Configuration method and apparatus for ai / ML model or function

Abstract

Embodiments of the present application provide a configuration method and apparatus for an AI / ML model / function. The method comprises: a terminal device sending first information to a network device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device; and the terminal device receiving second information sent by the network device, the second information being used to indicate a transform domain operation for the input of the AI / ML model / function.

Description

Configuration method and apparatus of AI / ML model or function TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of communication technology. BACKGROUND

[0002] In RAN1#118 meeting, companies selected the research cases (Case 0 / 1 / 2 / 3 / 4 / 5) for Rel-19 (Release 19), and gave priority to Case 2 (Case 2) and Case 3 (Case 3), and reached the following consensus (agreement).

[0003] In RAN1#116 meeting, in order to solve the problem of cross-vendor collaboration difficulty, 3GPP (3rd Generation Partnership Project) defined the following 5 options.

[0004] In the Rel-18 (Release) stage, 3GPP studied the AI (Artificial Intelligence) / ML (Machine Learning) based CSI (Channel state information) compression feedback use case in the space-frequency domain. In this use case, the AI / ML model (AI encoder / CSI generation part) on the terminal side (UE side) compresses the measured precoding and generates the corresponding PMI (Precoding Matrix Indicator) feedback bits, and the AI / ML model (AI decoder / CSI generation part) on the network side (NW side) reconstructs / recoveres / decompresses the precoding vector based on the PMI feedback bits.

[0005] For the input of the AI / ML model, in addition to considering the space-frequency domain in the Rel-18 stage, the “angle-delay” transform domain is further considered, that is, the space-frequency domain is subjected to 2-dimensional Fourier transform to obtain the angle-delay domain, so as to reduce the sparsity or dimension of the model input, thereby reducing the complexity / performance of the AI / ML model. Corresponding to the reduction of the dimension, since the “angle-delay” transform domain is more sparse, the amplitudes of many coefficients on the basis are small and can be discarded, and only the strong coefficients on the transform domain basis are retained.

[0006] It should be noted that the above introduction to the technical background is only for the convenience of clearly and completely describing the technical scheme of the present application and facilitating the understanding of those skilled in the art. The above technical scheme cannot be considered as known to those skilled in the art merely because it is described in the background section of the present application.

[0007] SUMMARY

[0008] The inventors find that in Rel-19 (Release) phase, 3GPP further studies the time-domain CSI compression feedback use case Case 3, expecting to obtain higher performance gain. However, for use case 3, there is currently no specific transform domain scheme. In addition, when the UE and / or NW do not know whether a transform domain scheme is adopted and / or the specific design of the transform domain scheme, the UE and / or NW can only train / reason / performance monitor the AI / ML model through guessing / trial and error, etc., at this time the AI / ML model of both ends may not be completely matched, or the performance is not optimal.

[0009] To at least one of the above problems or other similar problems, the embodiments of the present application provide an AI / ML model / function configuration method and device.

[0010] According to an aspect of the embodiments of the present application, an AI / ML model / function configuration method is provided, applied to a terminal device, the method comprising:

[0011] The terminal device sends first information to a network device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device;

[0012] The terminal device receives second information sent by the network device, the second information being used to indicate a transform domain operation of an input of an AI / ML model / function.

[0013] According to another aspect of the embodiments of the present application, an AI / ML model / function configuration method is provided, applied to a network device, the method comprising:

[0014] The network device receives first information sent by a terminal device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device;

[0015] The network device sends second information to the terminal device, the second information being used to indicate a transform domain operation of an input of an AI / ML model / function.

[0016] According to still another aspect of the embodiments of the present application, an AI / ML model / function configuration device is provided, configured in a terminal device, the device comprising:

[0017] a sending unit configured to send, to the network device, first information related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device;

[0018] a receiving unit configured to receive second information sent by the network device, the second information being used to indicate a transform domain operation of an input of an AI / ML model / function.

[0019] According to a further aspect of the embodiments of the present application, an AI / ML model / function configuration apparatus is provided, configured in a network device, the apparatus comprising:

[0020] a receiving unit configured to receive first information sent by the terminal device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device;

[0021] a sending unit configured to send, to the terminal device, second information used to indicate a transform domain operation of an input of an AI / ML model / function.

[0022] One of the beneficial effects of the embodiments of the present application is that the terminal device and the network device interact the information related to the transform domain of the input of the AI / ML model / function, such as whether the input of the AI / ML model / function adopts a transform domain, a specific transform domain scheme, and the like, so as to facilitate the training, inference, and performance monitoring of the AI / ML model / function, reduce the complexity, avoid performance loss, and improve the performance of the communication system.

[0023] Specific embodiments of the application are disclosed herein, and represented in the accompanying drawings, illustrating the principles of the application in a manner that is sufficiently detailed to be understood by those skilled in the art. It should be understood that the embodiments of the application are not limited in scope to the specific embodiments described herein. Embodiments of the application include many changes, modifications, and equivalents.

[0024] Features described and / or illustrated with respect to one implementation can be used in the same or similar manner in one or more other implementations, in combination with other features in the other implementations, or in place of other features in the other implementations.

[0025] It should be emphasized that the term "comprises / comprising" when used in this text is taken to mean the presence of stated features, integers, steps or components, but not to the exclusion of one or more other features, integers, steps or components, or groups thereof. BRIEF DESCRIPTION OF DRAWINGS

[0026] Elements and features depicted in one drawing or embodiment of the application can be combined with elements and features depicted in one or more other drawings or embodiments, as

[0027] FIG. 1 is a schematic diagram of a communication system according to an embodiment of the application;

[0028] FIG. 2 is a schematic diagram of an AI / ML model corresponding to use case 3;

[0029] FIG. 3 is a schematic diagram of a configuration method of an AI / ML model / function according to an embodiment of the application;

[0030] FIG. 4 is a schematic diagram of a CSI compression feedback based on an angle-delay-Doppler domain;

[0031] FIG. 5 is a schematic diagram of a configuration method of an AI / ML model / function according to an embodiment of the application;

[0032] FIG. 6 is a schematic diagram of a configuration apparatus of an AI / ML model / function according to an embodiment of the application;

[0033] FIG. 7 is another schematic diagram of a configuration apparatus of an AI / ML model / function according to an embodiment of the application;

[0034] FIG. 8 is a schematic block diagram of a system configuration of a network device according to an embodiment of the application;

[0035] FIG. 9 is a schematic block diagram of a system configuration of a terminal device according to an embodiment of the application. DETAILED DESCRIPTION

[0036] The foregoing and other features of the application are hereinafter more fully described and understood when considered in connection with the following drawings. In the drawings, specific embodiments of the application are illustrated, which show, by way of illustration, the principles of the application. It will be appreciated that the application is not limited to the particular embodiments described herein but rather the application includes all modifications, variations, and equivalents that fall within the scope of the appended claims. Various embodiments of the application are discussed in conjunction with the attached drawings. These embodiments are illustrative of the application and do not limit the scope of the application as defined by the appended claims.

[0037] In the embodiments of the present application, the terms "first", "second" and the like are used to distinguish different elements from each other, but do not indicate spatial arrangement or time sequence of the elements, and the elements should not be limited by these terms. The term "and / or" includes any one and all combinations of the associated listed terms. The terms "comprise", "include", "have" and the like mean the presence of the stated features, elements, elements or components, but do not exclude the presence or addition of one or more other features, elements, elements or components.

[0038] In the embodiments of the present application, the singular form "a", "an" and the like includes the plural form, should be broadly understood as "one" or "a kind of", and not limited to the meaning of "one"; in addition, the term "said" should be understood to include both singular and plural forms, unless the context clearly indicates otherwise. In addition, the term "according to" should be understood as "at least partially according to", and the term "based on" should be understood as "at least partially based on", unless the context clearly indicates otherwise.

[0039] In the embodiments of the present application, the term "communication network" or "wireless communication network" can refer to a network that complies with any communication standard, such as New Radio (NR), Long Term Evolution (LTE), LTE-Advanced (LTE-A), Wideband Code Division Multiple Access (WCDMA), High-Speed Packet Access (HSPA), and the like.

[0040] In addition, the communication between devices in the communication system can be carried out according to any stage communication protocol, which can include but is not limited to the following communication protocols: 1G (generation), 2G, 2.5G, 2.75G, 3G, 4G, 4.5G and 5G, New Radio (NR), 6G and future communication, and the like, and / or other currently known or future to be developed communication protocols.

[0041] In the embodiments of the present application, the term "network device" refers to, for example, a device that accesses a terminal device to a communication network and provides services for the terminal device in a communication system. The network device can include, but is not limited to, the following devices: base station (BS), access point (AP), transmission reception point (TRP), broadcast transmitter, mobile management entity (MME), gateway, server, radio network controller (RNC), base station controller (BSC), and the like.

[0042] Wherein, the base station can include, but is not limited to: Node B (NodeB or NB), evolved Node B (eNodeB or eNB), 5G base station (gNB), 6G base station, and future base station, etc., and can also include remote radio head (RRH), remote radio unit (RRU), relay, or low-power node (such as femto, pico, etc.). And the term "base station" can include some or all functions of them, and each base station can provide communication coverage for a specific geographic area. The term "cell" can refer to a base station and / or its coverage area, depending on the context in which the term is used.

[0043] In the embodiments of the present application, the term "user equipment" (UE) or "terminal equipment" (TE) refers to, for example, a device that accesses a communication network through a network device and receives network services. The user equipment can be fixed or mobile, and can also be referred to as a mobile station (MS), a terminal, a user, a subscriber station (SS), an access terminal (AT), a station, a mobile terminal (MT), and the like.

[0044] Wherein, the terminal device can include, but is not limited to, the following devices: cellular phone, personal digital assistant (PDA), wireless modem, wireless communication device, handheld device, machine type communication device, laptop computer, cordless phone, smart phone, smart watch, digital camera, and the like.

[0045] For example, in scenarios such as Internet of Things (IoT), user equipment can also be a machine or device for monitoring or measurement, which can include but is not limited to: machine type communication (MTC) terminal, vehicle-mounted communication terminal, device to device (D2D) terminal, machine to machine (M2M) terminal, terminal supporting sidelink communication, and the like.

[0046] In addition, the term "network side" or "network device side" refers to the side of the network, which can be a certain base station, or can include one or more network devices as described above. The term "user side" or "terminal side" or "terminal device side" refers to the side of the user or terminal, which can be a certain UE, or can include one or more terminal devices as described above. In this article, "device" can refer to a network device or a terminal device without special indication.

[0047] In the embodiments of the present application, "at least one" and "one or more" can be interchangeable, "multiple" and "more than one" can be interchangeable, and "multiple" means at least two, or two or more.

[0048] In the embodiments of the present application, pre-defined means defined by a protocol or determined according to rules defined by a protocol, without additional configuration. Configuration / indication means direct or indirect configuration / indication by a network device through high layer signaling and / or physical layer signaling. High layer parameters can be configured / indicated by introducing high layer parameters in high layer signaling, which refer to information fields and / or information elements / units / members (IE) in high layer signaling, etc. Physical layer signaling refers to control information (DCI) carried by physical downlink control channel or sequence, but is not limited thereto.

[0049] In the embodiments of the present application, "time" and "slot" are expressions of time units, and in the following description, time or slot is taken as an example for illustration, but the present application is not limited thereto. According to different specific real-time scenarios, time or slot in the text can also be replaced by other time units, such as slot group, symbol, symbol group, etc.

[0050] In the following description, "if", "in the case of" and "when" can be used interchangeably without causing confusion.

[0051] The scenarios of the embodiments of the present application are described below by way of example, but the present application is not limited thereto.

[0052] FIG. 1 is a schematic diagram of a communication system according to an embodiment of the present application, which illustrates an example of a terminal device and a network device, as shown in FIG. 1, the communication system 100 can include a network device 101, a terminal device 102, and a terminal device 103. For simplicity, FIG. 1 only illustrates an example of two terminal devices and one network device, but the embodiments of the present application are not limited thereto.

[0053] In the embodiments of the present application, the network device 101, the terminal device 102, and the terminal device 103 can perform existing services or future implementable services transmission. For example, these services can include but are not limited to: enhanced mobile broadband (eMBB), massive machine type communication (mMTC), ultra-reliable and low-latency communication (URLLC), and related communication of reduced-capability terminal devices, etc.

[0054] Among them, the terminal devices 102 and 103 can be in an RRC_IDLE state, or an RRC_INACTIVE state or an RRC_CONNECTED state, and the terminal devices 102 and 103 can also communicate with the network device 101, for example, taking the terminal device 102 as an example, the terminal device 102 can send data to the network device 101, or can perform data retransmission. The network device 101 can send a paging message to the terminal device 102, and can also send data to the terminal device 102, and the terminal device 102 receives the data sent by the network device 101. In addition, different terminal devices can also communicate with each other, for example, the terminal device 102 and the terminal device 103 can exchange data.

[0055] It is worth noting that FIG. 1 shows that the terminal device 102 and the terminal device 103 are both within the coverage of the network device 101, but the present application is not limited thereto. The terminal device 102 and the terminal device 103 can both be outside the coverage of the network device 101, or one of the terminal device 102 and the terminal device 103 is within the coverage of the network device 101 and the other is outside the coverage of the network device 101.

[0056] In the embodiments of the present application, one or more AI / ML models can be configured and run in the network device and / or the terminal device. The AI / ML model can be used for various signal processing functions of wireless communication, such as CSI prediction, CSI compression, beam prediction, positioning management, etc.; the present application is not limited thereto.

[0057] FIG. 2 is a schematic diagram of an AI / ML model corresponding to Case 3. As shown in FIG. 2, the CSI information of multiple time instants / slots is jointly compressed to achieve a lower compression rate or a higher feedback accuracy. The CSI information of multiple time instants / slots can be obtained by prediction and then fed into the CSI compression model. Due to the correlation between time slots, the performance of the joint CSI compression scheme (Case 3) will be better than that of the independent (non-joint) CSI compression scheme (e.g., Case 0).

[0058] However, as mentioned above, currently, there is no specific transform domain scheme for Case 3. In addition, when the UE / NW does not know whether a transform domain scheme is adopted and / or the specific design of the transform domain scheme, the UE / NW can only train / reason / performance monitor the AI / ML model by guessing / trial and error, etc., in which case the AI / ML models of both ends may not be completely matched, or the performance is not optimal. To address at least one of the above problems or other similar problems, the present application is proposed.

[0059] In embodiments of the present application, CSI compression can also be referred to as "CSI encoding", "CSI generation", and the like. The result of the CSI compression, or CSI encoding, or CSI generation, and the like, is referred to as CSI feedback information, CSI reporting information, and the like.

[0060] In embodiments of the present application, CSI decompression can also be referred to as "CSI decoding", "CSI reconstruction", "CSI recovery", "CSI reconstruction", and the like.

[0061] In embodiments of the present application, the AI / ML model can also be referred to as an AI / ML method, an AI / ML unit, an AI / ML function, or an AI / ML element, and the like. On the UE side, the AI / ML model can also be referred to as an encoder, a CSI generation part, and the like. On the network device side, the AI / ML model can also be referred to as a decoder, a CSI reconstruction part, and the like.

[0062] In embodiments of the present application, for convenience of description, "training an AI / ML model or training a model supporting an AI / ML function" is collectively referred to as "training an AI / ML model / function", and "deploying an AI / ML model or deploying a model supporting an AI / ML function" is collectively referred to as "deploying an AI / ML model / function".

[0063] It should be noted that the operation processing method of the AI / ML model or function of the embodiments of the present application is applicable to use cases including but not limited to CSI compression feedback, but those skilled in the art can understand that the embodiments of the present application are also applicable to other use cases and / or scenarios of various application AI / ML models or functions (referred to as AI / ML models / functions).

[0064] The specific embodiments of the embodiments of the present application will be exemplarily described below in conjunction with the accompanying drawings.

[0065] Embodiments of the first aspect

[0066] The embodiments of the present application provide a configuration method of an AI / ML model / function, which is described from the side of a terminal device.

[0067] FIG. 3 is a schematic diagram of a configuration method of an AI / ML model / function according to an embodiment of the present application, as shown in FIG. 3, the method comprises:

[0068] 310: The terminal device sends first information to the network device, and the first information is related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device; and

[0069] 320: The terminal device receives second information sent by the network device, and the second information is used to indicate a transform domain operation of the input of the AI / ML model / function.

[0070] In the above embodiments, "related to a transform domain of an input of an AI / ML model / function" for example refers to whether the input of the AI / ML model / function adopts a transform domain, and / or specific content of the transform domain adopted by the input of the AI / ML model / function. The input of the AI / ML model / function adopts a transform domain for example refers to that the input of the AI / ML model / function has a transform domain operation, for example a two-dimensional transform, including but not limited to converting from a space-frequency domain to an angle-time delay domain, and for example a three-dimensional transform, including but not limited to converting from a space-frequency-time domain to an angle-time delay-Doppler domain.

[0071] It is worth noting that the above FIG. 3 only exemplarily illustrates the embodiments of the present application, but the present application is not limited thereto. For example, some other operations can be added. Those skilled in the art can make appropriate modifications according to the above content, and not only limited to the description of the above FIG. 3.

[0072] According to the above embodiments, the terminal device and the network device interact the related information of the transform domain of the input of the AI / ML model / function, such as whether the input of the AI / ML model / function adopts a transform domain, a specific transform domain scheme, and the like, so as to facilitate the training, inference and performance monitoring of the AI / ML model / function, reduce the complexity, avoid performance loss, and improve the performance of the communication system.

[0073] In some embodiments, the second information indicates whether the AI / ML model / function configured by the network device adopts a transform domain operation and / or the content of the transform domain operation adopted by the AI / ML model / function used by the network device, for example, the number of retained bases and / or coefficients, the index of the retained bases and / or coefficients, and the like.

[0074] In some embodiments, the method of the embodiments of the present application is applied to the model training phase.

[0075] In the above embodiments, the first information is related to the transform domain of the input of the AI / ML model / function expected by the terminal device, that is, the terminal device reports the relevant information of the transform domain of the input of the AI / ML model / function expected by the terminal device to the network device through the first information, for example, whether the input of the AI / ML model / function expected by the terminal device adopts a transform domain operation, and / or the content of the transform domain operation adopted by the input of the AI / ML model / function expected by the terminal device, and the like.

[0076] In the above embodiments, the first information can be sent through uplink RRC (Radio Resource Control) signaling or uplink RRC UAI (UE Assistance Information) signaling, and the present application is not limited thereto.

[0077] In the above embodiments, the terminal device can train an AI / ML model or a model supporting an AI / ML function according to the second information. For the convenience of description, it is collectively referred to as "training an AI / ML model / function".

[0078] In some possible implementations, the first information indicates an AI / ML model expected by the terminal device to support a transform domain input, and the second information includes relevant information of the AI / ML model supporting the transform domain input configured by the network device for the terminal device (for example, domain information (angle-time delay-Doppler) of the AI / ML input, the number of retained bases / coefficients, the index value of the retained bases / coefficients, and the like), and optionally, the specific scheme of the transform domain operation. The terminal device re-trains the AI / ML model supporting the transform domain input according to the above information, or directly uses the AI / ML model supporting the transform domain input.

[0079] For example, if the AI / ML model / function provided by the network device supports transform domain input, the terminal device can request the AI / ML model supporting transform domain input through the first information. At this time, the network device can deliver the AI / ML model supporting transform domain input to the terminal device, for example, deliver the structure and / or model parameters of the model to the terminal device, and optionally, also deliver the specific scheme of the transform domain to the terminal device. In this way, the terminal device can retrain the AI / ML model based on the delivered AI / ML model for inference / performance monitoring, or directly use the AI / ML model for inference / performance monitoring.

[0080] In some possible implementation manners, the first information indicates that the terminal device expects an AI / ML model supporting non-transform domain input, and the second information includes related information of the AI / ML model supporting non-transform domain input configured by the network device for the terminal device (for example, domain information (space, frequency, time) of AI / ML input, dimension size information, etc.). The terminal device can retrain the AI / ML model supporting non-transform domain input according to the related information, or directly use the AI / ML model supporting non-transform domain input.

[0081] For example, if the AI / ML model / function provided by the network device supports transform domain input, the terminal device can request a data set for training the AI / ML model supporting transform domain input through the first information. At this time, the network device can deliver the data set for training the AI / ML model supporting transform domain input to the terminal device, and optionally, also deliver the specific scheme of the transform domain, for example, the number of bases, the base index, the oversampling factor, etc., to the terminal device. In this way, the terminal device can train the AI / ML model based on the delivered data set for training to be used for inference / performance monitoring.

[0082] In some possible implementation manners, the first information indicates that the terminal device expects an AI / ML model supporting non-transform domain input, and the second information includes related information of the AI / ML model supporting non-transform domain input configured by the network device for the terminal device (for example, domain information (space, frequency, time) of AI / ML input, dimension size information, etc.). The terminal device can retrain the AI / ML model supporting non-transform domain input according to the related information, or directly use the AI / ML model supporting non-transform domain input.

[0083] In some possible implementation manners, the first information indicates that the terminal device expects an AI / ML model supporting non-transform domain input, and the second information includes related information of the AI / ML model supporting non-transform domain input configured by the network device for the terminal device (for example, domain information (space, frequency, time) of AI / ML input, dimension size information, etc.). The terminal device can retrain the AI / ML model supporting non-transform domain input according to the related information, or directly use the AI / ML model supporting non-transform domain input.

[0084] In the above embodiments, the AI / ML model supporting non-transform domain input means that the input of the AI / ML model does not support transform domain operation, or in other words, no transform domain operation is performed.

[0085] In the above embodiments, the network device can also indicate to the terminal device that the input of the AI / ML model configured by the network device for the terminal device is in the non-transform domain (i.e., no transform domain operation is performed).

[0086] For example, the second information includes indication information indicating that the input of the AI / ML model / function configured by the network device for the terminal device is in the non-transform domain. The application does not limit the specific indication method.

[0087] In the above embodiments, in some possible implementation manners, the terminal device trains the AI / ML model / function according to the second information, which can be pre-processing the CSI information of multiple time slots to obtain pre-processed multiple CSI information as the input of the AI / ML model / function, and the pre-processing includes angle-time-Doppler domain transformation.

[0088] For example, in the case where the terminal device and the network device confirm the transform domain operation (e.g., using the transform domain and / or specific transform domain scheme) of the input of the AI / ML model / function through the interaction of the first information and the second information, the terminal device can first pre-process (e.g., domain transform) the CSI information of multiple time slots according to the determined transform domain scheme, and then input the pre-processed CSI information (i.e., the CSI information after domain transform) as the input of the AI / ML model / function into the AI / ML model / function to compress and feed back the pre-processed multiple time slot CSI information, so that the input of the AI / ML model / function has undergone transform domain operation.

[0089] Taking the angle-time-Doppler domain transformation as an example.

[0090] In the above embodiments, the terminal device can transform the CSI information of multiple time slots from the space-frequency-time domain to the angle-time-Doppler domain to obtain transformed CSI information; and process the transformed CSI information according to the number of non-zero coefficients and / or basis to be retained to obtain coefficients on all or part of the transform domain basis as the input of the AI / ML model / function.

[0091] FIG. 4 is a schematic diagram of CSI compression feedback based on angle-time-Doppler domain.

[0092] As shown in FIG. 4, the terminal device first pre-processes the CSI information of multiple time slots, and then compresses and feeds back the CSI information of multiple time slots.

[0093] For example, at the terminal side, the terminal device performs angle-delay-Doppler domain transformation (three-dimensional Fourier transformation) on the precoding vector (CSI information) of the space-frequency-time domain and retains all or part of the coefficients on the transformed domain basis; correspondingly, at the network side, the network device performs post-processing on the result output by the AI / ML model to reconstruct the precoding vector, that is:

[0094] where i-j-k respectively represent the indexes corresponding to the angle-delay-Doppler domain basis, W' is the output of the AI / ML model, which is also the coefficient on the transformed domain basis, and Base(i,j,k) represents the Fourier basis corresponding to the i-j-k index.

[0095] In the above embodiment, the number of non-zero coefficients and / or bases that need to be retained can be determined based on L and Pv in the parameter combination (PC) configuration and the Q value; where L indicates the number of non-zero coefficients and / or bases retained in the angle domain, Pv indicates the ratio of the number of non-zero coefficients and / or bases retained in the delay domain to the total number of non-zero coefficients and / or bases, and the Q value indicates the number of non-zero coefficients and / or bases retained in the Doppler domain.

[0096] Table 1 below is an example of codebook parameter configuration, showing the case of codebook parameter combination of Release 18.

[0097] Table 1: Codebook parameter configurations for L, β and p υ

[0098] As can be seen from Table 1, when paramCombination-Doppler-r18 = 9 and rank = 2 (v in Table 1), for each layer of precoding vector V 32×13×4 (that is, 32 ports, 13 subbands, and 4 time slots) are all transformed into At this time, the input of the AI / ML model is transformed from the non-sparse space-frequency-time domain to the sparse angle-delay-Doppler domain, and the dimension is reduced from 32x13x4 to 6x4xQ. Thus, the transformation of the input of the AI / ML model / function is realized.

[0099] The above takes the codebook parameter combination of Release 18 as an example, but the present application is not limited thereto. In the embodiments of the present application, other codebook parameter combination configurations can also be used, such as a new combination of L and Pv.

[0100] In the above embodiments, the codebook parameter combination configuration and the Q value can be determined or indicated by at least one of the following manners:

[0101] determined in a predefined manner;

[0102] indicated by the network device through signaling;

[0103] determined by the terminal device according to its own channel state;

[0104] included in the model-related information exchanged between the network device and the terminal device.

[0105] For example, the codebook PC configuration and the Q value are predefined, and the terminal device and the network device have a consistent understanding thereof, so as to facilitate the implementation of pre-processing and post-processing.

[0106] For another example, the codebook PC configuration and the Q value are indicated by the network device to the terminal device through at least one of RRC signaling, MAC CE, and DCI.

[0107] For another example, the codebook PC configuration and the Q value are determined by the terminal device according to its own channel state and reported to the network side.

[0108] For another example, the codebook PC configuration and the Q value are included in the model-related information, and the network device and the terminal device reach a consistent understanding of the pre-processing and post-processing of the AI / ML model through the interaction of the model-related information.

[0109] In the above embodiments, the Q value can be indicated by additional information, and the application does not limit the specific indication manner. However, the Q value can also be included in the codebook parameter combination configuration, for example, as a column of the above table 1, so that L, Pv, and the Q value can be indicated by only one table, achieving more simplicity.

[0110] In the above embodiments, the index of the non-zero coefficient and / or the basis to be reserved can be determined or indicated by at least one of the following manners:

[0111] determined in a predefined manner;

[0112] indicated by the network device through signaling;

[0113] determined by the terminal device according to its own channel state;

[0114] included in the model-related information exchanged between the network device and the terminal device.

[0115] For example, the index is determined in a predefined manner, the interval of the reserved basis is centered on the zero frequency, and the index of the reserved basis can be determined in combination with the number of the reserved bases.

[0116] For another example, the index can be indicated by the network device to the terminal device through signaling. For example, the network device can indicate the terminal device through at least one of RRC signaling, MAC CE, and DCI, for example, in the form of a bitmap.

[0117] For another example, the index can be determined by the terminal device according to its own channel state and reported to the network device. For example, reported to the network device through UCI in the form of a bitmap.

[0118] For another example, the index can be included in the model-related information, and the network device and the terminal device obtain the index by interacting the model-related information, so as to reach an agreement on the pre-processing and post-processing of the AI / ML model.

[0119] In the above embodiment, in some possible implementation manners, the domain transformation processing described above can be performed by using a super-sampled Fourier transform basis. That is, the transformation from the space-frequency-time domain to the angle-delay-Doppler domain is performed by using a super-sampled Fourier transform basis.

[0120] For example, the terminal device performs angle-delay-Doppler domain transformation (that is, three-dimensional Fourier transform) on the precoding vector (CSI information) in the space-frequency-time domain by using a super-sampled Fourier transform basis. After the super-sampled Fourier transform, the coefficients in the transformed domain can be more sparse, and the information lost by retaining part of the basis coefficients is less, thereby reducing the complexity under the premise of ensuring performance.

[0121] In the above example, the space-frequency-time domain super-sampling factor corresponding to the Fourier transform basis can be configured by the network device (for example, configured through the second information described above), can be predefined, can be determined by the terminal device and reported to the network device, or can be indicated by the AI / ML model-related information interacted between the network device and the terminal device.

[0122] The above describes the pre-processing by taking the angle-delay-Doppler domain transformation as an example. In the embodiments of the present application, the pre-processing can also include other processing, such as normalization processing, phase correction processing, and the like. Through normalization processing, the features input to the AI / ML model can fall within a certain value range, so as to avoid that the difference in scale is too large and some features dominate; through phase correction processing, the correlation in the time-frequency domain can be improved, and the phase discontinuity caused by the transceiver can also be compensated, thereby improving the CSI compression rate.

[0123] The present application does not limit the order of different pre-processing. For example, normalization processing is performed first, and then domain transformation processing is performed; for another example, normalization processing is performed first, phase correction processing is performed, and then domain transformation processing is performed, and the like.

[0124] In some embodiments, the method of the embodiments of the present application is applied to a model inference stage.

[0125] In the above embodiments, the first information is related to a transform domain of an input of an AI / ML model / function supported by the terminal device, and after receiving the second information, the terminal device can deploy the AI / ML model / function according to the second information.

[0126] In some possible implementations, the first information indicates an AI / ML model supported by the terminal device that supports a transform domain input, and the second information includes related information of the AI / ML model that supports the transform domain input and is configured by the network device for the terminal device. Optionally, the second information can further include a specific scheme of a transform domain operation, and the terminal device can deploy the AI / ML model that supports the transform domain input according to the specific scheme.

[0127] In another possible implementation, the first information indicates an AI / ML model supported by the terminal device that does not support a transform domain input, and the second information includes an AI / ML model / function that does not support the transform domain input and is configured by the network device for the terminal device, and the terminal device can deploy the AI / ML model that does not support the transform domain input according to the second information.

[0128] In the above embodiments, the network device can further configure an AI / ML model with an input that is not a transform domain for the terminal device and indicate the same to the terminal device.

[0129] For example, the second information can further include indication information indicating that the input of the AI / ML model / function configured by the network device for the terminal device is not a transform domain. The present application is not limited thereto, and the indication information can also be independent of the second information and realized by other information.

[0130] In some possible implementations, the terminal device deploys the AI / ML model / function, which can be pre-processing of CSI information of multiple time slots to obtain pre-processed CSI information of multiple time slots as input of the deployed AI / ML model / function; the pre-processing includes angle-delay-Doppler domain transformation.

[0131] For example, the terminal device transforms the CSI information of the multiple time slots from a space-frequency-time domain to an angle-delay-Doppler domain to obtain transformed CSI information; and processes the transformed CSI information according to a number of non-zero coefficients to be retained and / or a number of bases to be retained to obtain coefficients on all or part of the transform domain bases as input of the deployed AI / ML model / function.

[0132] In the above embodiments, the determination manner of the number of non-zero coefficients to be retained and / or the number of bases to be retained and the determination manner of the index of the number of non-zero coefficients to be retained and / or the number of bases to be retained are the same as those in the training stage, which will not be described herein again.

[0133] In the above embodiments, similar to the training phase, the above domain transformation can also be performed using the oversampled Fourier transform basis, which will not be repeated here.

[0134] In the above embodiments, the CSI information of the plurality of time slots can include, but is not limited to, at least one of the following:

[0135] Only including the precoding vector of the historical measurement of the terminal device;

[0136] Including the precoding vector of the historical measurement of the terminal device and the future precoding vector predicted by the terminal device according to the precoding vector of the historical measurement;

[0137] Only including the future precoding vector predicted by the terminal device according to the precoding vector of the historical measurement.

[0138] In the above embodiments, as shown in FIG. 2, the terminal device can use the CSI prediction module to predict the CSI information of one or more time slots in the future by the CSI information obtained by measurement, and input the compressed feedback of the precoding vector of the input CSI information to the CSI compression module. For the precoding vector fed back It can only include the precoding vector of the historical measurement, or include the precoding vector of the historical measurement and the predicted future precoding vector, or only include the predicted future precoding vector.

[0139] In the above embodiments, the precoding vector of the historical measurement can be one or more time slots of the precoding vector, and the predicted future precoding vector can be one or more time slots of the precoding vector. In addition, the terminal device can predict the future precoding vector based on the precoding vector of the historical measurement based on the non-AI / ML method, or predict the future precoding vector based on the precoding vector of the historical measurement based on the AI / ML method.

[0140] In the above embodiments, the non-AI / ML method is, for example, autoregressive (AR, Autoregressive Model), Weiner filter, etc.

[0141] In the above embodiments, for the CSI compression feedback, the network device can also pass the specific scheme of the quantizer to the terminal device, for example, together with the specific scheme of the transform domain. Wherein, the specific scheme of the quantizer can include, but is not limited to, at least one of the following:

[0142] Quantizer type (vector / scalar);

[0143] Quantization bit number;

[0144] Vector length of vector quantization;

[0145] Vector / scalar quantization table.

[0146] Since the output of the AI / ML model is usually a floating point value (for example, Float32 / 64), it is too costly to directly transmit the floating point value over the air. According to the above embodiment, for CSI compression feedback, an additional quantizer is added at the output position of the terminal side AI / ML model, and an additional dequantizer is added at the input position of the network side AI / ML model, which solves the above problem.

[0147] The above embodiments only exemplarily illustrate the embodiments of the present application, but the present application is not limited thereto, and can be appropriately modified on the basis of the above embodiments. For example, the above embodiments can be used alone, or one or more of the above embodiments can be combined.

[0148] According to the embodiments of the present application, the terminal device and the network device interact the related information of the transform domain of the input of the AI / ML model / function, such as whether the input of the AI / ML model / function uses the transform domain, the specific transform domain scheme, and the like, which facilitates the training, inference and performance monitoring of the AI / ML model / function, reduces the complexity, avoids performance loss, and improves the performance of the communication system.

[0149] Embodiments of the second aspect

[0150] The embodiments of the present application provide an AI / ML model / function configuration method, which is applied to a network device and corresponds to the method of the embodiments of the first aspect. The same content as the embodiments of the first aspect will not be described again.

[0151] FIG. 5 is a schematic diagram of the AI / ML model / function configuration method according to the embodiments of the present application. As shown in FIG. 5, the method includes:

[0152] 510: The network device receives first information sent by the terminal device, and the first information is related to the transform domain of the input of the AI / ML model / function expected or supported by the terminal device; and

[0153] 520: The network device sends second information to the terminal device, and the second information is used to indicate the transform domain operation of the input of the AI / ML model / function.

[0154] In some embodiments, the second information indicates whether the AI / ML model / function configured by the network device uses the transform domain operation, and / or the content (i.e., the transform domain scheme) of the transform domain operation used by the AI / ML model / function used by the network device.

[0155] In some embodiments, the first information is related to a transform domain of an input of an AI / ML model / function expected by the terminal device, and the terminal device trains the AI / ML model or a model supporting the AI / ML function according to the second information after receiving the second information. For the convenience of description, the "training of the AI / ML model or the model supporting the AI / ML function" is collectively referred to as "training of the AI / ML model / function".

[0156] In the above embodiments, the first information is related to a transform domain of an input of an AI / ML model / function expected by the terminal device, including: the first information indicating whether a transform domain operation is adopted for an input of the AI / ML model expected by the terminal device, and / or content (i.e., a transform domain scheme) of a transform domain operation adopted by the AI / ML model / function expected by the terminal device.

[0157] In some possible implementations, the first information indicates an AI / ML model expected by the terminal device to support a transform domain input, the second information includes related information of the AI / ML model supporting the transform domain input configured by the network device for the terminal device, and optionally, includes specific content of the transform domain operation; the terminal device re-trains the AI / ML model supporting the transform domain input according to the related information, or directly uses the AI / ML model supporting the transform domain input.

[0158] In another possible implementation, the first information indicates a data set expected by the terminal device to train an AI / ML model supporting a transform domain input; the second information includes a data set configured by the network device for the terminal device to train the AI / ML model supporting the transform domain input, and optionally, includes specific content of the transform domain operation; the terminal device re-trains the AI / ML model supporting the transform domain input according to the data set.

[0159] In yet another possible implementation, the first information indicates an AI / ML model expected by the terminal device to support a non-transform domain input; the second information includes related information of the AI / ML model supporting the non-transform domain input configured by the network device for the terminal device; the terminal device re-trains the AI / ML model supporting the non-transform domain input according to the related information, or directly uses the AI / ML model supporting the non-transform domain input.

[0160] Optionally, the second information includes indication information indicating that an input of the AI / ML model / function configured by the network device for the terminal device is a non-transform domain.

[0161] In some possible implementation manners, the first information indicates that the terminal device expects a dataset for training an AI / ML model supporting non-transform domain input; the second information includes a dataset for training the AI / ML model supporting non-transform domain input configured by the network device for the terminal device; and the terminal device re-trains the AI / ML model supporting non-transform domain input according to the dataset.

[0162] Optionally, the second information includes indication information indicating that the input of the AI / ML model / function configured by the network device for the terminal device is in a non-transform domain.

[0163] In the above embodiment, the training of the AI / ML model / function by the terminal device according to the second information can include: pre-processing CSI information of multiple time slots to obtain pre-processed CSI information of the multiple time slots as input of the AI / ML model / function, and the pre-processing includes angle-time-Doppler domain transformation. That is, the terminal device can perform angle-time-Doppler domain transformation on the CSI of the multiple time slots. The content of the angle-time-Doppler domain transformation has been described in the embodiment of the first aspect, and will not be repeated here.

[0164] In the above embodiment, the first information can be sent through uplink RRC signaling or through uplink RRC UAI signaling.

[0165] In some embodiments, the first information is related to a transform domain of input of an AI / ML model / function supported by the terminal device, and the terminal device deploys the AI / ML model / function according to the second information after receiving the second information.

[0166] In some possible implementation manners, the deployment of the AI / ML model / function by the terminal device according to the second information includes: pre-processing CSI information of multiple time slots to obtain pre-processed CSI information of the multiple time slots as input of the deployed AI / ML model / function; and the pre-processing includes angle-time-Doppler domain transformation. That is, the terminal device can perform angle-time-Doppler domain transformation on the CSI information of the multiple time slots. The content of the angle-time-Doppler domain transformation has been described in the embodiment of the first aspect, and will not be repeated here.

[0167] In another possible implementation manner, the first information indicates that the terminal device supports an AI / ML model supporting transform domain input; the second information includes related information of the AI / ML model supporting transform domain input configured by the network device for the terminal device; and the terminal device deploys the AI / ML model supporting transform domain input according to the related information.

[0168] Optionally, the second information can further include a specific scheme of transform domain operation.

[0169] In some possible implementation manners, the first information indicates an AI / ML model of non-transform domain input supported by the terminal device; the second information includes an AI / ML model / function of non-transform domain input configured by the network device for the terminal device; and the terminal device deploys the AI / ML model / function of non-transform domain input according to the second information.

[0170] Optionally, the second information can include indication information indicating that the input of the AI / ML model / function configured by the network device for the terminal device is of non-transform domain.

[0171] The above embodiments are only exemplary for the embodiments of the present application, but the present application is not limited thereto, and can be appropriately modified on the basis of the above embodiments. For example, the above embodiments can be used alone or in combination of one or more of the above embodiments.

[0172] According to the embodiments of the present application, the terminal device and the network device interact the information related to the transform domain of the input of the AI / ML model / function, for example, whether the input of the AI / ML model / function adopts the transform domain, the specific transform domain scheme, and the like, so as to facilitate the training, inference and performance monitoring of the AI / ML model / function, reduce the complexity, avoid the performance loss, and improve the performance of the communication system.

[0173] Embodiments of the third aspect

[0174] The embodiments of the present application provide an AI / ML model / function configuration apparatus. The apparatus can be a terminal device, or can be one or more components or assemblies configured in the terminal device, which corresponds to the method applied to the terminal device side in the embodiments of the first aspect, and the same content as the embodiments of the first aspect will not be described herein.

[0175] FIG. 6 is a schematic diagram of an AI / ML model / function configuration apparatus according to an embodiment of the present application. As shown in FIG. 6, the AI / ML model / function configuration apparatus 600 according to an embodiment of the present application includes a sending unit 610 and a receiving unit 620.

[0176] The sending unit 610 sends first information to a network device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device; and the receiving unit 620 receives second information sent by the network device, the second information being used to indicate a transform domain operation of the input of the AI / ML model / function.

[0177] In some embodiments, the second information indicates whether the AI / ML model / function configured by the network device adopts the transform domain operation and / or the content of the transform domain operation adopted by the AI / ML model / function used by the network device.

[0178] In some embodiments, the first information relates to a transform domain of an input of an AI / ML model / function expected by the terminal device, i.e., the terminal device reports, to the network device through the first information, information related to a transform domain of an input of an AI / ML model / function expected by the terminal device, for example, whether the input of the AI / ML model / function expected by the terminal device adopts a transform domain operation, and / or content of the transform domain operation adopted by the input of the AI / ML model / function expected by the terminal device, etc.

[0179] In the above embodiments, the first information can be sent through uplink RRC signaling or uplink RRC UAI signaling.

[0180] In some embodiments, as shown in FIG. 6, the apparatus 600 further includes:

[0181] The processing unit 630 trains an AI / ML model or a model supporting an AI / ML function according to the second information after the receiving unit 620 receives the second information. For the convenience of description, it is collectively referred to as “training an AI / ML model / function”.

[0182] In some possible implementations, the first information indicates an AI / ML model expected by the terminal device to support a transform domain input, and the second information includes information related to an AI / ML model configured by the network device for the terminal device to support the transform domain input, and optionally, a specific scheme of the transform domain operation. The processing unit 630 re-trains the AI / ML model supporting the transform domain input according to the specific scheme of the transform domain operation, or directly uses the AI / ML model supporting the transform domain input.

[0183] In another possible implementation, the first information indicates a data set expected by the terminal device to train an AI / ML model supporting a transform domain input, and the second information includes a data set configured by the network device for the terminal device to train the AI / ML model supporting the transform domain input, and optionally, a specific scheme of the transform domain operation. The processing unit 630 re-trains the AI / ML model supporting the transform domain operation according to the data set.

[0184] In yet another possible implementation, the first information indicates an AI / ML model expected by the terminal device to support a non-transform domain input; the second information includes information related to an AI / ML model configured by the network device for the terminal device to support the non-transform domain input; and the processing unit 630 re-trains the AI / ML model supporting the non-transform domain input according to the information, or directly uses the AI / ML model supporting the non-transform domain input.

[0185] In some possible implementation manners, the first information indicates a dataset of the AI / ML model expected by the terminal device to support non-transform domain input; the second information includes a dataset of the AI / ML model configured by the network device for the terminal device to support non-transform domain input; and the processing unit 630 can retrain the AI / ML model supporting non-transform domain input according to the dataset.

[0186] In the above embodiments, the AI / ML model supporting non-transform domain input refers to that the input of the AI / ML model does not support transform domain operation, or in other words, no transform domain operation is performed.

[0187] In the above embodiments, the network device can also indicate the terminal device that the input of the AI / ML model configured by the network device for the terminal device is non-transform domain (that is, no transform domain operation is performed).

[0188] For example, the second information includes indication information indicating that the input of the AI / ML model / function configured by the network device for the terminal device is non-transform domain.

[0189] In the above embodiments, in some possible implementation manners, the processing unit 630 can train the AI / ML model / function according to the second information, which can include pre-processing the CSI information of the plurality of time slots to obtain pre-processed CSI information of the plurality of time slots as the input of the AI / ML model / function, and the pre-processing includes angle-delay-Doppler domain transform.

[0190] In the above embodiments, the processing unit 630 can transform the CSI information of the plurality of time slots from space-frequency-time domain to angle-delay-Doppler domain to obtain transformed CSI information; and process the transformed CSI information according to the number of non-zero coefficients and / or bases to be reserved to obtain coefficients on all or part of the transform domain bases as the input of the AI / ML model / function.

[0191] In the above embodiments, the number of non-zero coefficients and / or bases to be reserved can be determined based on L and Pv in the parameter combination (PC) configuration and a Q value; where L indicates the number of non-zero coefficients and / or bases reserved in the angle domain, Pv indicates the ratio of the number of non-zero coefficients and / or bases reserved in the delay domain to the total number of non-zero coefficients and / or bases, and the Q value indicates the number of non-zero coefficients and / or bases reserved in the Doppler domain.

[0192] In the above embodiments, the parameter combination configuration and the Q value can be determined or indicated by at least one of the following manners:

[0193] determined in a predefined manner;

[0194] indicated by the network device through signaling;

[0195] determined by the terminal device according to its own channel state;

[0196] included in the model-related information exchanged between the network device and the terminal device.

[0197] In the above embodiments, the Q value can be indicated by additional information or included in the codebook parameter combination configuration.

[0198] In the above embodiments, the indices of the non-zero coefficients and / or bases that need to be reserved can be determined or indicated in at least one of the following ways:

[0199] determined in a predefined manner;

[0200] indicated by the network device through signaling;

[0201] determined by the terminal device according to its own channel state;

[0202] included in the model-related information exchanged between the network device and the terminal device.

[0203] In the above embodiments, in some possible implementation manners, the processing unit 630 can perform the above domain transformation processing using a super-sampled Fourier transform basis. That is, the transformation from the space-frequency-time domain to the angle-delay-Doppler domain uses a super-sampled Fourier transform basis.

[0204] In the above examples, the space-frequency-time domain super-sampling factor corresponding to the Fourier transform basis can be configured by the network device (for example, configured through the above-mentioned second information), can be predefined, can be determined by the terminal device and reported to the network device, or can be indicated by the AI / ML model-related information exchanged between the network device and the terminal device.

[0205] In some embodiments, the preprocessing can also include other processing, such as normalization processing, phase correction processing, etc.

[0206] In some other embodiments, the first information is related to the transform domain of the input of the AI / ML model / function supported by the terminal device, and after receiving the above-mentioned second information, the processing unit 630 can deploy the AI / ML model / function according to the second information.

[0207] In some possible implementation manners, the first information indicates an AI / ML model supported by the terminal device and supporting transform domain input, the second information includes related information of the AI / ML model supporting transform domain input configured by the network device for the terminal device, and optionally, the second information can further include a specific scheme of the transform domain operation, and the processing unit 630 can deploy the AI / ML model supporting transform domain input according to the second information.

[0208] In some possible implementation manners, the first information indicates an AI / ML model supported by the terminal device and supporting transform domain input, the second information includes related information of the AI / ML model supporting transform domain input configured by the network device for the terminal device, and optionally, the second information can further include a specific scheme of the transform domain operation, and the processing unit 630 can deploy the AI / ML model supporting transform domain input according to the second information.

[0209] In the above embodiment, the network device can further configure the terminal device with an AI / ML model whose input is a non-transform domain.

[0210] For example, the second information can further include indication information indicating that the input of the AI / ML model / function configured by the network device for the terminal device is a non-transform domain.

[0211] In some possible implementation manners, the processing unit 630 deploys the AI / ML model / function, which can be, pre-processes the CSI information of the multiple time slots to obtain pre-processed multiple CSI information as the input of the deployed AI / ML model / function; the pre-processing includes angle-delay-Doppler domain transformation.

[0212] For example, the terminal device performs transformation from the space-frequency-time domain to the angle-delay-Doppler domain on the CSI information of the multiple time slots to obtain transformed CSI information; and performs processing on the transformed CSI information according to the number of non-zero coefficients and / or bases to be reserved, to obtain coefficients on all or part of the transformed domain bases as the input of the deployed AI / ML model / function.

[0213] In the above embodiment, the determination manner of the number of non-zero coefficients and / or bases to be reserved and the determination manner of the index of the number of non-zero coefficients and / or bases to be reserved are the same as those in the training phase, which will not be described herein again.

[0214] In the above embodiment, similar to the training phase, the processing unit 630 can also use the oversampled Fourier transform basis to perform the above domain transformation, which will not be described herein again.

[0215] In the above embodiment, the CSI information of the multiple time slots can include but is not limited to at least one of the following:

[0216] Only the precoding vector of the historical measurement of the terminal device;

[0217] a precoding vector comprising historical measurements of the terminal device and a predicted future precoding vector of the terminal device from the precoding vector of the historical measurements;

[0218] a predicted future precoding vector of the terminal device from the precoding vector of the historical measurements.

[0219] In the above embodiments, the precoding vector of the historical measurements can be a precoding vector of one or more time slots, and the predicted future precoding vector can be a precoding vector of one or more time slots. In addition, the processing unit 630 can predict the future precoding vector from the precoding vector of the historical measurements based on a non-AI / ML method, or predict the future precoding vector from the precoding vector of the historical measurements based on an AI / ML method.

[0220] The embodiments of the present application provide an apparatus for configuring an AI / ML model / function. The apparatus can be a network device, or can be a component or assembly configured in the network device, which corresponds to the method applied to the network device side in the embodiments of the second aspect, and the same content as the embodiments of the second aspect will not be repeated.

[0221] FIG. 7 is a schematic diagram of an apparatus for configuring an AI / ML model / function according to an embodiment of the present application. As shown in FIG. 7, the apparatus 700 for configuring an AI / ML model / function according to an embodiment of the present application comprises a receiving unit 710 and a sending unit 720.

[0222] The receiving unit 710 receives first information sent by a terminal device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device; and the sending unit 720 sends second information to the terminal device, the second information being used to indicate a transform domain operation of the input of the AI / ML model / function.

[0223] The related content of the first information and the second information has been described in the embodiments of the first aspect, which is incorporated herein, and will not be repeated here.

[0224] The above embodiments are only exemplary, but the present application is not limited thereto, and appropriate modifications can be made on the basis of the above embodiments. For example, the above embodiments can be used alone, or one or more of the above embodiments can be combined.

[0225] It should be noted that the above only describes the components or modules related to the present application, but the present application is not limited thereto. The apparatuses 600 and 700 can also include other components or modules, and the specific content of these components or modules can be referred to the related technology.

[0226] Further, for simplicity, only the connection relationship or signal direction between each component or module is exemplarily shown in FIG. 6 and FIG. 7, but it should be clear to those skilled in the art that various related technologies such as bus connection can be adopted. Each component or module described above can be implemented by hardware facilities such as processor, memory, transmitter, receiver, etc.; the implementation of the present application is not limited thereto.

[0227] According to the embodiments of the present application, the terminal device and the network device interact with the related information of the transform domain of the input of the AI / ML model / function, such as whether the input of the AI / ML model / function adopts the transform domain, the specific transform domain scheme, and the like, so as to facilitate the training, inference and performance monitoring of the AI / ML model / function, reduce the complexity, avoid the performance loss, and improve the performance of the communication system.

[0228] Embodiments of the fourth aspect

[0229] The embodiments of the present application provide a communication system, including a terminal device and a network device.

[0230] For example, the structure of the communication system can refer to FIG. 1. As shown in FIG. 1, the communication system 100 includes a network device 101 and terminal devices 102 and 103.

[0231] In some embodiments, the network device 101 performs the functions of the network device in the embodiments of the second aspect, and correspondingly, the terminal devices 102 and 103 perform the functions of the terminal device in the embodiments of the first aspect. Since the functions of the network device and the terminal device have been described in the embodiments of the first aspect to the second aspect, the contents are incorporated herein, and will not be repeated here.

[0232] The embodiments of the present application also provide a network device.

[0233] FIG. 8 is a schematic block diagram of the system structure of the network device according to the embodiments of the present application. As shown in FIG. 8, the network device 800 can include a processor 810 and a memory 820; the memory 820 is coupled to the processor 810. The memory 820 can store various data; in addition, it also stores a program 830 for information processing, and executes the program 830 under the control of the processor 810.

[0234] In one embodiment, the network device 800 is configured to implement the method described in the embodiments of the second aspect, for example, includes the functions of the apparatus 700 in the embodiments of the third aspect, which can be integrated into the processor 810, or can be configured separately from the processor 810, for example, the apparatus 700 can be configured as a chip connected with the processor 810, and the functions of the above apparatus are realized through the control of the processor 810.

[0235] In addition, as shown in FIG. 8, the network device 800 can further include a transceiver 840 and an antenna 850, etc. The functions of the above components are similar to those of the prior art, and thus are not described here. It is worth noting that the network device 800 does not necessarily include all the components shown in FIG. 8. In addition, the network device 800 can also include components not shown in FIG. 8, which can be referred to the prior art.

[0236] The embodiments of the present application also provide a terminal device.

[0237] FIG. 9 is a schematic block diagram of a system structure of a terminal device according to an embodiment of the present application. As shown in FIG. 9, the terminal device 900 can include a processor 910 and a memory 920; the memory 920 is coupled to the processor 910. It is worth noting that this figure is exemplary; other types of structures can also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0238] In one embodiment, the terminal device 900 is configured to implement the method described in the embodiments of the first aspect, including the functions of the apparatus 600 in the embodiments of the third aspect, which can be integrated into the processor 910 or configured separately from the processor 910, for example, the apparatus 600 can be configured as a chip connected to the processor 910 to implement the functions of the above apparatus through the control of the processor 910.

[0239] As shown in FIG. 9, the terminal device 900 can further include a communication module 930, an input unit 940, a display 950, and a power supply 960. It is worth noting that the terminal device 900 does not necessarily include all the components shown in FIG. 9. In addition, the terminal device 900 can also include components not shown in FIG. 9, which can be referred to the related art.

[0240] As shown in FIG. 9, the processor 910, also known as a controller or operation control, can include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of each component of the terminal device 900.

[0241] The memory 920, for example, can be one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, or other suitable device. Various data can be stored, and programs for performing related information can also be stored. The processor 910 can execute the programs stored in the memory 920 to achieve information storage or processing, etc. The functions of other components are similar to those of the prior art, and thus are not described here. The components of the terminal device 900 can be implemented by dedicated hardware, firmware, software, or a combination thereof, without departing from the scope of the present application.

[0242] The embodiments of the present application further provide a computer readable program, wherein the program enables the computer to execute the method of the embodiments of the second aspect when the program is executed in the network device.

[0243] The embodiments of the present application further provide a storage medium storing a computer readable program, wherein the computer readable program enables the computer to execute the method of the embodiments of the second aspect when the program is executed in the network device.

[0244] The embodiments of the present application further provide a computer readable program, wherein the program enables the computer to execute the method of the embodiments of the first aspect when the program is executed in the terminal device.

[0245] The embodiments of the present application further provide a storage medium storing a computer readable program, wherein the computer readable program enables the computer to execute the method of the embodiments of the first aspect when the program is executed in the terminal device.

[0246] The embodiments of the present application further provide a computer program product comprising a computer program, wherein the computer program is executed by a processor to implement the method of the embodiments of the first aspect or the second aspect.

[0247] The apparatus and method described in the above embodiments of the present application can be realized by hardware, or by hardware in combination with software. The present application relates to a computer readable program, which, when executed by a logic component, enables the logic component to realize the apparatus or constituent components described above, or enables the logic component to realize the various methods or steps described above. The present application also relates to a storage medium for storing the above program, such as a hard disk, a magnetic disk, an optical disk, a DVD, a flash memory, etc.

[0248] The method / apparatus described in combination with the embodiments of the present application can be directly embodied as hardware, a software module executed by a processor, or a combination of the two. For example, one or more of the functional block diagrams shown in the figures and / or a combination of one or more of the functional block diagrams can correspond to each software module of the computer program flow, or can correspond to each hardware module. The software modules can respectively correspond to each step shown in the figures. These hardware modules can be realized by, for example, fixing the software modules with a field programmable gate array (FPGA).

[0249] The software modules can reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, registers, a hard disk, a mobile disk, CD-ROM, or any other form of storage medium known in the art. One storage medium can be coupled to the processor, such that the processor can read information from, and write information to, the storage medium; or the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The software modules can be stored in a memory of the mobile terminal, or in a memory card that can be inserted into the mobile terminal. For example, if the device, such as the mobile terminal, uses a MEGA-SIM card or a flash memory device of large capacity, the software modules can be stored in the MEGA-SIM card or the flash memory device of large capacity.

[0250] One or more of the functional blocks depicted in the figures and / or one or more combinations of the functional blocks can be implemented as a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any appropriate combination thereof, for performing the functions described in this disclosure. One or more of the functional blocks described in relation to Figure 6 or Figure 7 and / or one or more combinations of the functional blocks can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0251] The present application has been described above with the attachment to the specific embodiments, but it should be clear to those skilled in the art that these descriptions are exemplary and are not a limitation on the scope of protection of the present application. Those skilled in the art can make various modifications and changes to the present application according to the spirit and principles of the present application, and these modifications and changes are also within the scope of the present application.

[0252] In addition to the above embodiments disclosed in the present embodiment, the following notes are also disclosed:

[0253] 1. A terminal device comprising a memory and a processor, the memory storing a computer program, the processor configured to execute the computer program to implement a method comprising:

[0254] sending first information to a network device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device;

[0255] receiving second information sent by the network device, the second information being used to indicate a transform domain operation of an input of an AI / ML model / function.

[0256] 2. A computer readable program, wherein the program, when executed in a terminal device, causes a computer to perform the method as follows in the terminal device:

[0257] sending first information to a network device, the first information being related to a transform domain of input of an AI / ML model / function expected or supported by the terminal device;

[0258] receiving second information sent by the network device, the second information being used to indicate a transform domain operation of input of the AI / ML model / function.

[0259] 3. A storage medium storing a computer readable program, wherein the computer readable program causes a computer to perform the method as follows in a terminal device:

[0260] sending first information to a network device, the first information being related to a transform domain of input of an AI / ML model / function expected or supported by the terminal device;

[0261] receiving second information sent by the network device, the second information being used to indicate a transform domain operation of input of the AI / ML model / function.

[0262] 4. A network device comprising a memory and a processor, the memory storing a computer program, and the processor being configured to execute the computer program to implement the method as follows:

[0263] receiving first information sent by a terminal device, the first information being related to a transform domain of input of an AI / ML model / function expected or supported by the terminal device;

[0264] sending second information to the terminal device, the second information being used to indicate a transform domain operation of input of the AI / ML model / function.

[0265] 5. A computer readable program, wherein the program, when executed in a network device, causes a computer to perform the method as follows in the network device:

[0266] receiving first information sent by a terminal device, the first information being related to a transform domain of input of an AI / ML model / function expected or supported by the terminal device;

[0267] sending second information to the terminal device, the second information being used to indicate a transform domain operation of input of the AI / ML model / function.

[0268] 6. A storage medium storing a computer readable program, wherein the computer readable program causes a computer to perform the method as follows in a network device:

[0269] Receive first information sent by the terminal device, wherein the first information is related to the transform domain of the input of the AI / ML model / function that the terminal device expects or supports;

[0270] Send a second message to the terminal device, the second message being used to instruct the transform domain operation of the input of the AI / ML model / function.

[0271] 7. A computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the following method:

[0272] Send a first message to the network device, the first message being related to the transform domain of the input of an AI / ML model / function that the terminal device expects or supports; receive a second message sent by the network device, the second message being used to indicate the transform domain operation of the input of the AI / ML model / function;

[0273] or,

[0274] The system receives first information sent by a terminal device, the first information being related to the transform domain of the input of an AI / ML model / function that the terminal device expects or supports; and sends second information to the terminal device, the second information being used to indicate the transform domain operation of the input of the AI / ML model / function.

[0275] 8. A communication system, comprising a first network device and a terminal device, wherein,

[0276] The network device includes a memory and a processor, the memory storing a computer program, and the processor being configured to execute the computer program to implement the following method:

[0277] Receive first information sent by the terminal device, the first information being related to the transform domain of the input of the AI / ML model / function that the terminal device expects or supports; and send...

[0278] The terminal device sends a second message, which is used to indicate the transform domain operation of the input of the AI / ML model / function;

[0279] The terminal device includes a memory and a processor. The memory stores a computer program, and the processor is configured to execute the computer program to implement the following method:

[0280] Send first information to the network device, the first information being related to the transform domain of the input of the AI / ML model / function that the terminal device expects or supports;

[0281] Receive second information sent by the network device, the second information being used to indicate the transform domain operation of the input of the AI / ML model / function.

Claims

1. An apparatus for configuring an AI (Artificial Intelligence) / ML (Machine Learning) model / function, configured in a terminal device, wherein, The apparatus comprises: a sending unit that sends first information to a network device, the first information being related to a transform domain of an input of an AI / ML model / function expected or supported by the terminal device; a receiving unit that receives second information sent by the network device, the second information being used to indicate a transform domain operation of an input of an AI / ML model / function.

2. The apparatus according to claim 1, wherein the second information indicates whether the AI / ML model / function configured by the network device adopts a transform domain operation, and / or the content of the transform domain operation adopted by the AI / ML model / function used by the network device.

3. The apparatus of claim 1, wherein, The apparatus further comprises a processing unit; the first information is related to a transform domain of an input of an AI / ML model / function expected by the terminal device, and the processing unit trains an AI / ML model according to the second information or trains a model supporting an AI / ML function according to the second information.

4. The apparatus according to claim 3, wherein the first information is related to a transform domain of an input of an AI / ML model / function expected by the terminal device, and the processing unit trains an AI / ML model according to the second information or trains a model supporting an AI / ML function according to the second information.

5. The apparatus according to claim 3, wherein the first information indicates an AI / ML model supporting a transform domain input expected by the terminal device; the second information comprises related information of an AI / ML model supporting a transform domain input configured by the network device for the terminal device; the processing unit re-trains an AI / ML model supporting a transform domain input according to the related information of the AI / ML model supporting a transform domain input configured by the network device for the terminal device, or directly uses the AI / ML model supporting a transform domain input.

6. The apparatus according to claim 3, wherein the first information indicates a data set for training an AI / ML model supporting a transform domain input expected by the terminal device; the second information comprises a data set for training an AI / ML model supporting a transform domain input configured by the network device for the terminal device; the processing unit re-trains an AI / ML model supporting a transform domain input according to the data set.

7. The apparatus according to claim 3, wherein the first information indicates an AI / ML model supporting a non-transform domain input expected by the terminal device; the second information comprises related information of an AI / ML model supporting a non-transform domain input configured by the network device for the terminal device; the processing unit re-trains an AI / ML model supporting a non-transform domain input according to the related information of the AI / ML model supporting a non-transform domain input configured by the network device for the terminal device, or directly uses the AI / ML model supporting a non-transform domain input.

8. The apparatus according to claim 3, wherein ​ The first information indicates a dataset of an AI / ML model that supports non-transform domain input expected by the terminal device; The second information includes a dataset of an AI / ML model that supports non-transform domain input configured by the network device for the terminal device; The processing unit re-trains the AI / ML model that supports non-transform domain input according to the dataset.

9. The apparatus of claim 3, wherein, The processing unit trains the AI / ML model or the model supporting AI / ML function according to the second information, including: The CSI (Channel State Information) information of multiple time slots is preprocessed to obtain preprocessed multiple CSI information as input of the AI / ML model or the model supporting AI / ML function; wherein the preprocessing includes angle-delay-Doppler domain transformation.

10. The apparatus of claim 9, wherein, The processing unit transforms the CSI information of the multiple time slots from space-frequency-time domain to angle-delay-Doppler domain to obtain transformed CSI information; The transformed CSI information is processed according to the number of non-zero coefficients and / or bases to be retained to retain coefficients on all or part of the transformed domain bases as input of the AI / ML model or the model supporting AI / ML function.

11. The apparatus of claim 10, wherein, The number of non-zero coefficients and / or bases to be retained is determined based on L and Pv and Q values in a codebook parameter combination configuration; The L indicates the number of angle domain non-zero coefficients and / or bases, the Pv indicates the ratio of the number of time delay domain non-zero coefficients and / or bases to the total number of non-zero coefficients and / or bases, and the Q value indicates the number of Doppler domain non-zero coefficients and / or bases.

12. The apparatus of claim 11, wherein, The codebook parameter combination configuration and the Q value are determined or indicated by at least one of the following ways: Determined by a predefined way; Indicated by the network device through signaling; Determined by the terminal device according to its own channel state; Included in the model related information exchanged between the network device and the terminal device.

13. The apparatus of claim 10, wherein, The index of the non-zero coefficients and / or bases to be retained is determined or indicated by at least one of the following ways: Determined by a predefined way; Indicated by the network device through signaling; Determined by the terminal device according to its own channel state; Included in the model related information exchanged between the network device and the terminal device.

14. The apparatus of claim 1, wherein, The apparatus further includes a processing unit, The first information is related to the transform domain of the input of the AI / ML model / function supported by the terminal device; The processing unit deploys the AI / ML model / function according to the second information.

15. The apparatus of claim 14, wherein, The processing unit deploys the AI / ML model / function according to the second information, including: The CSI information of multiple time slots is preprocessed to obtain preprocessed multiple CSI information as input of the deployed AI / ML model / function; wherein the preprocessing includes angle-delay-Doppler domain transformation.

16. The apparatus of claim 15, wherein, The processing unit transforms the CSI information of the plurality of time slots from a space-frequency-time domain to an angle-delay-Doppler domain to obtain transformed CSI information; According to the number of non-zero coefficients and / or bases to be retained, the transformed CSI information is processed to retain all or part of the coefficients on the transformed domain bases as inputs of the deployed AI / ML model or the AI / ML function supported model. The CSI information of the plurality of time slots includes at least one of:

17. The apparatus of claim 15, wherein, Only the precoding vectors of the historical measurements of the terminal device; Both the precoding vectors of the historical measurements of the terminal device and the future precoding vectors predicted by the terminal device based on the precoding vectors of the historical measurements; Only the future precoding vectors predicted by the terminal device based on the precoding vectors of the historical measurements.

18. The apparatus of claim 14, wherein The first information indicates an AI / ML model supported by the terminal device that supports transformed domain input; The second information includes related information of an AI / ML model configured by the network device for the terminal device that supports transformed domain input; The processing unit deploys an AI / ML model that supports transformed domain input according to the related information of the AI / ML model configured by the network device for the terminal device that supports transformed domain input.

19. The apparatus of claim 14, wherein The first information indicates an AI / ML model supported by the terminal device that does not support transformed domain input; The second information includes an AI / ML model / function configured by the network device for the terminal device that does not support transformed domain input; The processing unit deploys an AI / ML model / function that does not support transformed domain input according to the second information. The apparatus includes:

20. An apparatus for configuring an AI / ML model / functionality configured in a network device, wherein, A receiving unit that receives first information sent by a terminal device, the first information being related to a transformation domain of an input of an AI / ML model / function expected or supported by the terminal device; A sending unit that sends second information to the terminal device, the second information being used to indicate a transformation domain operation of an input of an AI / ML model / function. ​

Images