Abstract: This application provides an AI model switching or updating method. A change of a use environment (a moving speed of an inference network element, a channel environment, and the like) of an AI model can be learned of by monitoring an input or intermediate performance indicator of the AI model, for example, monitoring the input indicator of the AI model or monitoring the intermediate performance indicator of the AI model. In this way, whether the AI model needs to be switched or updated is determined based on correspondence information, to adapt to the change of the use environment of the AI model. This helps alleviate a problem that performance of the AI model decreases or deteriorates due to a great change of the use environment of the AI model.

Abstract: In the field of communication technologies, a random access preamble sending method and a communication apparatus are provided to avoid a problem that a conflict exists between a PRACH time-frequency resource occupied by a terminal device for sending a random access preamble and a PUSCH time-frequency resource occupied by another terminal device for sending uplink data, and consequently a network device cannot normally receive the random access preamble and the uplink data. The method includes: a terminal device receives configuration information sent by a network device, where the configuration information includes configuration information of a PRACH time-frequency resource set and configuration information of a PUSCH time-frequency resource set. The terminal device determines, in the PRACH time-frequency resource set, a PRACH time-frequency resource that does not conflict with the PUSCH time-frequency resource set.

Abstract: Embodiments of this application provide a communication method and a communication apparatus. The method includes: A terminal device determines a first parameter value of a first parameter based on a measurement result of a first downlink reference signal and a first precoder, where the first parameter reflects downlink channel quality; and obtains a parameter value of a second parameter based on a first compensation amount and the first parameter value of the first parameter, and reports the parameter value of the second parameter, where the first compensation amount indicates a difference between downlink channel quality in a case in which a network device uses the first precoder and downlink channel quality in a case in which the network device uses a second precoder. In the solutions in embodiments of this application, the parameter value reported by the terminal device can accurately reflect actual downlink channel quality.

Abstract: A first network element (a training data collection network element) receives first information from a second network element (an AI model training network element), the first information being for determining validity of candidate training data collected by the first network element. The first network element collects candidate training data of an AI model, and determines the validity of the candidate training data based on the first information. When the candidate training data is valid, the first network element sends valid candidate training data to the second network element, but does not send invalid candidate training data. When there is no valid candidate training data, the first network element indicates, to the second network element, that the training data collected this time is invalid, and does not send the candidate training data collected this time to the second network element. Waste of air interface resources can be reduced.

Abstract: Embodiments of this application provide a model matching method and a communication apparatus. The method includes: receiving an identifier of a first artificial intelligence AI model from a first communication apparatus; and determining, based on the identifier of the first AI model, whether a second AI model is deployed on a second communication apparatus, where the second AI model is an AI model that matches the first AI model. In this way, an identifier of an AI model can be exchanged to determine, based on the identifier of the AI model, an AI model that matches the AI model.

Abstract: This disclosure provides a beam management method, including: A communication device like a terminal device or an access network device maps a reference signal measurement quantity to input adaptation information by using an input adaptation layer, where a beam pattern corresponding to the reference signal measurement quantity is a first beam pattern; and obtains a first beam prediction result by using a beam prediction model, where an input of the beam prediction model includes the input adaptation information, the input of the beam prediction model matches a second beam pattern, and the first beam pattern is different from the second beam pattern. According to the method, system overheads can be reduced.

Abstract: A data quality measurement method and an apparatus are provided. The method includes: A data quality measurement apparatus sends first configuration information to a first data collection node, where the first configuration information is used to configure feature extraction in similarity measurement, the similarity measurement is used to determine a degree of similarity in statistical distribution between different datasets, and the feature extraction is used to extract a feature of a data sample included in the dataset. The data quality measurement apparatus receives a first feature set of a first dataset from the first data collection node, where the first feature set is based on the first configuration information. Accordingly, the data quality measurement apparatus can determine whether the dataset collected by the data collection node meets a low similarity requirement.

Abstract: In accordance with an embodiment, a method includes encoding channel information of N downlink transport layers using a first encoder to determine N pieces of first channel state indication information wherein N is a positive integer greater than 1; encoding the N pieces of first channel state indication information using a second encoder to determine second channel state indication information, wherein the second channel state indication information corresponds to the N downlink transport layers, and a sequence length corresponding to the second channel state indication information is less than a sum of sequence lengths corresponding to the N pieces of first channel state indication information; and sending the second channel state indication information.

Abstract: A model application method and an apparatus are provided. The method includes: determining first information based on a received reference signal, where the first information includes data of m ports for the reference signal; and mapping the data of the m ports to an input of a first model based on a correspondence between the m ports and n ports in an input dimension of the first model, to obtain an output of the first model, where both m and n are integers greater than or equal to 1, and m and n are not equal. By using the method and the apparatus in this application, self-adaptation to a port of the reference signal can be implemented, and model storage and training overheads can be reduced.

Abstract: In accordance with an embodiment, a method includes: obtaining a first manner, where the first manner indicates a first downlink reference signal that is among received downlink reference signals and that is used to obtain first channel environment information, the first channel environment information is used as an input of a first artificial intelligence (AI) model, and the first channel environment information is current channel environment information; generating a first uplink signal at an output of the first AI model; and sending the first uplink signal.

Abstract: A reference signal indication method includes receiving a plurality of reference signals, and sending at least one codeword on a first time-frequency resource. The at least one codeword includes a first codeword. The first codeword corresponds to a first reference signal in N reference signals of the plurality of reference signals. The first reference signal is a selected reference signal. The N reference signals are a subset of the plurality of reference signals, and N is a positive integer.

Abstract: A random access method and an apparatus. In the method, random access is completed in a manner in which an agent terminal performs access, so that an access requirement of a terminal can be satisfied under low energy consumption. The method and the apparatus may be applied to application scenarios such as an internet of things (IoT), machine type communication (MTC), massive machine type communication (mMTC), an internet of vehicles, and a connected vehicle.

Abstract: A method includes: When user equipment is in an RRC idle mode or inactive mode, the user equipment sends a first random access preamble to an access network device on a PRACH resource, and sends first uplink data to the access network device on a PUSCH resource. The user equipment sends, by using at least two beams respectively, SRSs to the access network device on at least two SRS resources associated with the PUSCH resource. The user equipment receives a random access response from the access network device, where the random access response includes beam indication information, and the beam indication information is used to indicate an optimal beam in the at least two beams.

Abstract: This disclosure provides a communication method and a communication apparatus. The method includes: receiving first information from a first node, where the first information is for configuring a first model structure of an intelligent model, the first model structure includes N operations, N is a positive integer, an operation in the N operations is a calculation for obtaining an output parameter by using an operation parameter and an operator, and the operator is an operation rule or a function; and determining the intelligent model based on the first information. This enables communication nodes to reach a consensus on the model structure, makes the intelligent model applicable to a communication network, and implements communication network intelligence.

Abstract: A data processing method for a model and an apparatus are provided. The method includes: A first node determines the data processing method for the model; and the first node implements, according to the data processing method for the model, at least one of the following: performing model training or performing model inference. Based on the method and the apparatus, inference tasks for different radio resource configurations can be completed by using one model, so that overheads for model training and storage are reduced.

Abstract: An uplink precoding method and an apparatus are provided. The method includes: A terminal device receives first information from an access network device; the terminal device determines a first precoding matrix based on the first information and a first model, where the first precoding matrix is determined based on an output of the first model, and an input of the first model is determined based on the first information; the terminal device precodes an uplink signal based on the first precoding matrix; and the terminal device sends a precoded uplink signal to the access network device. In the method, the terminal device determines the first precoding matrix based on the first information and the first model. The precoding matrix is no longer selected from an offline codebook, but is a floating-point matrix generated by the terminal device based on the first information.

Abstract: This disclosure provides a communication method and apparatus. An apparatus obtains an ith piece of first information that indicates an ith local model of each of K subnodes and i is a positive integer, and determines an ith federated learning model of a kth subnode based on the ith piece of first information and a type of the ith federated learning model of the kth subnode, where k is any positive integer from 1 to K. When i is greater than 1, the type of the ith federated learning model of the kth subnode is determined based on a type or performance of an (i?1)th federated learning model of the kth subnode. The apparatus sends an ith piece of second information, where the ith piece of second information indicates an ith federated learning model of each of the K subnodes.

Abstract: This disclosure provides a communication method and apparatus, to reduce transmission overheads of training data and improve performance of an artificial intelligence model. The method includes: obtaining first data and first information from a first node, where the first information indicates a data augmentation manner of the first data; and determining a first training dataset of a model based on the first data and the first information.

Abstract: A communication method and a communication apparatus. The method includes: the communication apparatus receives first information, where the first information indicates a training policy of a first intelligent model. The communication apparatus performs model training on the first intelligent model according to the training policy. In this way, an intelligent model that meets a communication performance requirement can be obtained while air interface resource overheads are reduced.

Abstract: This application provides a data processing method and a communication apparatus. The method includes: receiving or sending a plurality of reference signals; obtaining a first channel data set, where the first channel data set is a set of a plurality of pieces of channel data obtained based on the plurality of reference signals; obtaining a data quality screening policy, where the data quality screening policy is used to obtain, through screening, model training data that meets a data quality requirement; and screening the channel data in the first channel data set according to the data quality screening policy, to obtain a training data set. In this way, quality of the training data for model training is improved, so that a trained intelligent model meets a performance requirement.