station, a network controller, an access node in a WiFi system, a wireless relay node, a wireless backhaul node, etc. Optionally, the RAN node can also be a server, a wearable device, a vehicle or a vehicle-mounted device, etc. For example, the RAN node in V2X technology can be a road side unit (RSU).

[0115] For example, a base station can be a macro base station, a micro base station, or an indoor station, a relay node or a donor node, or a radio controller in a cloud radio access network (CRAN) scenario. A base station can be fixed, or mobile. For example, a helicopter or an unmanned aerial vehicle (UAV) can be configured to act as a mobile base station, and one or more cells can move according to the location of the mobile base station. In other examples, a helicopter or an unmanned aerial vehicle can be configured to act as a device that communicates with another base station.

[0116] As another possible implementation, a network device can include at least one of the following: a central unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP), a radio unit (RU), etc.

[0117] For example, a CU and a DU can be separately configured, or can also be included in the same network element, such as a baseband unit (BBU). An RU can be included in a radio frequency device or a radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH).

[0118] In a possible design, a processing unit in a BBU for implementing baseband functions is referred to as a base band high (BBH) unit, and a processing unit in an RRU / AAU / RRH for implementing baseband functions is referred to as a base band low (BBL) unit.

[0119] In different systems, the CU (or CU-CP and CU-UP), DU or RU can also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, the CU can also be referred to as the O-RAN central unit (O-CU), the DU can also be referred to as the O-RAN distributed unit (O-DU), the CU-CP can also be referred to as the O-RAN central unit control plane (O-CU-CP), the CU-UP can also be referred to as the O-RAN central unit user plane (O-CU-UP), and the RU can also be referred to as the O-RAN radio unit (O-RU). Any of the CU (or CU-CP, CU-UP), DU and RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0120] In a possible design, the devices such as the CU, the DU, and the RU communicate with each other through interfaces. For example, as shown in FIG. 9, the CUs communicate with each other through an Xn interface, the CUs and the DUs communicate with each other through an F1 interface, the DUs and the terminal communicate with each other through an air interface, and the CUs and the core network device communicate with each other through an NG interface. In addition, the core network device and the network device can be connected to an operation administration and maintenance (OAM) system.

[0121] Optionally, as shown in FIG. 9, one or more AI modules (only one is shown in FIG. 9 for clarity) can be arranged in the core network device, the CU, the DU, the terminal, or the OAM. The AI module is used to implement the corresponding AI function, and the AI modules deployed in different devices can be the same or different. One AI module can have one or more models. One model can infer an output, which includes one parameter or multiple parameters. The learning process, the training process, or the inference process of different models can be deployed in different devices, or can be deployed in the same device.

[0122] In addition, the DU and the RU can communicate through a fronthaul interface (not shown in FIG. 9). The DU and the RU can support one or more types of fronthaul interface, different fronthaul interfaces corresponding to DUs and RUs with different functions, respectively. For example, if the fronthaul interface between the DU and the RU is a common public radio interface (CPRI), the DU is configured to implement one or more of the baseband functions, and the RU is configured to implement one or more of the radio frequency functions. If the fronthaul interface between the DU and the RU is another interface, it will move part of the baseband functions of the downlink and / or uplink from the DU to the RU for implementation.

[0123] In a possible design, the interface between the DU and the RU can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the split manner between the DU and the RU is different, corresponding to different categories (Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, and F. For the function description of the DU and the RU corresponding to various types of eCPRI, reference can be made to the eCPRI protocol, which is not described herein.

[0124] In an embodiment of this application, the apparatus for implementing the function of the network device can be a network device; or can be an apparatus capable of supporting the network device to implement the function, for example, a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. The apparatus can be installed in the network device or used in matching with the network device. In this embodiment of this application, only the apparatus for implementing the function of the network device is taken as an example for description, and the scheme of this embodiment of this application is not limited in this way.

[0125] In a possible implementation, as shown in FIG. 10, the communication system applicable to this embodiment of this application can further include a RAN intelligent controller (RIC). The RIC can further include a non-real time RIC (Non-RT RIC or NRT RIC) and a near-real time RIC (Near-RT RIC or nRT RIC).

[0126] The non-real-time RIC is used to implement non-real-time intelligent management of the RAN, and processes non-real-time information, such as data that is not sensitive to latency, for example, data with a latency of seconds. The non-real-time RIC can also implement AI / ML including model training and model updating, and guide applications / functions in the near-real-time RIC based on a policy. The near-real-time RIC is used to implement near-real-time intelligent management of the RAN, and processes near-real-time information, such as data that is relatively sensitive to latency, for example, data with a latency of tens of milliseconds. The near-real-time RIC can also implement near-real-time control and optimization of modules and resources of the O-RAN through data collection and related operations on an E2 interface. The E2 interface can be understood as an open interface between two nodes (or endpoints).

[0127] In a possible design, the near-real-time RIC can be used to perform model training and inference. For example, the near-real-time RIC is used to train an AI model, and inference is performed using the AI model. The near-real-time RIC can obtain network-side and / or terminal-side information from a RAN node (for example, a CU, a CU-CP, a CU-UP, a DU, and / or a RU) and / or a terminal. The information can be used as training data or inference data. Optionally, the near-real-time RIC can deliver inference results to the RAN node and / or the terminal. Optionally, the inference results can be exchanged between the CU and the DU, and / or between the DU and the RU. For example, the near-real-time RIC delivers the inference results to the DU, and the DU delivers the inference results to the RU.

[0128] The non-real-time RIC can also be used to perform model training and inference. For details, refer to the description of the near-real-time RIC for model training and inference, which is not repeated here.

[0129] In a possible design, the near-real-time RIC and the non-real-time RIC can be separately configured as a network element. Alternatively, the near-real-time RIC and the non-real-time RIC can be part of another device. For example, the near-real-time RIC is configured in a RAN node (for example, a CU or a DU), and the non-real-time RIC is configured in an OAM, a cloud server, a core network device, or another device.

[0130] In a possible implementation, the network device and / or the terminal can be deployed on land, including indoors or outdoors, handheld, or vehicle-mounted; can be deployed on water; or can be deployed on an aircraft, a balloon, or a satellite in the air. The present application does not limit the scenario in which the network device and the terminal are located. In addition, the terminal and the network device can be a hardware device, a software function running on a dedicated hardware device, a software function running on a general-purpose hardware device, such as a virtualized function instantiated on a platform (for example, a cloud platform), or an entity including a dedicated or general-purpose hardware device and a software function. The present application does not limit the specific form of the terminal and the network device.

[0131] It should be noted that the system described in the embodiments of the present application is to more clearly illustrate the technical solutions of the embodiments of the present application, and does not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0132] The communication method provided by the embodiments of the present application will be described below by taking the interaction between the terminal and the network device as an example in the communication system shown in FIG. 7. It should be noted that in the following embodiments of the present application, the names of messages between the terminal and the network device, the names of parameters, or the names of information, etc. are only examples, and in other embodiments, they can also be other names. The method provided by the present application does not make specific limitations on this.

[0133] It can be understood that in the embodiments of the present application, the terminal or the network device can perform part or all of the steps in the embodiments of the present application. These steps or operations are only examples, and the embodiments of the present application can also perform other operations or variations of various operations. In addition, each step can be executed in a different order according to the embodiments of the present application, and it is possible that not all operations in the embodiments of the present application are executed.

[0134] It can be understood that the terminal and the network device are taken as an example of the execution subject of the interaction in the present application, but the present application does not limit the execution subject of the interaction. For example, the method executed by the terminal in the present application can also be executed by a module (such as a chip, a chip system, a processor, a communication module, or a modem, etc.) applied to the terminal, and can also be realized by a logic node, a logic module or software that can realize all or part of the terminal function; the method executed by the network device in the present application can also be executed by a module (such as a chip, a chip system, a processor, a communication module, or a modem, etc.) applied to the network device, and can also be realized by a logic node, a logic module or software that can realize all or part of the terminal function.

[0135] The communication method provided by the embodiments of the present application will be described below. As shown in FIG. 11, the communication method can include the following steps:

[0136] S1101, the network device sends first information to the terminal. Correspondingly, the terminal receives the first information from the network device. The first information includes the information of the LLM simulator.

[0137] As a possible implementation, the LLM simulator is determined according to the LLM. For example, the LLM simulator is obtained by compressing the LLM, such as compressing the LLM by pruning, distillation, etc., or the LLM simulator can also be obtained by other processing of the LLM, which is not limited. The LLM can be a pre-trained model of the network device. For example, before step S1101, the network device can pre-train at least one LLM, then select one LLM from the at least one LLM, determine the LLM simulator based on the selected LLM, and then send the first information including the information of the LLM simulator.

[0138] For example, the LLM selected by the network device can be an LLM related to a specific communication task, which can be a communication task that the network device needs to complete or perform or participate in, such as beam scanning, processing based on channel state information (CSI), positioning, etc., which is not limited.

[0139] Among them, the LLM simulator hides the structure and parameters of the (original) LLM, or in other words, the LLM simulator hides the original structure and original parameters of the LLM. That is, the LLM simulator does not expose the structure and parameters of the LLM, and the structure and parameters of the LLM cannot be obtained through the LLM simulator. That is, the privacy of the LLM can be protected through the LLM simulator. Therefore, even if the network device sends the information of the LLM simulator to the terminal, the terminal cannot access the structure and parameters of the LLM.

[0140] As a possible implementation, the information of the LLM simulator can be the information of the LLM simulator itself, or can be the calling information of the LLM simulator, or can be the download indication information of the LLM simulator, etc. The calling information or download indication information of the LLM simulator may, for example, include an application programming interface (API) file of the LLM simulator, or an open neural network exchange (ONNX) file of the LLM simulator.

[0141] S1102, the network device sends second information to the terminal. Correspondingly, the terminal receives the second information from the network device.

[0142] Among them, the second information is used to indicate the structure of the adaptation layer. The adaptation layer includes an input adaptation layer, and the output of the input adaptation layer can be used as the input of the LLM simulator.

[0143] Optionally, the adaptation layer can further include an output adaptation layer, and the output of the LLM simulator can be taken as the input of the output adaptation layer. The output of the output adaptation layer can be taken as the final model output. In the absence of the output adaptation layer, the output of the LLM simulator can be taken as the final model output.

[0144] That is, the structural relationship between the adaptation layer and the LLM simulator can be as shown in FIG. 12. Referring to FIG. 12, the input adaptation layer and the output adaptation layer are linked in series at both ends of the LLM simulator to finally obtain the initial joint model. The input adaptation layer can be located before the LLM simulator, and the output thereof can be taken as the input of the LLM simulator. The output adaptation layer can be located after the LLM simulator, and the output of the LLM simulator can be taken as the input of the output adaptation layer. Meanwhile, the LLM simulator is obtained according to the LLM. It can be understood that the output adaptation layer in FIG. 12 is optional.

[0145] As a possible implementation, the input adaptation layer is used to implement the conversion of data dimensions and / or information domains. For example, in the model training phase, the input adaptation layer is used to convert the dimensions of the data samples into the input dimensions of the LLM simulator, and / or is used to convert the wireless corpus (or information in the wireless domain or communication domain) into the natural language corresponding to the LLM simulator.

[0146] As a possible implementation, the output adaptation layer is also used to implement the conversion of data dimensions and / or information domains. For example, in the model training phase, the output adaptation layer is used to convert the dimensions of the output data of the LLM simulator into the dimensions of the sample labels, and / or is used to convert the natural language corresponding to the LLM simulator into the relevant parameters of the specific communication task.

[0147] As a possible implementation, the second information includes the structural information of the input adaptation layer, such as the number of neural network layers, the number of neurons in each layer of the neural network, or the connection relationship between layers, the type of input parameters, the dimensions of input parameters, the type of output parameters, or the dimensions of output parameters, etc.

[0148] Optionally, the second information can further include the initial parameters of the input adaptation layer, such as the initial weighting value corresponding to each neuron, the activation function of the neuron, or the bias in the activation function, etc., without limitation.

[0149] Optionally, in the presence of the output adaptation layer, the second information can further include the structural information of the output adaptation layer and / or the initial parameters of the output adaptation layer. Reference can be made to the above description of the structural information and initial parameters of the input adaptation layer, which will not be repeated here.

[0150] Optionally, the structure of the output adaptation layer and the structure of the input adaptation layer can be the same or different. In the case of being the same, the second information can not include the structure information of the output adaptation layer; in the case of being different, the second information can include the structure information of the output adaptation layer.

[0151] As a possible implementation, the first information and the second information can be carried in the same message, for example, the first information and the second information can be carried in different fields of the same message; or the first information and the second information can also be carried in different messages, without limitation.

[0152] For example, in the case of carrying the first information and the second information in different messages, the network device can first send the first information, and then send the second information; or can first send the second information, and then send the first information; or can send the first information and the second information at the same time. That is, step S1101 can be performed first, and then step S1102 can be performed; or step S1102 can be performed first, and then step S1101 can be performed; or step S1101 and step S1102 can be performed at the same time.

[0153] S1103, the terminal trains according to the first information, the second information and the training data to obtain the adaptation layer.

[0154] As a possible implementation, step S1103 can also be understood as that the terminal trains the parameters of the adaptation layer according to the first information, the second information and the training data to obtain the adaptation layer (such as the final parameters of the adaptation layer).

[0155] As a possible implementation, the terminal can first link the adaptation layer and the LLM simulator together according to the first information and the second information to obtain an initial joint model, and then train the initial joint model based on the training data.

[0156] As a possible implementation, the training data can include a measurement quantity of a reference signal and a sample label obtained based on the measurement quantity. For example, the measurement quantity of part of the reference signal and the sample label obtained based on the measurement quantity of all the reference signals.

[0157] For example, the input of the initial joint model is determined by the measurement quantity of the reference signal, such as the measurement quantity of the reference signal can be directly taken as the input of the input adaptation layer; or the measurement quantity of the reference signal can be preprocessed, and the preprocessed data can be taken as the input of the input adaptation layer, and the application does not make any limitation on the preprocessing manner. In addition, the sample label obtained based on the measurement quantity of the reference signal can be taken as the target to be fitted for the model output.

[0158] Exemplarily, the reference signal can be an SSB or a CSI-RS. Of course, it can also be other types of reference signals such as a positioning reference signal (PRS) and the like, without limitation. The measurement quantity of the reference signal can be an RSRP, and of course, it can also be other types of measurement quantities such as a reference signal received quality (RSRQ) and the like, without limitation.

[0159] Taking the reference signal as an SSB and the measurement quantity of the reference signal as an RSRP as an example, all the reference signals can be understood as SSBs carried by all SSB beams sent by the network device in the full-band direction, or can be understood as SSBs carried by all candidate SSB beams. For example, taking the network device as an example that can send 64 SSB beams in the beam sweeping process, all the reference signals can be SSBs carried by the 64 SSB beams. The part of the reference signals can be understood as SSBs carried by part of the 64 SSB beams (such as 16 SSB beams therein). Correspondingly, the sample label obtained based on the measurement quantity of the reference signal can be the RSRP of all SSB beams, or can be the identifier of the optimal SSB beam, or can be the probability of each SSB beam in all SSB beams as the optimal beam, and the like, without limitation.

[0160] Taking the reference signal as a CSI-RS and the measurement quantity of the reference signal as an RSRP as an example, all the reference signals can be understood as CSI-RSs carried by all candidate CSI-RS beams, and the part of the reference signals can be understood as CSI-RSs carried by part of the candidate CSI-RS beams. Correspondingly, the sample label obtained based on the measurement quantity of the reference signal can be the CSI in the future period of time, or can be the RSRP of all CSI-RS beams, and the like, without limitation.

[0161] As a possible implementation, the manner in which the terminal performs data training can be supervised learning, unsupervised learning, reinforcement learning, and the like, which can be referred to the foregoing related description, and will not be described herein. Of course, the terminal can also use other manners for training, and the training manner used by the terminal is not limited in the present application.

[0162] S1104, the terminal sends third information to the network device. Correspondingly, the terminal receives the third information from the network device.

[0163] The third information is used for indicating the adaptation layer. For example, the third information is used for indicating the adaptation layer, which can also be understood as: the third information is used for feeding back the adaptation layer, and the two can be replaced with each other. For example, the third information includes parameters of the adaptation layer, such as parameters of the input adaptation layer, and optionally, parameters of the output adaptation layer. The parameters of the adaptation layer can be obtained by the terminal through the above step S1103.

[0164] S1105, the network device performs data inference according to the adaptation layer and the LLM.

[0165] As a possible implementation, the network device can first link the adaptation layer and the LLM together, such as linking the input adaptation layer and the output adaptation layer at the beginning and the end of the LLM to obtain a joint model, and then performing data inference based on the joint model. That is, the joint model is composed of the adaptation layer and the LLM. The input of the input adaptation layer is the input of the joint model, and the output of the input adaptation layer is the input of the LLM. In the absence of the output adaptation layer, the output of the LLM is the output of the joint model; in the presence of the output adaptation layer, the output of the LLM can be the input of the output adaptation layer, and the output of the output adaptation layer can be the output of the joint model.

[0166] For example, the network device performs data inference based on the joint model, which can include: the network device inputs data obtained according to the measurement of the first reference signal into the joint model to obtain the first wireless channel parameter and / or the first wireless transmission parameter. Subsequently, the network device can perform related processing based on the first wireless channel parameter and / or the first wireless transmission parameter, which is not limited.

[0167] For example, the first reference signal can include at least one reference signal, such as at least one SSB or at least one CSI-RS. The measurement of the first reference signal can be reported by the terminal to the network device. For example, before step S1105, the network device sends the first reference signal to the terminal, the terminal measures the first reference signal, and reports the measurement of the first reference signal.

[0168] For example, the network device can send the first reference signal to the terminal when a specific communication task is triggered or needs to be executed. For example, taking the first reference signal SSB as an example, the network device can send at least one SSB to the terminal when beam scanning is triggered; taking the first reference signal CSI-RS as an example, the network device can send at least one CSI-RS to the terminal when CSI needs to be obtained.

[0169] For example, when the first reference signal includes at least one SSB, the first wireless channel parameter may, for example, be the RSRP of all SSB beams, and the first wireless transmission parameter may, for example, be the identification of the optimal SSB beam or the identification of the top K SSB beams, where K may be 3, 4, or 5, etc. When the first reference signal includes at least one CSI-RS, the first wireless channel parameter may, for example, be the CSI. The following takes the first reference signal as an SSB, the output of the joint model as the identification of the top K SSB beams, and K as 3 as an example to describe the application of the joint model.

[0170] As shown in FIG. 13, the network device may perform two rounds of beam sweeping. In the first round of beam sweeping, the network device transmits SSBs on part of the candidate SSB beams, which are represented by the black squares in FIG. 13. Correspondingly, the terminal performs beam sweeping measurement and reports the measurement quantity of the SSB beams (i.e., the measurement quantity of the first reference signal) to the network device. After receiving the measurement quantity of the SSB beams reported by the terminal, the network device inputs the measurement quantity into the joint model, and the joint model outputs the indices of the top K SSB beams among all the candidate SSB beams. FIG. 13 takes K as 3 as an example for description.

[0171] In the second round of beam sweeping, the network device continues to transmit SSBs on the top 3 SSB beams obtained in the first round of beam sweeping, and correspondingly, the terminal performs measurement on the 3 SSB beams transmitted by the network device, selects the optimal SSB beam (e.g., the SSB beam with the highest RSRP) from the 3 SSB beams based on the measurement quantity, and reports the optimal SSB beam to the network device.

[0172] In the example shown in FIG. 13, the network device does not need to transmit SSBs on all SSB beams, for example, the network device transmits SSB beams accounting for 1 / 4 of the total beams in FIG. 13, but through the joint model, the network device can obtain the top K SSB beams among all the SSB beams. That is, in the case of reducing the number of transmitted SSB beams, the prediction result of all the SSB beams can still be obtained, that is, compared with the traditional beam sweeping, the beam sweeping overhead can be reduced.

[0173] In the first reference signal CSI-RS, in a case that the output of the joint model is the CSI of the channel in a future period of time, the application process of the joint model can include: the network device sends the CSI-RS, and correspondingly, the terminal measures the CSI-RS and reports the measurement quantity of the CSI-RS to the network device. After receiving the measurement quantity of the CSI-RS, the network device can input the measurement quantity into the joint model to obtain the CSI of the channel in the future period of time. Subsequently, the network device can perform transmission based on the predicted CSI. In this scenario, since the network device can predict the CSI in the future period of time through the joint model, the network device can not send the CSI-RS in the period of time, thereby saving the resource overhead of the CSI-RS.

[0174] It should be noted that the above only takes beam scanning and CSI prediction as an example to describe the application scenario of the present application, and in addition, the communication method provided by the present application can also be applicable to other communication scenarios, which are not limited by the present application.

[0175] Based on the above scheme, the network device can send the information of the LLM simulator and the information for indicating the adaptation layer structure to the terminal. The terminal trains to obtain the adaptation layer (such as the parameters of the adaptation layer) according to the information of the LLM simulator, the information for indicating the adaptation layer structure, and the training data, and indicates the adaptation layer to the network device. After receiving the information indicating the adaptation layer, the network device performs data inference according to the adaptation layer and the LLM. The adaptation layer includes an input adaptation layer, and the output of the input adaptation layer can be used as the input of the LLM (or the LLM simulator). In this scheme, since the network side does not directly send the LLM to the terminal, but sends the information of the LLM simulator, generally, the LLM simulator can be obtained by compressing the LLM, so the privacy of the LLM model is not disclosed. In addition, the terminal does not need to continuously report the training data, which can reduce the air interface overhead and avoid disclosing the data privacy of the terminal. Furthermore, the terminal only needs to train the adaptation layer, which can reduce the computational complexity and the requirement for the capability of the terminal compared with retraining the complete model for the communication task.

[0176] In a possible implementation, before the step S1101, the communication method can further include: the terminal sends the capability information to the network device, and correspondingly, the network device receives the capability information from the terminal. The capability information can be used to determine the compression ratio, that is, the compression ratio of the LLM simulator and the LLM.

[0177] As a possible implementation, the capability information includes at least one of memory, computing capability or power. Illustratively, the memory can be the total amount of memory of the terminal, the remaining amount of memory, the available amount of memory, etc. The computing capability can include floating-point operations per second (FLOPS), multiply accumulate operations (MACs), graphic processing unit (GPU) capability, etc. The power can be the remaining power, the available power, the current power, etc.

[0178] As a possible implementation, after receiving the capability information of the terminal, the network device can determine the compression ratio based on the capability information, compress the LLM according to the compression ratio to obtain the LLM simulator, and then send the first information. Illustratively, the higher the computing capability of the terminal, the lower the compression ratio can be, and the lower the compression ratio, the more conducive to protecting the privacy of the LLM.

[0179] Based on this possible implementation, the network device can reasonably set the compression ratio based on the terminal capability, thereby providing a suitable LLM simulator, avoiding problems in training the adaptation layer by the terminal due to unreasonable setting of the compression ratio, and improving the reliability of the adaptation layer training. In addition, in the case where the terminal does not report the capability information, in order to determine a reasonable compression ratio, the network device can need to first determine the LLM simulator based on an arbitrary compression ratio, send the information of the LLM simulator to the terminal, and then perform feedback to the network device in the case where the terminal has problems in training, and the network device adjusts the compression ratio based on the feedback of the terminal, and then sends the information of the LLM simulator determined based on the new compression ratio to the terminal, and so on until the terminal can train the adaptation layer. In the case where the terminal reports the capability information, the network device can directly determine the compression ratio based on the capability information, without performing the above-mentioned iterative process, thereby reducing the signaling overhead between the network device and the terminal, and improving the training efficiency and reducing the training delay of the adaptation layer.

[0180] In a possible implementation, before the step S1102, the communication method can further include: the terminal sends fourth information to the network device, and correspondingly, the network device receives the fourth information from the terminal device. The fourth information is used to indicate the dimension of the input parameter of the expected input adaptation layer, and / or the structure of the expected input adaptation layer.

[0181] Optionally, the fourth information can also be used to indicate the dimension of the output parameter of the expected output adaptation layer, and / or the structure of the expected output adaptation layer.

[0182] As a possible implementation, after receiving the fourth information, the network device can determine the structure of the input adaptation layer based on the fourth information. For example, the structure of the input adaptation layer is determined as the structure of the input adaptation layer expected by the terminal, or the structure of the input adaptation layer is set as a simpler structure than the structure expected by the terminal. In addition, the dimension of the input parameter of the input adaptation layer can also be set as the dimension of the input parameter expected by the terminal, or a lower dimension. The implementation of the network device determining the structure of the output adaptation layer can refer to the related description of determining the structure of the input adaptation layer, which will not be repeated here.

[0183] Based on this possible implementation, the network device can set a reasonable structure of the adaptation layer and / or dimension of the input / output parameter based on the structure expected by the terminal and / or dimension of the input / output parameter, thereby providing a suitable adaptation layer structure, avoiding problems in training the adaptation layer by the terminal due to unreasonable setting of the adaptation layer structure, and improving the efficiency of adaptation layer training.

[0184] In a possible implementation, after step S1105, the communication method can further include: the network device sending fifth information to the terminal, and correspondingly, the terminal receiving the fifth information from the network device. The fifth information is used to indicate updating the adaptation layer.

[0185] As a possible implementation, the network device can periodically or aperiodically send the fifth information based on the performance monitoring result. The performance monitoring result can be the communication performance of the network device using the output of the joint model for communication. For example, in the case of poor communication performance or substandard communication indicators, the network device can consider that the joint model can still be optimized, and therefore can send the fifth information to instruct the terminal to update the adaptation layer.

[0186] As a possible implementation, the fifth information can include at least one of the following: the structure and / or initial parameter of the updated input adaptation layer, the structure and / or initial parameter of the updated output adaptation layer, and the information of the updated LLM simulator. That is, the network device can first update the structure and / or initial parameter of the adaptation layer or the LLM simulator, and then instruct the terminal to update the adaptation layer based on the updated information.

[0187] As another possible implementation, the fifth information includes an update command or an update indication, which is used to instruct updating the adaptation layer. For example, the update command or the update indication can be 1-bit information.

[0188] As a possible implementation, after receiving the fifth information, the terminal can continue training based on the training data to optimize the adaptation layer, thereby updating the adaptation layer. For example, the training data can be different from the training data used for initial training of the adaptation layer (such as the training data used in step 1103).

[0189] Based on the possible implementation, the network device can timely instruct the terminal to update the adaptation layer to optimize the adaptation layer, thereby improving the performance of the communication task based on the joint model and improving the communication quality.

[0190] In the above method, the adaptation layer training by the terminal is taken as an example for illustration. In addition, a network element other than the terminal, such as an AI network element or a host or a cloud server of an OTT system, can also perform the training of the adaptation layer. In this scenario, as shown in FIG. 14, the communication method provided by the present application can include the following steps:

[0191] S1401, the network device sends first information to the AI network element / host / server of the OTT system. Correspondingly, the AI network element / host / server of the OTT system receives the first information from the network device.

[0192] The first information includes the information of the LLM simulator. For details, refer to the related description and implementation in step S1101 above, which will not be repeated here.

[0193] S1402, the network device sends second information to the AI network element / host / server of the OTT system. Correspondingly, the AI network element / host / server of the OTT system receives the second information from the network device.

[0194] The second information is used to indicate the structure of the adaptation layer. The adaptation layer includes an input adaptation layer. Optionally, the adaptation layer can also include an output adaptation layer. For details, refer to the related description and implementation in step S1102 above, which will not be repeated here.

[0195] S1403, the terminal sends training data to the AI network element / host / server of the OTT system. Correspondingly, the AI network element / host / server of the OTT system receives the training data from the terminal.

[0196] The training data can refer to the related description in step S1103 above, which will not be repeated here.

[0197] S1404, the AI network element / host / server of the OTT system performs training according to the first information, the second information, and the training data to obtain the adaptation layer.

[0198] For details of the training by the AI network element / host / server of the OTT system, refer to the related description and implementation of the training by the terminal in step S1103 above, which will not be repeated here.

[0199] S1405, the AI network element / host / server of the OTT system sends third information to the network device. Correspondingly, the network device receives the third information from the AI network element / host / server of the OTT system.

[0200] The third information indicates an adaptation layer. For details, refer to the description and implementation of step S1104.

[0201] S1406. The network device performs data inference according to the adaptation layer and the LLM. For details, refer to the description and implementation of step S1105.

[0202] Based on the scheme, the adaptation layer is trained on the host / server side of the AI network element / OTT system, and the terminal does not need to perform the training task, thereby reducing the computational complexity of the terminal and reducing the requirement on the capability of the terminal.

[0203] In a possible implementation, for the method embodiments described above, the functions of the network device in the CU-DU architecture or the ORAN system can be implemented by the CU, the DU, the CU and the DU. For example, the functions related to the model (or the functions related to the AI) implemented by the network device can be implemented by the AI module in the CU, the DU, the CU and the DU. In addition, the functions of the network device interacting with the terminal can be implemented by the DU or the O-DU. The information sent by the network device to the terminal can be generated by the DU or the O-DU, or can be generated by the CU or the O-CU and sent to the DU or the O-DU. The functions of the network device interacting with the core network can be implemented by the CU or the O-CU. The processing functions of the network device can be implemented by the CU or the O-CU, or can be implemented by the DU or the O-DU, or can be implemented jointly by the CU and the DU (or the O-CU and the O-DU), without limitation.

[0204] The above describes the method provided by the application, and the application further provides a communication apparatus for implementing the functions described in the above method embodiments.

[0205] It can be understood that, to implement the above functions, the communication apparatus includes hardware structures and / or software modules corresponding to the functions. Those skilled in the art can easily realize that, in combination with the units and algorithm steps of the examples described in the embodiments disclosed in the present application, the application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is realized in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0206] The embodiments of the present application can divide the functions of the communication device according to the method embodiments described above. For example, each function module can be divided according to each function, or two or more functions can be integrated into one processing module. The integrated module can be implemented in the form of hardware or software function module. It should be noted that the division of the modules in the embodiments of the present application is illustrative, and is only a logical function division. In actual implementation, there can be another division manner.

[0207] FIG. 15 shows a structural schematic diagram of a communication device 150. The communication device 150 includes a processing module 1501 and a transceiver module 1502. The communication device 150 can be used to implement the functions of the terminal or RAN node described above.

[0208] In some embodiments, the communication device 150 can further include a storage module (not shown in FIG. 15) for storing program instructions and data.

[0209] In some embodiments, the transceiver module 1502, which can also be referred to as a transceiver unit, is used to implement the sending and / or receiving functions. The transceiver module 1502 can be composed of a transceiver circuit, a transceiver, a transceiver, or a communication interface.

[0210] In some embodiments, the transceiver module 1502 can include a receiving module and a sending module, which are respectively used to perform the receiving and sending steps of the terminal or RAN node in the method embodiments described above, and / or other processes for supporting the technologies described herein; the processing module 1501 can be used to perform the processing steps of the terminal or RAN node in the method embodiments described above, and / or other processes for supporting the technologies described herein.

[0211] When the communication device 150 is used to implement the functions of the terminal:

[0212] The transceiver module 1502 is configured to receive first information from a network device, the first information including information of a large language model (LLM) simulator; the transceiver module 1502 is further configured to receive second information from the network device, the second information being used to indicate a structure of an adaptation layer, the adaptation layer including an input adaptation layer, an output of the input adaptation layer being an input of the LLM simulator; the processing module 1501 is configured to perform training according to the first information, the second information, and training data to obtain the adaptation layer; and the transceiver module 1502 is further configured to send third information to the network device, the third information being used to indicate the adaptation layer.

[0213] Optionally, the transceiver module 1502 is further configured to send capability information to the network device, the capability information being used to determine a compression ratio, the compression ratio being a compression ratio of the LLM simulator and an LLM, and the capability information including at least one of the following: memory, computing capability, or power.

[0214] Optionally, the transceiver 1502 is further configured to send, to the network device, fourth information, where the fourth information is used to indicate a dimension of an input parameter of a desired input adaptation layer and / or a structure of the desired input adaptation layer.

[0215] Optionally, the transceiver 1502 is further configured to receive, from the network device, fifth information, where the fifth information is used to indicate an updated adaptation layer.

[0216] When the communication apparatus 150 is configured to implement a function of a RAN node, the transceiver 1502 is further configured to receive, from a terminal, first information, where the first information includes information of a large language model (LLM) simulator; the transceiver 1502 is further configured to send, to the terminal, second information, where the second information is used to indicate a structure of an adaptation layer, and the adaptation layer includes an input adaptation layer; the first information and the second information are used to train the adaptation layer; the transceiver 1502 is further configured to receive, from the terminal, third information, where the third information is used to indicate the adaptation layer; and the processor 1501 is configured to perform data inference according to the adaptation layer and the LLM.

[0217] The transceiver 1502 is configured to send, to a terminal, first information, where the first information includes information of a large language model (LLM) simulator; the transceiver 1502 is further configured to send, to the terminal, second information, where the second information is used to indicate a structure of an adaptation layer, and the adaptation layer includes an input adaptation layer; the first information and the second information are used to train the adaptation layer; the transceiver 1502 is further configured to receive, from the terminal, third information, where the third information is used to indicate the adaptation layer; and the processor 1501 is configured to perform data inference according to the adaptation layer and the LLM.

[0218] Optionally, the processor 1501 is configured to perform data inference according to the adaptation layer and the LLM, including: inputting, into a joint model, data obtained according to a measurement quantity of a first reference signal, to obtain a first wireless channel parameter and / or a first wireless transmission parameter, where the joint model is composed of the adaptation layer and the LLM, an input of the input adaptation layer is an input of the joint model, and an output of the input adaptation layer is an input of the LLM.

[0219] Optionally, the transceiver 1502 is further configured to receive, from the terminal, capability information, where the capability information is used to determine a compression ratio, the compression ratio is a compression ratio of the LLM simulator and the LLM, and the capability information includes at least one of the following: memory, computing capability, or power.

[0220] Optionally, the transceiver 1502 is further configured to receive, from the terminal, fourth information, where the fourth information is used to indicate a dimension of an input parameter of a desired input adaptation layer and / or a structure of the desired input adaptation layer.

[0221] Optionally, the transceiver 1502 is further configured to send, to the terminal, fifth information, where the fifth information is used to indicate an updated adaptation layer.

[0222] The above method embodiments involve all related contents of each step, which can be referred to the function description of the corresponding functional module, and thus will not be repeated here.

[0223] In the present application, the communication apparatus 150 can be presented in the form of integrated division of various functional modules. The "module" here can refer to a specific application-specific integrated circuit (ASIC), a circuit, a processor and a memory executing one or more software or firmware programs, an integrated logic circuit, and / or other devices that can provide the above functions.

[0224] In some embodiments, when the communication apparatus 150 in FIG. 15 is a chip or a chip system, the function / implementation process of the transceiver module 1502 can be implemented through the input / output interface (or communication interface) of the chip or chip system, and the function / implementation process of the processing module 1501 can be implemented through the processor (or processing circuit) of the chip or chip system.

[0225] Since the communication apparatus 150 provided by the present embodiment can execute the above method, the technical effects it can obtain can refer to the above method embodiments, which will not be repeated here.

[0226] As a possible product form, the terminal or RAN node described in the embodiments of the present application can be implemented using one or more field programmable gate arrays (FPGAs), programmable logic devices (PLDs), controllers, state machines, gate logic, discrete hardware components, any other suitable circuits, or any combination of circuits capable of performing the various functions described throughout the present application.

[0227] As another possible product form, the terminal or RAN node described in the embodiments of the present application can be implemented by a general bus architecture. For ease of illustration, refer to FIG. 16, which is a structural schematic diagram of a communication apparatus 1600 provided by the embodiments of the present application, which includes a processor 1601 and a transceiver 1602. The communication apparatus 1600 can be a terminal, or a chip or chip system therein; or the communication apparatus 1600 can be a RAN node, or a chip or module therein. FIG. 16 only shows the main components of the communication apparatus 1600. In addition to the processor 1601 and the transceiver 1602, the communication apparatus can further include a memory 1603, and an input / output device (not shown in FIG. 16).

[0228] Optionally, the processor 1601 is mainly used for processing communication protocols and communication data, and controlling the whole communication apparatus, executing software programs, processing data of the software programs, so as to implement the methods provided in the above method embodiments. For example, the processor can be used for processing the measurement quantity of the reference signal, and performing the training task.

[0229] For example, the processor 1601 can implement a communication control function. In the case where the communication device 1600 is a terminal, the processor 1601 can control the communication device 1600 to establish a communication connection with a network device.

[0230] Optionally, the memory 1603 is mainly used for storing software programs and data. For example, the memory 1603 can store training data, information of an adaptation layer, information of an LLM simulator, and the like.

[0231] Optionally, the transceiver 1602 can include a radio frequency circuit and an antenna. The radio frequency circuit is mainly used for conversion between a baseband signal and a radio frequency signal and processing of the radio frequency signal. The radio frequency circuit can also be understood as a receiver / transmitter. The antenna is mainly used for transmitting and receiving radio frequency signals in the form of electromagnetic waves. For example, the transceiver can be used to transmit information between a terminal and a network device, such as the first information, the second information, and the third information described above.

[0232] Optionally, the radio frequency circuit and the antenna can be arranged independently of the processor that performs baseband processing. For example, in a distributed scenario, the radio frequency circuit and the antenna can be arranged in a remote manner independently of the communication device.

[0233] Optionally, the input / output device, for example, can include a touch screen, a display screen, a keyboard, and the like, and is mainly used for receiving user input data and outputting data to a user.

[0234] Optionally, the processor 1601, the transceiver 1602, and the memory 1603 can be connected through a communication bus.

[0235] When the communication device is powered on, the processor 1601 can read software programs in the memory 1603, execute instructions of the software programs, and process data of the software programs. When data needs to be transmitted wirelessly, the processor 1601 performs baseband processing on the data to be transmitted, and outputs a baseband signal to the radio frequency circuit. The radio frequency circuit performs radio frequency processing on the baseband signal, and transmits a radio frequency signal in the form of electromagnetic waves through the antenna. When data is transmitted to the communication device, the radio frequency circuit receives a radio frequency signal through the antenna, converts the radio frequency signal into a baseband signal, and outputs the baseband signal to the processor 1601. The processor 1601 converts the baseband signal into data and processes the data. In some embodiments, in terms of hardware implementation, those skilled in the art can conceive that the above-described communication device 150 can adopt the form of the communication device 1600 shown in FIG. 16.

[0236] As an example, the function / implementation process of the processing module 1501 in FIG. 15 can be implemented by invoking the computer-executed instructions stored in the memory 1603 by the processor 1601 in the communication apparatus 1600 shown in FIG. 16. The function / implementation process of the transceiver module 1502 in FIG. 15 can be implemented by the transceiver 1602 in the communication apparatus 1600 shown in FIG. 16.

[0237] As yet another possible product form, the terminal or RAN node in the present application can adopt the constituent structure shown in FIG. 17, or include the components shown in FIG. 17. FIG. 17 is a constituent diagram of a communication apparatus 1700 provided in the present application, which can be a terminal or a chip or system on chip in the terminal; or can be a RAN node or a chip or system on chip in the RAN node.

[0238] As shown in FIG. 17, the communication apparatus 1700 includes at least one processor 1701, and at least one communication interface (only one communication interface 1704 is shown in FIG. 17 by way of example, and the processor 1701 is taken as an example for description). Optionally, the communication apparatus 1700 can further include at least one of a communication bus 1702, a memory 1703, an output device 1705, and an input device 1706.

[0239] The processor 1701 can be a general central processing unit (CPU), a general processor, a network processor (NP), a digital signal processor (DSP), a microprocessor (such as X86, ARM), a microcontroller, an FPGA, a PLD, a state machine, gate logic, discrete hardware circuits, other suitable hardware configured to perform various functions, or any combination thereof. The processor 1701 can also be other apparatuses with processing capabilities, such as a circuit, a device, or a software module, without limitation.

[0240] The communication bus 1702 is used to connect different components in the communication device 1700 so that different components can communicate. The communication bus 1702 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is shown in FIG. 17, but it does not mean that there is only one bus or only one type of bus. For example, the communication bus 1702 can include any number of interconnected buses and bridges, depending on the specific application of the communication device and the overall design constraints. In addition, the communication bus 1702 can also link various other circuits, such as timing sources, peripherals, voltage regulators, and power management circuits, etc.

[0241] The communication interface 1704 is used for communication with other devices or communication networks. For example, the communication interface 1704 can be a module, a circuit or any device capable of realizing communication.

[0242] As a possible implementation, the communication interface 1704 can also be an input and output interface located in the processor 1701, used to realize the signal input and signal output of the processor.

[0243] As another possible implementation, the communication interface 1704 can also be understood as a bus interface. It is used to provide an interface between the communication bus and the transceiver. The transceiver can provide an interface or device for communicating with various other devices through wireless / wired transmission media. The transceiver can be coupled to an antenna array, and the transceiver and the antenna array can be used together to communicate with the network of the corresponding type.

[0244] The memory 1703 can be a device having a storage function, configured to store instructions and / or data. The instructions can be a computer program. For example, the memory 1703 can be a read-only memory (ROM) or another type of static storage device that can store static information and / or instructions, or a random access memory (RAM) or another type of dynamic storage device that can store information and / or instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or another optical disk storage, an optical disk storage (including a compact disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk storage medium or another magnetic storage device, etc., without limitation.

[0245] It should be noted that the memory 1703 can exist independently of the processor 1701, or can be integrated with the processor 1701. The memory 1703 can be located within the communication device 1700, or can be located outside the communication device 1700, without limitation.

[0246] The processor 1701 can be configured to execute instructions stored in the memory 1703, or to execute a computer program or instructions stored in the computer-readable storage medium 1707, to implement the methods provided by the embodiments described above.

[0247] For example, the processor 1701 can also implement the following functions, or the processor 1701 executes the instructions or computer programs stored in the memory 1703 or the computer-readable storage medium to implement the following functions: encoding, decoding, rate matching, de-rate matching, scrambling, descrambling, modulation, demodulation, layer mapping, fast Fourier transform (FFT), inverse fast Fourier transform (IFFT), inverse discrete Fourier transform (IDFT), precoding, resource element (RE) mapping, channel equalization, RE demapping, digital beam forming (BF), adding a cyclic prefix (CP), removing a CP, etc.

[0248] Optionally, the processor 1701 and / or the memory 1703 can include an AI module for implementing AI-related functions. The AI module can be implemented in software, hardware, or a combination of software and hardware. For example, the AI module can include a RIC module. For example, the AI module can be a near-real-time RIC or a non-real-time RIC.

[0249] As an optional implementation, the output device 1705 is in communication with the processor 1701 and can display information in a variety of ways. For example, the output device 1705 can be a liquid crystal display (LCD), a light emitting diode (LED) display device, a cathode ray tube (CRT) display device, or a projector, etc. The input device 1706 is in communication with the processor 1701 and can receive user input in a variety of ways. For example, the input device 1706 can be a mouse, a keyboard, a touch screen device, a sensor device, etc.

[0250] In some embodiments, on a hardware implementation, those skilled in the art can conceive that the communication apparatus 150 shown in FIG. 15 can take the form of the communication apparatus 1700 shown in FIG. 17.

[0251] As an example, the functions / implementation processes of the processing module 1501 in FIG. 15 can be implemented by the processor 1701 in the communication apparatus 1700 in FIG. 17 invoking computer execution instructions stored in the memory 1703. The functions / implementation processes of the transceiver module 1502 in FIG. 15 can be implemented by the communication interface 1704 in the communication apparatus 1700 in FIG. 17.

[0252] It should be noted that the structures shown in FIGS. 16 and 17 do not constitute a specific limitation on the terminal or RAN node. For example, in some other embodiments of the present application, the terminal or RAN node can include more or fewer components than shown, or combine certain components, or split certain components, or different component arrangements. The illustrated components can be implemented in hardware, software, or a combination of software and hardware.

[0253] In a possible implementation, the processor in the embodiments of the present application can include communication and processing circuitry. The communication and processing circuitry can include one or more hardware components that provide a physical structure that performs various processes related to wireless communication or sensing (such as signal reception and / or signal transmission). The communication and processing circuitry can include two or more transmit / receive chains. The functions implemented by the communication and processing circuitry can also be processed on a computer readable medium.

[0254] In some embodiments, this application also provides a communication device, which includes a processor for implementing the methods in any of the above method embodiments.

[0255] As one possible implementation, the communication device also includes a memory. This memory stores necessary computer programs and data. The computer program may include instructions, which a processor can invoke to instruct the communication device to execute the methods described in any of the above method embodiments. Alternatively, the memory may not be present in the communication device.

[0256] As another possible implementation, the communication device also includes an interface circuit, which is a code / data read / write interface circuit, used to receive computer execution instructions (which are stored in memory and may be read directly from memory or may be transmitted through other devices) and transmit them to the processor.

[0257] As another possible implementation, the communication device also includes a communication interface for communicating with modules outside the communication device.

[0258] It is understood that the communication device can be a chip or a chip system. When the communication device is a chip system, it can be composed of chips or may include chips and other discrete devices. This application does not specifically limit this.

[0259] This application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a computer, implements the functions of any of the above-described method embodiments.

[0260] This application also provides a computer program product that, when executed by a computer, implements the functions of any of the above method embodiments.

[0261] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0262] It is understood that the systems, apparatuses, and methods described in this application can also be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0263] The units described as separate components may or may not be physically separate, i.e., may be located in one place, or may be distributed to multiple network units. The components shown as units may or may not be physical units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0264] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit.

[0265] In the above embodiments, all or part can be realized by software, hardware, firmware or any combination thereof. When realized by software, all or part can be realized in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that a computer can access or include one or more data storage devices such as servers, data centers, etc. integrated with the medium. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state drive (SSD)), etc. In the embodiments of the present application, the computer can include the device described above.

[0266] Although the application has been described in connection with the embodiments thereof with reference to the various drawings, it will be understood that other variations and modifications of the details, and specific embodiments disclosed can be effected without departing from the application. In its broadest form, the application comprises the combinations of features of the application as described hereinabove. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit can fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0267] Although the application has been described in connection with specific embodiments thereof, it will be understood that it is capable of further modifications and this application is intended to cover any variations, uses or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. Accordingly, the description and drawings are to be regarded as illustrative in nature and not as restrictive.

Claims

1. A communication method characterized by comprising: The method comprises: receiving first information from a network device, the first information comprising information of a large language model (LLM) simulator; receiving second information from the network device, the second information being used to indicate a structure of an adaptation layer, the adaptation layer comprising an input adaptation layer, an output of the input adaptation layer being an input of the LLM simulator; training, according to the first information, the second information and training data, to obtain the adaptation layer; sending third information to the network device, the third information being used to indicate the adaptation layer.

2. The method of claim 1, wherein, The training data comprises a measurement quantity of a reference signal and a sample label obtained based on the measurement quantity, and an input of the input adaptation layer is determined by the measurement quantity of the reference signal.

3. The method according to claim 1 or 2, characterized in that, The adaptation layer further comprises an output adaptation layer, and an output of the LLM simulator is an input of the output adaptation layer.

4. The method according to any one of claims 1 to 3, characterized in that, The information of the LLM simulator comprises an application programming interface (API) file of the LLM simulator, or an open neural network exchange (ONNX) file of the LLM simulator.

5. The method according to any one of claims 1 to 4, characterized in that, The second information comprises structure information and / or initial parameters of the input adaptation layer.

6. The method of claim 5, wherein, The second information further comprises structure information and / or initial parameters of an output adaptation layer.

7. The method according to any one of claims 1 to 6, characterized in that, The LLM simulator is obtained by compressing an LLM.

8. The method according to any one of claims 1 to 7, characterized in that, The method further comprises: sending capability information to the network device, the capability information being used to determine a compression ratio, the compression ratio being a compression ratio of the LLM simulator and an LLM, and the capability information comprising at least one of the following: memory, computing capability or power.

9. The method according to any one of claims 1 to 8, characterized in that, The method further comprises: sending fourth information to the network device, the fourth information being used to indicate a desired dimension of an input parameter of the input adaptation layer, and / or a desired structure of the input adaptation layer.

10. The method of claim 9, wherein, The fourth information is further used to indicate a desired dimension of an output parameter of an output adaptation layer, and / or a desired structure of the output adaptation layer.

11. The method according to any one of claims 1 to 10, characterized in that, The method further comprises: receiving fifth information from the network device, the fifth information being used to indicate updating the adaptation layer.

12. The method of claim 11, wherein, The fifth information comprises at least one of the following: updated structure and / or initial parameters of the input adaptation layer, updated structure and / or initial parameters of an output adaptation layer, and updated information of the LLM simulator; Or, The fifth information comprises an update command, the update command being used to indicate updating the adaptation layer.

13. A method of communication, comprising: The method comprises: sending first information to a terminal, the first information comprising information of a large language model (LLM) simulator; sending second information to the terminal, the second information being used to indicate a structure of an adaptation layer, the adaptation layer comprising an input adaptation layer; the first information and the second information being used to train the adaptation layer; receiving third information from the terminal, the third information being used to indicate the adaptation layer; performing data inference according to the adaptation layer and an LLM.

14. The method of claim 13, wherein, The data inference according to the adaptation layer and the LLM comprises: inputting data obtained according to a measurement quantity of a first reference signal into a joint model to obtain a first wireless channel parameter and / or a first wireless transmission parameter, the joint model being composed of the adaptation layer and the LLM, an input of the input adaptation layer being an input of the joint model, and an output of the input adaptation layer being an input of the LLM.

15. The method according to claim 13 or 14, characterized in that, The adaptation layer further comprises an output adaptation layer, and an output of the LLM is an input of the output adaptation layer.

16. The method according to any one of claims 13-15, characterized in that, The information of the LLM simulator comprises an application program interface (API) file of the LLM simulator, or an open neural network exchange (ONNX) file of the LLM simulator.

17. The method according to any one of claims 13-16, characterized by, The second information comprises structure information and / or initial parameters of the input adaptation layer.

18. The method of claim 17, wherein, The second information further comprises structure information and / or initial parameters of the output adaptation layer.

19. The method according to any of claims 13-18, characterized by, The LLM simulator is obtained by compressing an LLM.

20. The method according to any one of claims 13-19, characterized in that, The method further comprises: receiving capability information from the terminal, the capability information being used to determine a compression ratio, the compression ratio being a compression ratio of the LLM simulator and the LLM, and the capability information comprising at least one of the following: memory, computing capability, or power.

21. The method according to any one of claims 13-20, characterized in that, The method further comprises: receiving fourth information from the terminal, the fourth information being used to indicate a dimension of an input parameter of the input adaptation layer expected and / or a structure of the input adaptation layer expected.

22. The method of claim 21, wherein, The fourth information is further used to indicate a dimension of an output parameter of the output adaptation layer expected and / or a structure of the output adaptation layer expected.

23. The method according to any one of claims 13-22, characterized by, The method further comprises: sending fifth information to the terminal, the fifth information being used to indicate updating of the adaptation layer.

24. The method of claim 23, wherein, The fifth information comprises at least one of the following: updated structure and / or initial parameters of the input adaptation layer, updated structure and / or initial parameters of the output adaptation layer, and updated information of the LLM simulator. Or, The fifth information comprises an update command, the update command being used to indicate updating of the adaptation layer.

25. A communications device, characterized by The communication device comprises a processor, and the processor is used to run a computer program or instructions to enable the communication device to perform the method according to any one of claims 1 to 12, or to enable the communication device to perform the method according to any one of claims 13 to 24.

26. A computer readable storage medium, characterized in that, A computer readable storage medium stores computer instructions or programs, and when the computer instructions or programs are run on a computer, the method according to any one of claims 1 to 12 is performed, or the method according to any one of claims 13 to 24 is performed.

27. A computer program product, characterised in that, The computer program product comprises computer instructions, and when part or all of the computer instructions are run on a computer, the method according to any one of claims 1 to 12 is performed, or the method according to any one of claims 13 to 24 is performed.

Images