Model training method and apparatus, model inference method and apparatus, and node, system and medium

Abstract

The embodiments of the present disclosure relate to a model training method and apparatus, a model inference method and apparatus, and a node, a system and a medium. The model training method may be executed by means of a first node. The method comprises: obtaining a plurality of groups of sample data obtained on the basis of different modulation modes, wherein each group of sample data comprises first data to be demodulated and first demodulated data, the first demodulated data corresponds to the first data to be demodulated, and the first demodulated data in different groups of sample data has the same data length; and on the basis of the first data to be demodulated and the first demodulated data, training a first model, wherein the first model is used for data demodulation. In this way, in the embodiments of the present disclosure, the difference between different modulation modes can be ignored, so that it is not necessary to train corresponding demodulation models for different modulation modes, thereby reducing training overheads and storage overheads of the demodulation models.

Description

Model training method, model inference method, device, node, system and medium TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of communication, and particularly relates to a model training method, a model inference method, a device, a node, a system and a medium. BACKGROUND

[0002] Signal demodulation is one of the basic modules of a wireless communication system for realizing low error rate high-speed transmission, and affects the overall performance of the communication system. Under different modulation modes, the mapping between the to-be-demodulated data and the demodulated data is different, and one to-be-demodulated data symbol can correspond to different data lengths. Therefore, for to-be-demodulated data using different modulation modes, a matched demodulation mode needs to be used for demodulation.

[0003] SUMMARY

[0004] In the case of introducing an AI model based demodulation, the influence of different modulation modes on demodulation also needs to be considered. Therefore, different AI demodulation models need to be trained for different modulation modes, resulting in large model training overhead and large model storage overhead.

[0005] Embodiments of the present disclosure provide a model training method, a model inference method, a device, a node, a system and a medium, which can reduce the training overhead and storage overhead of the demodulation model.

[0006] According to a first aspect of embodiments of the present disclosure, a model training method is provided, which is executed by a first node and includes: obtaining a plurality of groups of sample data obtained based on different modulation modes; each group of sample data includes first to-be-demodulated data and first demodulated data, the first demodulated data and the first to-be-demodulated data correspond to each other, and the data lengths of the first demodulated data in different groups of sample data are the same; and training a first model based on the first to-be-demodulated data and the first demodulated data, the first model being used for data demodulation.

[0007] According to a second aspect of embodiments of the present disclosure, a model inference method is provided, which is executed by a second node and includes: obtaining second to-be-demodulated data from a sending end; inputting the second to-be-demodulated data into the trained first model for demodulation to obtain third demodulated data; wherein the first model is trained based on a plurality of groups of sample data obtained based on different modulation modes, each group of sample data includes first to-be-demodulated data and first demodulated data, the first demodulated data and the first to-be-demodulated data correspond to each other, and the data lengths of the first demodulated data in different groups of sample data are the same.

[0008] According to a third aspect of the embodiments of the present disclosure, a model training apparatus is provided, which comprises: a first transceiver configured to obtain a plurality of groups of sample data obtained based on different modulation modes; each group of sample data comprises first to-be-demodulated data and first demodulated data corresponding to the first to-be-demodulated data, and the data lengths of the first demodulated data in different groups of sample data are the same; and a first processing module configured to train a first model based on the first to-be-demodulated data and the first demodulated data, the first model being used for data demodulation.

[0009] According to a fourth aspect of the embodiments of the present disclosure, a model inference apparatus is provided, which comprises: a second transceiver configured to obtain second to-be-demodulated data from a sending end; and a second processing module configured to input the second to-be-demodulated data into the trained first model for demodulation to obtain third demodulated data; wherein the first model is trained based on a plurality of groups of sample data obtained based on different modulation modes, each group of sample data comprising first to-be-demodulated data and first demodulated data corresponding to the first to-be-demodulated data, and the data lengths of the first demodulated data in different groups of sample data are the same.

[0010] According to a fifth aspect of the embodiments of the present disclosure, a first node is provided. The first node comprises at least one processor and a memory storing instructions. The instructions, when executed by the first node, cause the first node to implement the model training method according to the first aspect.

[0011] According to a sixth aspect of the embodiments of the present disclosure, a second node is provided. The second node comprises at least one processor and a memory storing instructions. The instructions, when executed by the second node, cause the second node to implement the model inference method according to the second aspect.

[0012] According to a seventh aspect of the embodiments of the present disclosure, a communication system is provided. The communication system comprises a first node and a second node. The first node is configured to perform the model training method according to the first aspect, and the second node is configured to perform the model inference method according to the second aspect.

[0013] According to an eighth aspect of the embodiments of the present disclosure, a storage medium is provided, which stores instructions. When the instructions are executed on a communication device, the communication device performs the method according to the first aspect or the second aspect.

[0014] According to a ninth aspect of the embodiments of the present disclosure, a computer program product is provided. When the computer program product is executed on a communication device, the communication device performs the method according to the first aspect or the second aspect.

[0015] According to a tenth aspect of the embodiments of the present disclosure, a computer program is provided. The computer program, when running on a computer, causes the computer to perform the method according to the first aspect or the second aspect.

[0016] According to an eleventh aspect of the embodiments of the present disclosure, a chip or a chip system is provided. The chip or the chip system includes processing circuitry. The processing circuitry is configured to perform the method according to the first aspect or the second aspect.

[0017] According to the embodiments of the present disclosure, the first model is trained using the to-be-demodulated data in the sample data corresponding to different modulation modes and the demodulated data with the same data length, so that the trained first model can demodulate data for different modulation modes, and the differences between different modulation modes are ignored, thereby reducing the training overhead and storage overhead of the demodulation model without training the corresponding demodulation model for different modulation modes.

[0018] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and do not constitute a limitation on the embodiments of the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following describes the drawings required for the embodiment description. The following drawings are only some embodiments of the present disclosure, and do not specifically limit the protection scope of the present disclosure.

[0020] FIG. 1A is a schematic diagram of an architecture of a communication system according to an embodiment of the present disclosure.

[0021] FIG. 1B is a schematic diagram of an AI demodulation process with known modulation modes according to an embodiment of the present disclosure.

[0022] FIG. 1C is a schematic diagram of an AI demodulation process with unknown modulation modes according to an embodiment of the present disclosure.

[0023] FIG. 2A is a schematic diagram of an interaction of a model-based communication method according to an embodiment of the present disclosure.

[0024] FIG. 2B is a schematic diagram of a demodulated data processing process under different modulation modes according to an embodiment of the present disclosure.

[0025] FIG. 2C is a schematic diagram of an interaction of a model training method according to an embodiment of the present disclosure.

[0026] FIG. 2D is a schematic diagram of an interaction of a model inference method according to an embodiment of the present disclosure.

[0027] FIG. 3A is a schematic diagram of a flow of a model training method performed at a first node side according to an embodiment of the present disclosure.

[0028] FIG. 3B is a flow diagram of a method for model inference performed by a second node, according to an embodiment of the present disclosure.

[0029] FIG. 3C is a flow diagram of a method for model-based communication performed by a sending end, according to an embodiment of the present disclosure.

[0030] FIG. 4A is a flow diagram of a method for model training performed by a first node, according to an embodiment of the present disclosure.

[0031] FIG. 4B is a flow diagram of a method for model inference performed by a second node, according to an embodiment of the present disclosure.

[0032] FIG. 5A is a structural diagram of a model training apparatus, according to an embodiment of the present disclosure.

[0033] FIG. 5B is a structural diagram of a model inference apparatus, according to an embodiment of the present disclosure.

[0034] FIG. 6A is a structural diagram of a communication device, according to an embodiment of the present disclosure.

[0035] FIG. 6B is a structural diagram of a chip, according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0036] Embodiments of the present disclosure provide a model training method, a model inference method, an apparatus, a node, a system, and a medium.

[0037] In a first aspect, a model training method is provided. The method is performed by a first node and includes obtaining a plurality of groups of sample data based on different modulation modes, each group of sample data including first to-be-demodulated data and first demodulated data, the first to-be-demodulated data and the first demodulated data corresponding to each other, and the first demodulated data in different groups of sample data having a same data length. The method also includes training a first model based on the first to-be-demodulated data and the first demodulated data, the first model being used for data demodulation.

[0038] In embodiments of the present disclosure, the first model is trained based on the to-be-demodulated data and the demodulated data having the same data length in the sample data corresponding to different modulation modes. After training, the first model can demodulate data for different modulation modes, ignoring the differences between different modulation modes. Thus, it is not necessary to train a corresponding demodulation model for different modulation modes, reducing the training overhead and storage overhead of the demodulation model.

[0039] In some embodiments, the data length of the first demodulated data is a preset value.

[0040] In some embodiments, the preset value is a maximum value in modulation orders corresponding to different modulation modes.

[0041] In the embodiment of the present disclosure, in the case that the data length of the first demodulation data is the maximum value in the modulation orders corresponding to different modulation modes, the data length of the output data of the first model is also the maximum value.

[0042] In some embodiments, the multiple groups of sample data obtained based on different modulation modes include: obtaining multiple first to-be-demodulated data obtained based on different modulation modes; demodulating each first to-be-demodulated data based on the modulation mode corresponding to the first to-be-demodulated data to obtain second demodulation data corresponding to each first to-be-demodulated data; performing data padding on the second demodulation data with a data length less than a preset value to obtain first demodulation data; and taking the first to-be-demodulated data and the first demodulation data obtained based on the same modulation mode as a group of sample data to obtain the multiple groups of sample data.

[0043] In the embodiment of the present disclosure, by performing data padding on the second demodulation data with a data length less than a preset value to obtain the first demodulation data, the data lengths of the first demodulation data corresponding to each first to-be-demodulated data are the same, thereby ensuring that the data dimensions in the multiple groups of sample data are determined.

[0044] In some embodiments, the padded values are different from the values at each position in the second demodulation data.

[0045] In the embodiment of the present disclosure, by padding values different from the values at each position in the second demodulation data, the values at each position in the second demodulation data and the padded values can be distinguished.

[0046] In some embodiments, the values at each position in the second demodulation data include one of the following: 0, 1, and a first value, the first value being greater than 0 and less than 1.

[0047] In some embodiments, the first model is an AI model or an ML model.

[0048] In some embodiments, the first node is deployed in at least one of the following: a terminal, an access network device, a first core network element, and the first core network element for training the first model.

[0049] In a second aspect, the embodiments of the present disclosure provide a model inference method, executed by a second node, including: obtaining second to-be-demodulated data from a sending end; inputting the second to-be-demodulated data into a trained first model for demodulation to obtain third demodulated data; wherein the first model is trained according to a plurality of groups of sample data obtained based on different modulation modes, each group of sample data including first to-be-demodulated data and first demodulated data, the first demodulated data corresponding to the first to-be-demodulated data, and the data lengths of the first demodulated data in different groups of sample data being the same.

[0050] In the embodiments of the present disclosure, since the first model is trained by the first to-be-demodulated data and the first demodulated data with the same data length, the trained first model can be universal for multiple modulation modes, and can demodulate data for different modulation modes through the trained first model, ignoring the differences between different modulation modes, thereby reducing the training overhead and storage overhead of the demodulation model, and without the sending end indicating the modulation mode information, the practicality of the AI-based demodulation scheme can be effectively improved.

[0051] In some embodiments, the method further includes: dividing the third demodulated data into a first N positions and a second M positions; the numerical value of each position in the first N positions is different from the numerical value of each position in the second M positions, or the numerical value range of each position in the first N positions is different from the numerical value range of each position in the second M positions; and retaining the first N positions to obtain fourth demodulated data.

[0052] In the embodiments of the present disclosure, since the first demodulated data is obtained by data padding the second demodulated data with a data length less than a preset value when obtaining the plurality of groups of sample data obtained based on different modulation modes, and the padded numerical value is different from the numerical value of each position in the second demodulated data, the third demodulated data obtained by demodulating the second to-be-demodulated data by the first model trained based on the plurality of groups of sample data can be divided into the first N positions and the second M positions according to the numerical value, and the actual demodulated data (i.e., the fourth demodulated data) can be obtained by retaining the first N positions.

[0053] In some embodiments, the numerical value of each position in the first N positions includes one of the following: 0, 1, and a first value, the first value being greater than 0 and less than 1.

[0054] In some embodiments, the second to-be-demodulated data is obtained based on a first modulation mode, the fourth demodulated data is the same as fifth demodulated data corresponding to the second to-be-demodulated data, and the fifth demodulated data is data obtained by demodulating the second to-be-demodulated data based on the first modulation mode.

[0055] In the embodiments of the present disclosure, the fourth demodulation data is actual demodulation data of the second to-be-demodulated data obtained based on the first modulation mode, and a demodulation mode of the fourth demodulation data corresponds to the first modulation mode.

[0056] In some embodiments, the data length of the third demodulation data is a preset value.

[0057] In some embodiments, the preset value is a maximum value in modulation orders corresponding to different modulation modes.

[0058] In some embodiments, the first model is an AI model or an ML model.

[0059] In some embodiments, the second node is deployed in at least one of the following: a terminal; an access network device; a second core network network element; and the second core network network element is configured to perform data demodulation based on the trained first model.

[0060] In a third aspect, the embodiments of the present disclosure provide a model training apparatus, comprising: a first transceiver module configured to obtain a plurality of groups of sample data obtained based on different modulation modes; each group of sample data comprises first to-be-demodulated data and first demodulation data, the first demodulation data and the first to-be-demodulated data correspond to each other, and the data lengths of the first demodulation data in different groups of sample data are the same; and a first processing module configured to train a first model based on the first to-be-demodulated data and the first demodulation data, the first model being used for data demodulation.

[0061] In some embodiments, the data length of the first demodulation data is a preset value.

[0062] In some embodiments, the preset value is a maximum value in modulation orders corresponding to different modulation modes.

[0063] In some embodiments, the first transceiver module is configured to: obtain a plurality of first to-be-demodulated data obtained based on different modulation modes; demodulate each first to-be-demodulated data based on a modulation mode corresponding to the first to-be-demodulated data, to obtain second demodulation data corresponding to the first to-be-demodulated data; perform data padding on the second demodulation data with a data length less than a preset value, to obtain the first demodulation data; and take the first to-be-demodulated data and the first demodulation data obtained based on the same modulation mode as a group of sample data, to obtain a plurality of groups of sample data.

[0064] In some embodiments, the padded values in the data padding are all different from the values at each position in the second demodulation data.

[0065] In some embodiments, the values at each position in the second demodulation data comprise one of the following: 0, 1, and a first value, the first value being greater than 0 and less than 1.

[0066] In some embodiments, the first model is an AI model or an ML model.

[0067] In some embodiments, the first node is deployed in at least one of: a terminal; an access network device; a first core network element; and the first core network element is configured to train the first model.

[0068] In a fourth aspect, the embodiments of the present disclosure provide a model inference apparatus, comprising: a second transceiving module configured to obtain second to-be-demodulated data from a sending end; and a second processing module configured to input the second to-be-demodulated data into the trained first model for demodulation to obtain third demodulated data, wherein the first model is trained according to a plurality of groups of sample data obtained based on different modulation modes, each group of sample data comprises first to-be-demodulated data and first demodulated data, the first demodulated data corresponds to the first to-be-demodulated data, and the data lengths of the first demodulated data in different groups of sample data are the same.

[0069] In some embodiments, the second processing module is further configured to: divide the third demodulated data into a first N positions and a second M positions; the numerical value of each position in the first N positions is different from the numerical value of each position in the second M positions, or the numerical value range of each position in the first N positions is different from the numerical value range of each position in the second M positions; and retain the first N positions to obtain fourth demodulated data.

[0070] In some embodiments, the numerical value of each position in the first N positions comprises one of: 0, 1, and a first value, the first value being greater than 0 and less than 1.

[0071] In some embodiments, the second to-be-demodulated data is obtained based on a first modulation mode, the fourth demodulated data is the same as fifth demodulated data corresponding to the second to-be-demodulated data, and the fifth demodulated data is data obtained by demodulating the second to-be-demodulated data based on the first modulation mode.

[0072] In some embodiments, the data length of the third demodulated data is a preset value.

[0073] In some embodiments, the preset value is a maximum value in modulation orders corresponding to different modulation modes.

[0074] In some embodiments, the first model is an AI model or an ML model.

[0075] In some embodiments, the second node is deployed in at least one of: a terminal; an access network device; a second core network element; and the second core network element is configured to perform data demodulation based on the trained first model.

[0076] In a fifth aspect, an embodiment of the present disclosure provides a first node. The first node comprises at least one processor and a memory storing instructions. The instructions, when executed by the first node, enable the first node to implement the model training method in the first aspect and possible implementation manners thereof.

[0077] In a sixth aspect, an embodiment of the present disclosure provides a second node. The second node comprises at least one processor and a memory storing instructions. The instructions, when executed by the second node, enable the second node to implement the model inference method in the second aspect and possible implementation manners thereof.

[0078] In a seventh aspect, an embodiment of the present disclosure provides a communication system. The communication system comprises the first node and the second node. The first node is configured to perform the model training method in the first aspect and possible implementation manners thereof, and the second node is configured to perform the model inference method in the second aspect and possible implementation manners thereof.

[0079] In an eighth aspect, an embodiment of the present disclosure provides a storage medium. The storage medium stores instructions, which, when executed on a communication device, enable the communication device to perform the method in the first aspect, the second aspect and possible implementation manners thereof.

[0080] In a ninth aspect, an embodiment of the present disclosure provides a computer program product. The computer program product, when executed on a communication device, enables the communication device to perform the method in the first aspect, the second aspect and possible implementation manners thereof.

[0081] In a tenth aspect, an embodiment of the present disclosure provides a computer program. The computer program, when executed on a computer, enables the computer to perform the method in the first aspect, the second aspect and possible implementation manners thereof.

[0082] In an eleventh aspect, an embodiment of the present disclosure provides a chip or chip system. The chip or chip system comprises a processing circuit. The processing circuit is configured to perform the method in the first aspect, the second aspect and possible implementation manners thereof.

[0083] It can be understood that the above-mentioned first node, second node, communication system, storage medium, computer program product, computer program, chip or chip system are all used to execute the method provided by the embodiments of the present disclosure. Therefore, the beneficial effects they can achieve can refer to the beneficial effects in the corresponding method, which will not be described here.

[0084] The embodiments of the present disclosure provide a model training method, a model inference method, an apparatus, a node, a system and a medium. In some embodiments, the terms such as model training method, model training method, model-based communication method, model training method, information processing method, etc. can be replaced with each other. The terms such as model inference method, inference method, model-based communication method, information processing method, etc. can be replaced with each other. The terms such as first node, model training apparatus, communication apparatus, information processing apparatus, model training apparatus, model training apparatus, model-based communication apparatus, etc. can be replaced with each other. The terms such as second node, communication apparatus, information processing apparatus, model inference apparatus, inference apparatus, etc. can be replaced with each other. The terms such as information processing system, communication system, model-based communication system, etc. can be replaced with each other.

[0085] The embodiments of the present disclosure are not exhaustive, but only illustrate some embodiments, and are not specific limitations on the protection scope of the present disclosure. In the case of no contradiction, each step in an embodiment can be implemented as an independent embodiment, and the steps can be combined arbitrarily, for example, the scheme after removing part of the steps in an embodiment can also be implemented as an independent embodiment, and the order of the steps in an embodiment can be exchanged arbitrarily, in addition, the optional implementation manners in an embodiment can be combined arbitrarily; in addition, the embodiments can be combined arbitrarily, for example, part or all of the steps of different embodiments can be combined arbitrarily, an embodiment can be combined with the optional implementation manners of other embodiments.

[0086] In each embodiment of the present disclosure, the terms and / or descriptions between the embodiments are consistent if there is no special description and logical conflict, and can be referred to each other, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0087] The terms used in the embodiments of the present disclosure are only for the purpose of describing the specific embodiments, and not as a limitation on the present disclosure.

[0088] In the embodiments of the present disclosure, unless otherwise specified, the elements expressed in singular form, such as "one", "a", "the", "above", "said", "preceding", "this" and the like, can represent "one and only one", or "one or more", "at least one" and the like. For example, in the case of using articles such as "a", "an", "the" in English, the noun after the article can be understood as singular expression, or as plural expression.

[0089] In the embodiments of the present disclosure, "a plurality of" means two or more.

[0090] In some embodiments, the terms "at least one of," "one or more of," "a plurality of," "multiple," and the like can be used interchangeably.

[0091] In some embodiments, the recitations "at least one of A, B," "A and / or B," "in one case A, in another case B," "in response to a case A, in response to a case B," and the like can include the following technical solutions according to the case: in some embodiments A (A is executed regardless of B); in some embodiments B (B is executed regardless of A); in some embodiments, A and B are selectively executed (A and B are selectively executed); in some embodiments, A and B (A and B are executed). When there are more branches such as A, B, C, and the like, the above is similar.

[0092] In some embodiments, the recitations "A or B" and the like can include the following technical solutions according to the case: in some embodiments A (A is executed regardless of B); in some embodiments B (B is executed regardless of A); in some embodiments, A and B are selectively executed (A and B are selectively executed). When there are more branches such as A, B, C, and the like, the above is similar.

[0093] The prefix words "first", "second", and the like in the embodiments of the present disclosure are only used to distinguish different description objects, and do not constitute a limitation on the position, order, priority, quantity, or content of the description objects. The description of the description objects should refer to the description in the context of the claims or embodiments, and should not constitute an additional limitation because of the use of the prefix words. For example, the description object is "field", and the ordinal words before "field" in "first field" and "second field" do not limit the position or order between "fields", and "first" and "second" do not limit whether the "fields" modified thereby are in the same message or not, nor do they limit the order of "first field" and "second field". For another example, the description object is "level", and the ordinal words before "level" in "first level" and "second level" do not limit the priority between "levels". For another example, the quantity of the description object is not limited by the ordinal words, and can be one or more. For example, "first device", wherein the quantity of "device" can be one or more. In addition, the objects modified by different prefix words can be the same or different, for example, the description object is "device", and "first device" and "second device" can be the same device or different devices, and their types can be the same or different; for another example, the description object is "information", and "first information" and "second information" can be the same information or different information, and their contents can be the same or different.

[0094] In some embodiments, "comprising", "including", "to indicate", "carrying", can be interpreted as directly carrying A, or indirectly indicating A.

[0095] In some embodiments, the terms "in response to", "in response to determining", "in the case of", "when", "when", "if", "if" and the like can be replaced with each other.

[0096] In some embodiments, the terms "greater than", "greater than or equal to", "not less than", "more than", "more than or equal to", "not less than", "higher than", "higher than or equal to", "not less than", "above" and the like can be replaced with each other, and the terms "less than", "less than or equal to", "not greater than", "less than", "less than or equal to", "not more than", "lower than", "lower than or equal to", "not higher than", "below" and the like can be replaced with each other.

[0097] In some embodiments, the device and the like can be interpreted as physical or virtual, and the name is not limited to the name described in the embodiments. The terms "device", "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "subject" and the like can be replaced with each other.

[0098] In some embodiments, "network" can be interpreted as a device (for example, access network device, core network device, etc.) contained in the network.

[0099] In some embodiments, the terms “network devices,” “access network devices (AN devices),” “radio access network devices (RAN devices),” “base stations (BSs),” “radio base stations,” “fixed stations,” “nodes,” “access network nodes,” “access points,” “transmission points (TPs),” “reception points (RPs),” “transmission / reception points (TRPs),” “panels,” “antenna panels,” “antenna arrays,” “cells,” “macro cells,” “small cells,” “femtocells,” “picocells,” “sectors,” “cell groups,” “serving cells,” “carriers,” “component carriers,” “bandwidth parts (BWPs),” and the like can be used interchangeably.

[0100] In some embodiments, the terms "terminal," "terminal device," "user equipment (UE)," "user terminal," "mobile station (MS)," "mobile terminal (MT)," "subscriber station," "mobile unit," "subscriber unit," "wireless unit," "remote unit," "mobile device," "wireless device," "wireless communication device," "remote device," "mobile subscriber station," "access terminal," "mobile terminal," "wireless terminal," "remote terminal," "handset," "user agent," "mobile client," "client," and so on can be replaced with each other.

[0101] In some embodiments, the access network device, the core network device, or the network device can be replaced with a terminal. For example, the embodiments of the present disclosure can also be applied to a structure in which communication between the access network device, the core network device, or the network device and the terminal is replaced with communication between a plurality of terminals (e.g., device-to-device (D2D), vehicle-to-everything (V2X), etc.). In this case, the terminal can also be configured to have all or part of the functions of the access network device. In addition, the terms "uplink," "downlink," and the like can also be replaced with terms corresponding to the inter-terminal communication (e.g., "side"). For example, the uplink channel, the downlink channel, and the like can be replaced with the side channel, and the uplink, the downlink, and the like can be replaced with the sidelink.

[0102] In some embodiments, the terminal can be replaced with the access network device, the core network device, or the network device. In this case, the access network device, the core network device, or the network device can also be configured to have all or part of the functions of the terminal.

[0103] In some embodiments, obtaining data, information, and the like can comply with laws and regulations of the country where the location is.

[0104] In some embodiments, data, information, and the like can be obtained after obtaining the consent of the user.

[0105] In addition, each element, each row, or each column in the table of the embodiments of the present disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0106] FIG. 1A is a schematic diagram of an architecture of a communication system according to an embodiment of the present disclosure. The communication system 100 includes a terminal 101, an access network device 102, and a core network device 103.

[0107] In some embodiments, the terminal 101 includes at least one of a mobile phone, a wearable device, an Internet of Things device, a car with communication function, a smart car, a Pad, a computer with wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal device in industrial control, a wireless terminal device in self-driving, a wireless terminal device in remote medical surgery, a wireless terminal device in a smart grid, a wireless terminal device in transportation safety, a wireless terminal device in a smart city, a wireless terminal device in a smart home, and the like, but is not limited thereto.

[0108] In some embodiments, the access network device 102, for example, a node or device that accesses a terminal to a wireless network, can include at least one of an evolved node B (eNB), a next generation eNB (ng-eNB), a next generation node B (gNB), a node B (NB), a home node B (HNB), a home evolved node B (HeNB), a wireless backhaul device, a radio network controller (RNC), a base station controller (BSC), a base transceiver station (BTS), a base band unit (BBU), a mobile switching center, a base station in a 6th generation mobile communication technology (6G) communication system, an open base station (Open RAN), a cloud base station (Cloud RAN), a base station in other communication systems, an access node in a Wi-Fi system, but is not limited thereto.

[0109] In some embodiments, the technical solutions of the present disclosure can be applicable to an open radio access network (Open RAN) architecture, at this time, the interfaces between or within the access network devices involved in the embodiments of the present disclosure can become internal interfaces of the Open RAN, and the processes and information interactions between these internal interfaces can be realized through software or programs.

[0110] In some embodiments, the access network device can be composed of a central unit (CU) and a distributed unit (DU), where the CU can also be referred to as a control unit. The CU-DU structure can split the protocol layers of the access network device, and the functions of part of the protocol layers are controlled by the CU, and the functions of the remaining part or all of the protocol layers are distributed in the DU and controlled by the CU, but the present disclosure is not limited thereto.

[0111] In some embodiments, the core network device 103 can be one device including one or more network elements, or can be multiple devices or device groups including all or part of the one or more network elements respectively. The network element can be virtual or physical. The core network includes at least one of an evolved packet core (EPC) network, a 5G core (5GC) network, a next generation core (NGC) network, and the like.

[0112] In some embodiments, the core network can be an EPC network in a 4G system. In this case, the access network device 102 can be an eNB, for example.

[0113] In some embodiments, the core network device 103 can include a first core network element.

[0114] In some embodiments, the first core network element can be used to support data collection, training of an AI model / machine learning (ML) model, and the like.

[0115] In some embodiments, the core network device 103 can include a second core network element.

[0116] In some embodiments, the second core network element can be used to perform a model-based inference operation.

[0117] In some embodiments, the core network can be a 5GC. In this case, the access network device 102 can be a gNB, for example.

[0118] In some embodiments, the core network device 103 can include a first core network element, which can be a network data analytics function (NWDAF), an operations administration and maintenance (OAM) function, and the like, for example.

[0119] In some embodiments, the first core network element can be used to support data collection, training of an AI model / ML model, and the like, and its name is not limited thereto.

[0120] In some embodiments, the core network device 103 can include a second core network element.

[0121] In some embodiments, the second core network element can be used to perform a model-based inference operation, a model-based application operation, and the like.

[0122] In some embodiments, each network element in the core network device 103 can also be referred to as a network device, a network function, a network entity, etc., and the name thereof is not limited.

[0123] It can be understood that the communication system described in the embodiments of the present disclosure is for more clearly illustrating the technical solutions of the embodiments of the present disclosure, and does not constitute a limitation on the technical solutions proposed in the embodiments of the present disclosure. It can be known by those skilled in the art that, as the system architecture evolves and new business scenarios appear, the technical solutions proposed in the embodiments of the present disclosure are also applicable to similar technical problems.

[0124] The following embodiments of the present disclosure can be applied to the communication system 100 shown in FIG. 1A or part of the subjects in the communication system 100, but are not limited thereto. The subjects shown in FIG. 1A are exemplary, the communication system 100 can include all or part of the subjects in FIG. 1A, or other subjects other than FIG. 1A, the number and form of each subject is arbitrary, each subject can be real or virtual, the connection relationship between each subject is exemplary, each subject can not be connected or can be connected, the connection can be in any way, can be direct connection or indirect connection, can be wired connection or wireless connection.

[0125] Embodiments of the present disclosure can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G new radio (NR), Future Radio Access (FRA), New-Radio Access Technology (RAT), new radio (NR), new radio access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM (registered trademark)), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi (registered trademark)), IEEE 802.16 (WiMAX (registered trademark)), IEEE 802.20, Ultra-Wide Band (UWB), Bluetooth (Bluetooth (registered trademark)), Public Land Mobile Network (PLMN) network, Device-to-Device (D2D) system, Machine to Machine (M2M) system, Internet of Things (IoT) system, Vehicle-to-Everything (V2X), system using other communication methods, next-generation system expanded based thereon, and the like. In addition, a plurality of systems can be combined (for example, combination of LTE or LTE-A and 5G, and the like).

[0126] Hereinafter, terms related to the embodiments of the present disclosure are explained and explained.

[0127] I. Introduction to demodulator

[0128] To meet the service requirements of high-speed data transmission, large-scale traffic broadband, reliable communication in mobile scenarios, and other services in 5G and future 6G mobile communication systems, and to realize the future wireless communication vision of smart cities, smart transportation, and other future wireless communication visions, the current wireless communication network still needs to be improved and enhanced in many aspects. Among them, accurate signal demodulation is one of the basic modules of a wireless communication system that realizes low error rate high-speed transmission, and affects the overall performance of the communication system.

[0129] The demodulator in the traditional wireless communication system can be implemented by a classical method derived by the Neyman-Pearson theorem and the Bayes theorem, and such a demodulator usually needs accurate channel state information (CSI) and channel noise distribution, and its performance depends on the parameter settings of multiple modules, which can include filters, phase-locked loops, product modulators, analog-to-digital converters, etc.

[0130] The limitations of the demodulator in the traditional wireless communication system can include the following points:

[0131] 1. Large delay and complex implementation: The use of traditional demodulators will cause significant delay, and the implementation process is relatively complex.

[0132] 2. Unstable module performance: Due to factors such as module vibration, acceleration, temperature fluctuation, aging, and instability of discrete components, module performance can change, resulting in a decline in the overall performance of the receiving end.

[0133] 3. Poor environmental adaptability: Actual wireless communication channels may be subject to multipath fading, impulse noise, clutter, or discrete interference, etc., which can reduce demodulation performance. For example, some coherent demodulators require carrier synchronization, and when there is a phase error and frequency offset in the synchronization process, it will cause demodulation errors.

[0134] 4. High dependence on prior knowledge such as CSI: Traditional demodulation methods usually have a high dependence on prior knowledge, and in actual communication, especially in fast fading scenarios, it is difficult to accurately estimate CSI, so the channel model may be unknown at the receiving end.

[0135] 5. Traditional demodulation methods do not fully utilize prior waveform information and time series information.

[0136] Traditional wireless communication systems are generally designed according to mathematical theories and system models. However, as the demand for wireless services such as terminals, virtual reality, and the Internet of Things continues to increase, wireless communication systems need to handle more complex and diverse communication requirements, and traditional mathematical models may not be able to meet these requirements. In this case, deep learning is introduced into wireless communication systems as a solution.

[0137] Since the information of the data to be demodulated is represented by amplitude and phase, feature extraction is crucial for data demodulation. A deep learning-based demodulator (also referred to as a learning-based demodulator) can automatically learn and extract features in the data to be demodulated using techniques such as neural networks, thereby achieving accurate and robust demodulation, which can have advantages including the following:

[0138] 1. Flexibility and adaptability through learning from data sets.

[0139] 2. Stronger noise resistance due to the model learning features in various situations during training.

[0140] 3. Compared to traditional demodulators, deep learning-based demodulators can have fewer or even no dependence on prior knowledge such as CSI and channel noise.

[0141] Therefore, deep learning-based demodulators have great potential in dealing with complex wireless communication environments and improving communication performance.

[0142] II. Introduction to AI demodulation process with known modulation mode

[0143] Under different modulation modes such as quadrature phase shift keying (QPSK), 16 quadrature amplitude modulation (QAM), and 64 QAM, the mapping between the data to be demodulated (such as constellation symbols) and the demodulated data is different, and one data to be demodulated symbol can correspond to different data lengths. Therefore, for data to be demodulated using different modulation modes, a matching demodulation mode is needed for data demodulation. For AI model-based demodulation schemes, different modulation modes also need to be considered.

[0144] The training of an AI model requires the use of input and output data of known dimensions, while under different modulation modes, the data length of the demodulated data corresponding to one data to be demodulated symbol (such as a constellation point) may be different, for example, under 16 QAM modulation, one constellation point corresponds to 4 bits of data, while under 64 QAM modulation, one constellation point corresponds to 6 bits of data. Therefore, the outputs of AI demodulation models for 16 QAM and 64 QAM modulation are 4 bits and 6 bits of data, respectively, and the mapping relationship between constellation points and demodulated data under the two modulation modes needs to be learned separately.

[0145] FIG. 1B is a schematic diagram of an AI demodulation process with known modulation mode according to an embodiment of the present disclosure. As shown in FIG. 1B, the sending end sends the to-be-demodulated data 1 and the modulation mode to the receiving end in the process of data transmission. The modulation mode can be 16QAM, 64QAM, etc., which is not limited in the embodiments of the present disclosure. The receiving end determines the corresponding AI demodulation model according to the modulation mode, and demodulates the to-be-demodulated data 1 through the AI demodulation model corresponding to the modulation mode to obtain the demodulated data (e.g., demodulated data 1 or demodulated data 2).

[0146] In an example, when the modulation mode is 16QAM, the receiving end demodulates the to-be-demodulated data 1 through the AI demodulation model 1 corresponding to 16QAM to obtain the demodulated data 1. In an example, when the modulation mode information is 64QAM, the receiving end demodulates the to-be-demodulated data 1 through the AI demodulation model 2 corresponding to 64QAM to obtain the demodulated data 2.

[0147] In some embodiments, the sending end can be understood as a data sender. In some embodiments, the sending end can include a terminal, an access network device, a third core network element, etc.

[0148] In some embodiments, the third core network element is used for transmitting data. In some embodiments, the third core network element, for example, can be a user plane function (UPF). In some embodiments, the third core network element can be used to support the routing and forwarding of user plane data packets, and the name is not limited thereto.

[0149] III. Introduction to AI demodulation process with unknown modulation mode

[0150] FIG. 1C is a schematic diagram of an AI demodulation process with unknown modulation mode according to an embodiment of the present disclosure. As shown in FIG. 1C, when the sending end does not send the modulation mode (i.e., the modulation mode is unknown), the AI modulation mode identification model 1 is used to identify the modulation mode of the to-be-demodulated data 2 to obtain the modulation mode corresponding to the to-be-demodulated data 2. The corresponding AI demodulation model is selected according to the modulation mode, and the AI demodulation model 3 is used to demodulate the to-be-demodulated data 2 to obtain the demodulated data 3.

[0151] In some embodiments, the to-be-demodulated data and the corresponding modulation mode can be used as training data to train the AI modulation mode identification model, so that the AI modulation mode identification model can determine the modulation mode corresponding to the to-be-demodulated data based on the characteristics of the to-be-demodulated data.

[0152] For the AI demodulation process known to the modulation mode and the AI demodulation process unknown to the modulation mode, the modulation mode judgment of the data to be demodulated is needed first, and then the AI demodulation model corresponding to the modulation mode is used for data demodulation, so the two schemes need to train and deploy different AI demodulation models for different modulation modes. Due to different modulation modes used for data transmission under different application scenarios, channel environments and signal-to-noise ratios, and considering the influence of user mobility and real-time changes of wireless communication environment, the modulation mode used for data transmission is also different, therefore, training multiple AI demodulation models for multiple modulation modes needs to collect training data under each modulation mode, which has high data collection cost and large model training and storage overhead.

[0153] To solve the above problems, the embodiment of the present disclosure proposes a model training method common to modulation modes, which can realize the demodulation of data to be demodulated under different modulation modes through the trained first model.

[0154] FIG. 2A is an interaction schematic diagram of a model-based communication method according to an embodiment of the present disclosure. As shown in FIG. 2A, the embodiment of the present disclosure relates to a model training method and a model inference method, and the above method includes steps S2101 to S2107.

[0155] In the embodiment of the present disclosure, the first node, the second node and the sending end are taken as examples for illustration.

[0156] In some embodiments, the first node can be deployed in at least one of a terminal, an access network device and a first core network element.

[0157] In some embodiments, the first core network element can be used to train the first model.

[0158] In some embodiments, the first core network element can be, for example, NWDAF, OAM, etc.

[0159] In some embodiments, the first core network element can be used to support data collection, implement AI model / ML model training and other functions, and its name is not limited thereto.

[0160] In some embodiments, when the first node is deployed in a terminal, the terminal has the capability to support AI model / ML model training.

[0161] In some embodiments, the second node can be deployed in at least one of a terminal, an access network device and a second core network element.

[0162] In some embodiments, the second core network element can be configured to perform data demodulation based on the trained first model. In some embodiments, the second core network element can be configured to perform model-based inference operations, model-based application operations, and the like.

[0163] In some embodiments, the sending end can be understood as a data sender. In some embodiments, the sending end can include a terminal, an access network device, a third core network element, and the like.

[0164] In some embodiments, the third core network element is configured to transmit data. In some embodiments, the third core network element can be a UPF, for example. In some embodiments, the third core network element can be configured to support routing and forwarding of user plane data packets, without limitation of the name.

[0165] In step S2101, the first node obtains a plurality of sets of sample data obtained based on different modulation modes.

[0166] In some embodiments, each set of sample data includes first to-be-demodulated data and first demodulated data (which can be used as the true value in the first model training process). In some embodiments, the first demodulated data and the first to-be-demodulated data included in each set of sample data are corresponding. In some embodiments, the data length of the first demodulated data in different sets of sample data is the same.

[0167] In some embodiments, the data length of the first demodulated data is a preset value.

[0168] In some embodiments, the preset value can be the maximum value in the modulation orders corresponding to different modulation modes. In some embodiments, the different modulation modes are the modulation modes adopted when obtaining the plurality of first to-be-demodulated data.

[0169] In some embodiments, the plurality of sets of sample data can be pre-stored, and the first node can directly obtain the plurality of sets of sample data from a third node storing the plurality of sets of sample data.

[0170] In some embodiments, the first node can send a first message to the third node. In some embodiments, the third node can receive the first message. In some embodiments, the third node can be a sending end, a UPF, and the like, without limitation in the embodiments of the present disclosure.

[0171] In some embodiments, the first message can be used to obtain the plurality of sets of sample data.

[0172] In some embodiments, the name of the first message is not limited, for example, it can be a sample data request message, a sample data acquisition message, and the like.

[0173] In some embodiments, the third node can send, to the first node, a plurality of groups of sample data according to the first message.

[0174] In some embodiments, the first node can receive the plurality of groups of sample data.

[0175] In some embodiments, the plurality of groups of sample data can be obtained through a data processing process. In some embodiments, the data processing process can be performed by the first node, or by a fourth node having a data processing function, and the embodiments of the present disclosure do not limit this.

[0176] In some embodiments, in the case that the data processing process is performed by the fourth node having the data processing function, the first node can obtain the plurality of groups of sample data from the fourth node.

[0177] In some embodiments, the data processing process can include the following steps: Step 1, obtaining a plurality of first to-be-demodulated data obtained in different modulation modes; Step 2, demodulating each first to-be-demodulated data based on the modulation mode corresponding to each first to-be-demodulated data, to obtain second demodulated data corresponding to each first to-be-demodulated data; Step 3, performing data padding on the second demodulated data with a data length less than a preset value, to obtain first demodulated data; and Step 4, taking the first to-be-demodulated data and the first demodulated data obtained based on the same modulation mode as a group of sample data, to obtain the plurality of groups of sample data.

[0178] In some embodiments, in the case that the data length of the second demodulated data is equal to the preset value, the second demodulated data is the first demodulated data without the need for data padding on the second demodulated data.

[0179] In the following, Step 1 is described by taking the first node as an example.

[0180] In some embodiments, the first node can obtain the plurality of first to-be-demodulated data from a sending end.

[0181] In some embodiments, the first node can send a second message to the sending end. In some embodiments, the sending end can receive the second message.

[0182] In some embodiments, the second message can be used to obtain the plurality of first to-be-demodulated data.

[0183] In some embodiments, the name of the second message is not limited, for example, it can be a training data request message, a training data acquisition message, a first to-be-demodulated data acquisition message, etc.

[0184] In some embodiments, the sending end can send, to the first node, the plurality of first to-be-demodulated data according to the second message.

[0185] In some embodiments, the first node can receive a plurality of first to-be-demodulated data.

[0186] In some embodiments, the steps performed by the sending end can also be performed by the terminal, the access network device, the fourth core network element, etc. In some embodiments, the fourth core network element is configured to collect the plurality of first to-be-demodulated data.

[0187] In some embodiments, the first node can autonomously collect the plurality of first to-be-demodulated data.

[0188] In some embodiments, the first to-be-demodulated data can be to-be-demodulated symbol data. In some embodiments, the first to-be-demodulated data can be a plurality of modulation data obtained based on different modulation modes.

[0189] In some embodiments, the different modulation modes can include at least two modulation modes. In some embodiments, the at least two modulation modes can be determined from QPSK, 16QAM, 64QAM, 256QAM, etc., which are not limited in the embodiments of the present disclosure.

[0190] In an example, the different modulation modes can include 16QAM, 64QAM, and 256QAM. In this case, the plurality of first to-be-demodulated data can be a plurality of modulation data obtained based on the three modulation modes of 16QAM, 64QAM, and 256QAM.

[0191] In some embodiments, the plurality of first to-be-demodulated data can be used as input data of the first model.

[0192] In some embodiments, the plurality of first to-be-demodulated data can be used as input data of the first model training function.

[0193] In some embodiments, the first model can be an AI model or an ML model. In some embodiments, the first model can be an AI / ML function.

[0194] In some embodiments, the first model can be used for data demodulation.

[0195] The following takes the first node as an example to describe step 2.

[0196] In some embodiments, after the first node obtains the plurality of first to-be-demodulated data, the first node demodulates each first to-be-demodulated data by using a demodulation mode corresponding to the modulation mode corresponding to each first to-be-demodulated data, to obtain second demodulation data corresponding to each first to-be-demodulated data.

[0197] In an example, the plurality of first to-be-demodulated data can be modulation data obtained based on three modulation modes of 16QAM, 64QAM and 256QAM. At this time, if the modulation mode corresponding to the first to-be-demodulated data is 16QAM, the first to-be-demodulated data is demodulated based on 16QAM, and the second demodulation data corresponding to the first to-be-demodulated data can be obtained; if the modulation mode corresponding to the first to-be-demodulated data is 64QAM, the first to-be-demodulated data is demodulated based on 64QAM, and the second demodulation data corresponding to the first to-be-demodulated data can be obtained; if the modulation mode corresponding to the first to-be-demodulated data is 256QAM, the first to-be-demodulated data is demodulated based on 256QAM, and the second demodulation data corresponding to the first to-be-demodulated data can be obtained.

[0198] In some embodiments, the modulation mode to which the first to-be-demodulated data belongs corresponds to the demodulation mode to which the second demodulation data belongs.

[0199] In some embodiments, the second demodulation data can be used to determine the first demodulation data. In some embodiments, each first to-be-demodulated data has corresponding first demodulation data.

[0200] In some embodiments, the first demodulation data can be used as the true value in the first model training process. In some embodiments, the loss value of the first model can be calculated by the true value and the output data of the first model, and based on the loss value, the parameters of the first model can be adjusted until the first model converges.

[0201] The following takes the first node as an example to illustrate step 3.

[0202] In some embodiments, after the first node obtains the second demodulation data corresponding to each first to-be-demodulated data, the first node can determine the first demodulation data corresponding to each first to-be-demodulated data according to the second demodulation data.

[0203] In some embodiments, the first node judges the data length of the second demodulation data, and in the case that the data length of the second demodulation data is less than a preset value, the second demodulation data with a data length less than the preset value is filled with data so that the filled data length is the preset value, thereby obtaining the first demodulation data.

[0204] In some embodiments, when the second demodulation data with a data length less than the preset value is filled with data, the values filled by the data filling are all different from the values at each position in the second demodulation data.

[0205] In some embodiments, the values at each position in the second demodulation data can include one of 0, 1 and a first value.

[0206] In some embodiments, the first value is greater than 0 and less than 1. In some embodiments, the first value is a pure decimal number.

[0207] In an example, in the case that the value of each position in the second demodulated data is 0 or 1, the value filled by the data padding can be -1, 100, etc. integer values different from 0 and 1.

[0208] In an example, in the case that the value of each position in the second demodulated data is the first value, the value filled by the data padding can be 1.5, 9.9, etc. decimal values different from the value range of the first value.

[0209] In some embodiments, the number of values filled by the data padding is the difference between the preset value and the second value.

[0210] In some embodiments, the second value is the value of the data length corresponding to the second demodulated data whose data length is less than the preset value.

[0211] In an example, the different modulation modes can include 16QAM, 64QAM and 256QAM, and the modulation orders corresponding to 16QAM, 64QAM and 256QAM are 4, 6 and 8 respectively. In this case, the maximum value in the modulation orders corresponding to the three modulation modes is 8, i.e. the preset value is 8.

[0212] In an example, if the second demodulated data is obtained based on 16QAM demodulation, the data length of the second demodulated data is 4, and the number of values filled by the data padding when the data padding is performed on the second demodulated data is 4. In an example, if the second demodulated data is obtained based on 64QAM demodulation, the data length of the second demodulated data is 6, and the number of values filled by the data padding when the data padding is performed on the second demodulated data is 2. In an example, if the second demodulated data is obtained based on 256QAM demodulation, the data length of the second demodulated data is 8, and the data padding is not needed for the second demodulated data.

[0213] FIG. 2B is a schematic diagram of a demodulation data processing process in different modulation modes according to an embodiment of the present disclosure. As shown in FIG. 2B, there are three modulation modes, namely, modulation mode 1, modulation mode 2, and modulation mode 3. Taking 16QAM as an example of the modulation mode 1, the second demodulation data obtained based on the modulation mode 1 is {a1, a2, a3, a4}, taking 64QAM as an example of the modulation mode 2, the second demodulation data obtained based on the modulation mode 2 is {b1, b2, b3, b4, b5, b6}, and taking 256QAM as an example of the modulation mode 3, the second demodulation data obtained based on the modulation mode 3 is {c1, c2, c3, c4, c5, c6, c7, c8}. The second demodulation data obtained based on the modulation mode 1 is filled with 4-bit data by using the above data filling mode, and the first demodulation data {a1, a2, a3, a4, d1, d2, d3, d4} is obtained. The second demodulation data obtained based on the modulation mode 2 is filled with 2-bit data, and the first demodulation data {b1, b2, b3, b4, b5, b6, d5, d6} is obtained. The second demodulation data obtained based on the modulation mode 3 does not need to be filled, and the first demodulation data is still {c1, c2, c3, c4, c5, c6, c7, c8}.

[0214] It should be noted that the three modulation modes shown in FIG. 2B are only examples, and the embodiments of the present disclosure do not limit this.

[0215] The following takes the first node as an example to explain step 4.

[0216] In some embodiments, after the first node obtains the plurality of first demodulation data and the first demodulation data corresponding to each first demodulation data, the first node takes the first demodulation data and the first demodulation data obtained based on the same modulation mode as a group of sample data, thereby obtaining a plurality of groups of sample data.

[0217] In step S2102, the first node trains the first model based on the first demodulation data and the first demodulation data.

[0218] In some embodiments, after the first node obtains a plurality of groups of sample data obtained based on different modulation modes, the first node trains the first model based on the first demodulation data and the first demodulation data included in each group of sample data.

[0219] In some embodiments, the first demodulation data in each group of sample data is input into the first model to obtain output data of the first model. According to the output data and the true value (i.e., the first demodulation data corresponding to the input data of the first model), a loss value can be calculated by using a preset loss function.

[0220] In some embodiments, based on the loss value, the parameters of the first model can be adjusted until the first model converges.

[0221] In some embodiments, the preset loss function can include at least one of a square loss function, a cross-entropy loss function, a logarithmic loss function, a mean absolute error loss function, and the like, without limitation of the present disclosure.

[0222] In step S2103, the first node sends the trained first model.

[0223] In some embodiments, the second node receives the trained first model.

[0224] In some embodiments, after obtaining the trained first model, the first node can directly send the first model to the second node.

[0225] In some embodiments, the second node can send a third message to the first node.

[0226] In some embodiments, the first node can receive the third message.

[0227] In some embodiments, the third message can be used to obtain the trained first model.

[0228] In some embodiments, the name of the third message is not limited, for example, it can be a model request message, a model obtaining message, and the like.

[0229] In some embodiments, the first node can send the trained first model to the second node according to the third message.

[0230] In step S2104, the sending end sends second to-be-demodulated data.

[0231] In some embodiments, the second node receives the second to-be-demodulated data.

[0232] In some embodiments, the execution order of step S2103 and step S2104 is not limited. In an example, step S2103 and step S2104 can be executed simultaneously. In an example, step S2103 can be executed before step S2104. In an example, step S2104 can be executed before step S2103.

[0233] In some embodiments, the modulation mode of the second to-be-demodulated data is unknown.

[0234] In step S2105, the second node inputs the second to-be-demodulated data into the trained first model for demodulation to obtain third demodulated data.

[0235] In some embodiments, after receiving the second to-be-demodulated data, the second node inputs the second to-be-demodulated data into the trained first model for demodulation to obtain the third demodulated data.

[0236] In some embodiments, the data length of the third demodulation data is a preset value.

[0237] In some embodiments, the preset value is a maximum value among modulation orders corresponding to different modulation modes.

[0238] In step S2106, the second node divides the third demodulation data into the first N positions and the last M positions.

[0239] In some embodiments, after obtaining the third demodulation data, the second node judges the value of each position in the third demodulation data, and divides the third demodulation data into the first N positions and the last M positions based on the value of each position.

[0240] In some embodiments, N is a positive integer, and M is an integer. In an example, N can be an integer greater than 0 and less than or equal to the preset value, and M can be an integer greater than or equal to 0 and less than the preset value.

[0241] In some embodiments, the value of each position in the first N positions is different from the value of each position in the last M positions.

[0242] In some embodiments, the value range of each position in the first N positions is different from the value range of each position in the last M positions.

[0243] In some embodiments, the value of each position in the first N positions includes one of 0, 1 and a first value.

[0244] In some embodiments, through the above division, it can be determined which positions of the third demodulation data need to be retained and which positions of the third demodulation data do not need to be retained.

[0245] In step S2107, the second node retains the first N positions to obtain fourth demodulation data.

[0246] In some embodiments, the second node retains the data of the first N positions to obtain the fourth demodulation data.

[0247] In some embodiments, in the case that M is 0, there is no data different from 0, 1 or the first value in the third demodulation data, and at this time, the third demodulation data is the fourth demodulation data.

[0248] In some embodiments, in the case that M is a positive integer, there is data different from 0, 1 or the first value in the third demodulation data, i.e., the data of the last M positions, and at this time, the data of the first N positions in the third demodulation data is the fourth demodulation data.

[0249] In some embodiments, the fourth demodulation data is the same as fifth demodulation data corresponding to the second demodulation data.

[0250] In some embodiments, the second to-be-demodulated data is obtained based on the first modulation mode, and the fifth demodulated data is data obtained by demodulating the second to-be-demodulated data based on the first modulation mode.

[0251] In some embodiments, the fifth demodulated data can be understood as actual demodulated data of the second to-be-demodulated data, that is, the demodulation mode of the fifth demodulated data corresponds to the modulation mode of the second to-be-demodulated data.

[0252] In an example, the different modulation modes can include 16QAM, 64QAM, and 256QAM, in the case that the output data of the first model contains 0, 1-bit data, the data length of the third demodulated data is 8, and the third demodulated data can be divided into the first 6 positions and the last 2 positions, that is, the values of the 7th and 8th positions of the third demodulated data are different from 0 and 1, and close to the values filled by the data padding, which indicates that the modulation mode of the second to-be-demodulated data is 64QAM, at this time, the data of the first 6 positions is reserved as the fourth demodulated data.

[0253] The model-based communication method disclosed in the embodiments of the present disclosure can include at least one of steps S2101 to S2107. For example, steps S2101 and S2102 can be implemented as independent embodiments. For example, steps S2104 and S2105 can be implemented as independent embodiments. For example, a combination of steps S2101 to S2103 can be implemented as an independent embodiment. For example, a combination of steps S2103 to S2105 can be implemented as an independent embodiment. For example, a combination of steps S2104 to S2107 can be implemented as an independent embodiment. For example, a combination of steps S2103 to S2107 can be implemented as an independent embodiment, but is not limited thereto.

[0254] In some embodiments, steps S2103 and S2104 can be exchanged in order or executed simultaneously.

[0255] In some embodiments, steps S2103 to S2107 are optional, and one or more of these steps can be omitted or replaced in different embodiments.

[0256] In some embodiments, steps S2101 to S2103, S2106, and S2107 are optional, and one or more of these steps can be omitted or replaced in different embodiments.

[0257] In some embodiments, the names of information and the like are not limited to the names described in the embodiments, and terms such as "information", "message", "signal", "signaling", "report", "configuration", "indication", "instruction", "command", "channel", "parameter", "domain", "field", "symbol", "symbol", "codebook", "codeword", "code point", "bit", "data", "program", "chip", and the like can be replaced with each other.

[0258] In some embodiments, terms such as "carrying", "including", "containing", and the like can be replaced with each other.

[0259] In some embodiments, terms such as "first type of mobility", "second type of mobility", "first mobility", "second mobility", and the like can be replaced with each other, terms such as "downlink", "physical downlink", and the like can be replaced with each other, terms such as "side", "sidelink", "sidelink communication", "direct", "direct link", "direct communication", and the like can be replaced with each other.

[0260] In some embodiments, terms such as "acquiring", "obtaining", "receiving", "transmitting", "bidirectional transmission", "sending and / or receiving" can be replaced with each other, which can be interpreted as receiving from other subjects, acquiring from protocols, acquiring from higher layers, obtaining by processing oneself, and the like.

[0261] In some embodiments, terms such as "sending", "transmitting", "reporting", "issuing", "transmitting", "requesting", "bidirectional transmission", "sending and / or receiving", and the like can be replaced with each other.

[0262] In some embodiments, the terms "certain", "preset", "pre-set", "set", "indicated", "any", "first", and the like can be replaced with each other, and "certain A", "preset A", "pre-set A", "set A", "indicated A", "any A", "first A" can be interpreted as A predetermined in a protocol or the like, or A obtained by setting, configuring, or indicating, or a specific A, any A, or first A, but are not limited thereto.

[0263] In some embodiments, the determination or judgment can be made by a value represented by 1 bit (0 or 1), or by a true or false value (Boolean value) represented by true or false, or by comparison of numerical values (for example, comparison with a predetermined value), but is not limited thereto.

[0264] FIG. 2C is an interaction schematic diagram of a model training method according to an embodiment of the present disclosure. As shown in FIG. 2C, the present disclosure relates to a model training method, and the method includes steps S2201 to S2203.

[0265] In the embodiments of the present disclosure, the first node and the second node are taken as examples for illustration.

[0266] In step S2201, the first node obtains a plurality of groups of sample data obtained based on different modulation modes.

[0267] The optional implementation of step S2201 can refer to step S2101 in FIG. 2A and other associated parts in the embodiments involved in FIG. 2A, which will not be repeated here.

[0268] In step S2202, the first node trains the first model based on the first to-be-demodulated data and the first demodulated data.

[0269] The optional implementation of step S2202 can refer to step S2102 in FIG. 2A and other associated parts in the embodiments involved in FIG. 2A, which will not be repeated here.

[0270] In step S2203, the first node sends the trained first model.

[0271] The optional implementation of step S2203 can refer to step S2103 in FIG. 2A and other associated parts in the embodiments involved in FIG. 2A, which will not be repeated here.

[0272] The model training method related by the embodiments of the present disclosure can comprise at least one of steps S2201 to S2203. For example, steps S2201 and S2202 can be implemented as independent embodiments. For example, the combination of steps S2201 to S2203 can be implemented as an independent embodiment, but is not limited thereto.

[0273] In some embodiments, step S2203 is optional, and in different embodiments, this step can be omitted or replaced.

[0274] FIG. 2D is an interaction schematic diagram of a model inference method according to an embodiment of the present disclosure. As shown in FIG. 2D, the embodiments of the present disclosure relate to a model inference method, and the method comprises steps S2301 to S2305.

[0275] In the embodiments of the present disclosure, the first node, the second node and the sending end are taken as examples for illustration.

[0276] In step S2301, the trained first model is sent.

[0277] The optional implementation of step S2301 can refer to steps S2103 of FIG. 2A, S2203 of FIG. 2C and other associated parts in the embodiments related by FIG. 2A and FIG. 2C, which will not be repeated here.

[0278] In step S2302, the second to-be-demodulated data is sent.

[0279] The optional implementation of step S2302 can refer to step S2104 of FIG. 2A and other associated parts in the embodiments related by FIG. 2A, which will not be repeated here.

[0280] In step S2303, the second to-be-demodulated data is input into the trained first model for demodulation to obtain third demodulated data.

[0281] The optional implementation of step S2303 can refer to step S2105 of FIG. 2A and other associated parts in the embodiments related by FIG. 2A, which will not be repeated here.

[0282] In step S2304, the third demodulated data is divided into the first N positions and the last M positions.

[0283] The optional implementation of step S2304 can refer to step S2106 of FIG. 2A and other associated parts in the embodiments related by FIG. 2A, which will not be repeated here.

[0284] In step S2305, the first N positions are reserved to obtain fourth demodulated data.

[0285] The optional implementation of step S2305 can refer to step S2107 in FIG. 2A and other associated parts in the embodiments related to FIG. 2A, which will not be repeated here.

[0286] The model inference method related to the embodiments of the present disclosure can include at least one of steps S2301 to S2305. For example, steps S2302 and S2303 can be implemented as independent embodiments. For example, a combination of steps S2301 to S2303 can be implemented as an independent embodiment. For example, a combination of steps S2302 to S2305 can be implemented as an independent embodiment, but is not limited thereto.

[0287] In some embodiments, steps S2301 and S2302 can be exchanged in order or executed simultaneously.

[0288] In some embodiments, steps S2301, S2304, and S2305 are optional, and one or more of these steps can be omitted or replaced in different embodiments.

[0289] In some embodiments, steps S2304 and S2305 are optional, and one or more of these steps can be omitted or replaced in different embodiments.

[0290] FIG. 3A is a flow diagram of a model training method performed by a first node according to an embodiment of the present disclosure. As shown in FIG. 3A, the present disclosure relates to a model training method performed by the first node, and the method includes steps S3101 to S3103.

[0291] In step S3101, a plurality of groups of sample data obtained based on different modulation modes are obtained.

[0292] The optional implementation of step S3101 can refer to step S2101 in FIG. 2A, step S2201 in FIG. 2C, and other associated parts in the embodiments related to FIG. 2A and FIG. 2C, which will not be repeated here.

[0293] In step S3102, the first model is trained based on the first to-be-demodulated data and the first demodulated data.

[0294] The optional implementation of step S3102 can refer to step S2102 in FIG. 2A, step S2202 in FIG. 2C, and other associated parts in the embodiments related to FIG. 2A and FIG. 2C, which will not be repeated here.

[0295] In step S3103, the trained first model is sent.

[0296] The optional implementation of step S3103 can be referred to step S2103 in FIG. 2A, step S2203 in FIG. 2C, and other associated parts in the embodiments involved in FIG. 2A and FIG. 2C, which will not be repeated here.

[0297] The model training method involved in the embodiments of the present disclosure can include at least one of steps S3101 to S3103. For example, steps S3101 and S3102 can be implemented as independent embodiments. For example, the combination of steps S3101 to S3103 can be implemented as an independent embodiment, but is not limited thereto.

[0298] In some embodiments, step S3103 is optional, which can be omitted or replaced in different embodiments.

[0299] FIG. 3B is a flow diagram of a model inference method performed by a second node according to an embodiment of the present disclosure. As shown in FIG. 3B, the present disclosure involves a model inference method performed by a second node, and the above method includes steps S3201 to S3205.

[0300] In step S3201, a trained first model is obtained.

[0301] The optional implementation of step S3201 can be referred to step S2103 in FIG. 2A, step S2203 in FIG. 2C, step S2301 in FIG. 2D, and other associated parts in the embodiments involved in FIG. 2A, FIG. 2C, and FIG. 2D, which will not be repeated here.

[0302] In step S3202, second to-be-demodulated data is obtained.

[0303] The optional implementation of step S3202 can be referred to step S2104 in FIG. 2A, step S2302 in FIG. 2D, and other associated parts in the embodiments involved in FIG. 2A and FIG. 2D, which will not be repeated here.

[0304] In step S3203, the second to-be-demodulated data is input into the trained first model for demodulation to obtain third demodulated data.

[0305] The optional implementation of step S3203 can be referred to step S2105 in FIG. 2A, step S2303 in FIG. 2D, and other associated parts in the embodiments involved in FIG. 2A and FIG. 2D, which will not be repeated here.

[0306] In step S3204, the third demodulated data is divided into a first N positions and a second M positions.

[0307] The optional implementation of step S3204 can refer to step S2106 in FIG. 2A, step S2304 in FIG. 2D, and other associated parts in the embodiments related to FIG. 2A and FIG. 2D, which will not be repeated here.

[0308] In step S3205, the first N positions are reserved to obtain fourth demodulation data.

[0309] The optional implementation of step S3205 can refer to step S2107 in FIG. 2A, step S2305 in FIG. 2D, and other associated parts in the embodiments related to FIG. 2A and FIG. 2D, which will not be repeated here.

[0310] The model inference method related to the embodiments of the present disclosure can include at least one of steps S3201 to S3205. For example, steps S3202 and S3203 can be implemented as an independent embodiment. For example, the combination of steps S3201 to S3203 can be implemented as an independent embodiment. For example, the combination of steps S3202 to S3205 can be implemented as an independent embodiment, but is not limited thereto.

[0311] In some embodiments, the order of steps S3201 and S3202 can be exchanged or performed simultaneously.

[0312] In some embodiments, steps S3201, S3204, and S3205 are optional, and one or more of these steps can be omitted or replaced in different embodiments.

[0313] In some embodiments, steps S3204 and S3205 are optional, and one or more of these steps can be omitted or replaced in different embodiments.

[0314] FIG. 3C is a flow diagram of a method for performing model-based communication at a sending end according to an embodiment of the present disclosure. As shown in FIG. 3C, the present embodiment relates to a model-based communication method applied to the above-mentioned sending end, and the above-mentioned method includes step S3301.

[0315] In step S3301, second to-be-demodulated data is sent.

[0316] The optional implementation of step S3301 can refer to step S2104 in FIG. 2A, step S2302 in FIG. 2D, and other associated parts in the embodiments related to FIG. 2A and FIG. 2D, which will not be repeated here.

[0317] FIG. 4A is a flow diagram of a method for performing model training at a first node according to an embodiment of the present disclosure. As shown in FIG. 4A, the present embodiment relates to a model training method applied to the above-mentioned first node, and the above-mentioned method includes steps S4101 to S4102.

[0318] In step S4101, a plurality of groups of sample data obtained based on different modulation modes are obtained.

[0319] Optional implementation of step S4101 can be referred to step S2101 in FIG. 2A, step S2201 in FIG. 2C, and other associated parts in the embodiments related to FIG. 2A and FIG. 2C, which will not be repeated here.

[0320] In step S4102, the first model is trained based on the first to-be-demodulated data and the first demodulated data.

[0321] Optional implementation of step S4102 can be referred to step S2102 in FIG. 2A, step S2202 in FIG. 2C, and other associated parts in the embodiments related to FIG. 2A and FIG. 2C, which will not be repeated here.

[0322] FIG. 4B is a flow diagram of a method of performing model inference on the second node side according to an embodiment of the present disclosure. As shown in FIG. 4B, the embodiment of the present disclosure relates to a method of model inference, which is applied to the second node, and the method includes steps S4201 to S4202.

[0323] In step S4201, the second to-be-demodulated data is obtained.

[0324] Optional implementation of step S4201 can be referred to step S2104 in FIG. 2A, step S2302 in FIG. 2D, and other associated parts in the embodiments related to FIG. 2A and FIG. 2D, which will not be repeated here.

[0325] In step S4202, the second to-be-demodulated data is input into the trained first model for demodulation to obtain third demodulated data.

[0326] Optional implementation of step S4202 can be referred to step S2105 in FIG. 2A, step S2303 in FIG. 2D, and other associated parts in the embodiments related to FIG. 2A and FIG. 2D, which will not be repeated here.

[0327] In the following, the technical solutions of the embodiments of the present disclosure are exemplarily described through specific embodiments.

[0328] In some embodiments, since the number of demodulated data bits corresponding to the to-be-demodulated data symbol under different modulation modes may be different, the existing scheme usually uses different AI demodulation models to realize data demodulation under different modulation modes. In order to use one model to complete data demodulation under different modulation modes, the following data processing method is considered: for a to-be-demodulated data symbol S, the length of the demodulated data corresponding to the case of modulation mode d i (i = 1 ~ N) is L iFor the N modulation modes mainly considered in the actual communication system, the longest demodulation data length corresponding to one to-be-demodulated data symbol is L max The model training data under various modulation modes are collected, and if the modulation mode d i The demodulation data length L i <L max The L max -L i padding data are supplemented after the demodulation data in the training data, so that the total data length is L max .

[0329] In some embodiments, the padding data refers to data that is obviously different from the demodulation output data. For example, if the demodulation data is 0 and 1 bits, the padding data can be set to -1, 100, etc.

[0330] In some embodiments, for the case of directly demodulating output 0 and 1 bits for 16QAM modulation mode, the demodulation data length corresponding to one to-be-demodulated data symbol is 4.

[0331] In some embodiments, if 16QAM, 64QAM and 256QAM are considered as the data modulation modes in the actual system, and the demodulation data includes 0 and 1 bits, N = 3, and the longest demodulation data length L max is 8. For 16QAM and 64QAM modulation modes, the demodulation data length corresponding to the coordinates (m, n) of one constellation point in the constellation diagram for one to-be-demodulated data symbol is 4 and 6, respectively, so 4 and 2 padding data are needed to be supplemented, respectively, so that the demodulation data length becomes 8.

[0332] In some embodiments, the AI demodulation model data set under different modulation modes is processed using the above data processing method, so as to obtain a model training data set with determined input and output data dimensions, wherein the model input data is to-be-demodulated data, and the model output data length is L max The data set is used for AI demodulation model training that is common to modulation modes.

[0333] In some embodiments, for the application of the above modulation mode common AI demodulation model in the actual system, the model is deployed at the receiving end, the to-be-demodulated data is input, and the demodulation data with a length of L max is output, and then numerical judgment and selection are performed on the output data. If the model output data from the jth position is obviously different from the demodulation data under the normal condition, only the data at the 1st to (j-1)th positions are retained as the final demodulation data.

[0334] In some embodiments, for the case of considering three modulation modes of 16QAM, 64QAM and 256QAM and the demodulated data containing 0, 1-bit data, the output length of the AI demodulation model is 8, the model output data at the 7th and 8th positions are quite different from 0 and 1, and are closer to the invalid padding data used when processing the model data, indicating that the current data modulation mode is 64QAM, and only the model output data at the 1st to 6th positions are retained as the final demodulated data.

[0335] The embodiments of the present disclosure also propose a device for implementing any of the above methods, for example, a device comprising units or modules for implementing the steps performed by the first node in any of the above methods. For another example, another device is also proposed, comprising units or modules for implementing the steps performed by the second node in any of the above methods.

[0336] It should be understood that the division of each unit or module in the above device is only a logical functional division, and all or part of the units or modules can be integrated into one physical entity, or can be physically separated. In addition, the units or modules in the device can be implemented in the form of processor calling software: for example, the device comprises a processor connected with a memory, the memory stores instructions, and the processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules of the device, wherein the processor is, for example, a general processor such as a central processing unit (CPU) or a microprocessor, and the memory is a memory in the device or a memory outside the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuit, and the functions of part or all of the units or modules can be implemented by the design of the hardware circuit, and the hardware circuit can be understood as one or more processors; for example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC), and the functions of part or all of the units or modules are implemented by the design of the logical relationship between the elements in the circuit; for another example, in another implementation, the hardware circuit is a programmable logic device (PLD), and a field programmable gate array (FPGA) is taken as an example, which can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured by a configuration file, so as to implement the functions of part or all of the units or modules. All units or modules of the above device can be implemented in the form of processor calling software, or all units or modules can be implemented in the form of hardware circuit, or part of the units or modules can be implemented in the form of processor calling software, and the remaining part can be implemented in the form of hardware circuit.

[0337] In embodiments of the present disclosure, the processor is a circuit with signal processing capability. In one implementation, the processor can be a circuit with instruction reading and running capability, such as a CPU, a microprocessor, a graphics processing unit (GPU) (which can also be understood as a microprocessor), a digital signal processor (DSP), or the like. In another implementation, the processor can implement certain functions through a logical relationship of hardware circuits, which is fixed or can be reconfigured. For example, the processor is an ASIC or a PLD implemented hardware circuit, such as an FPGA. In a reconfigurable hardware circuit, the processor loads a configuration document to implement the hardware circuit configuration. It can be understood that the processor loads instructions to implement the functions of the above part or all units or modules. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a neural network processing unit (NPU), a tensor processing unit (TPU), a deep learning processing unit (DPU), and the like.

[0338] As shown in FIG. 5A, FIG. 5A is a structural schematic diagram of a model training apparatus provided by embodiments of the present disclosure. The structure of the model training apparatus 5100 can be as shown in FIG. 5A. The model training apparatus 5100 can be a first node. The model training apparatus 5100 includes a first transceiver module 5101 and a first processing module 5102. In some embodiments, the first transceiver module 5101 is configured to obtain a plurality of groups of sample data obtained based on different modulation modes; each group of sample data includes first to-be-demodulated data and first demodulated data, the first demodulated data and the first to-be-demodulated data correspond to each other, and the data lengths of the first demodulated data in different groups of sample data are the same; and the first processing module 5102 is configured to train a first model based on the first to-be-demodulated data and the first demodulated data, the first model being used for data demodulation.

[0339] In some embodiments, the first transceiver module 5101 described above is configured to perform at least one of the communication steps (for example, step S3103) performed by the first node in any of the above methods, and details are not repeated here. In some embodiments, the first processing module 5102 described above is configured to perform at least one of the processing steps (for example, steps S3101 and S3102) performed by the first node in any of the above methods, and details are not repeated here.

[0340] In some embodiments, the first transceiver module 5101 can include a transmitting module and / or a receiving module, which can be separate or integrated together. Alternatively, the first transceiver module 5101 can be replaced by a transceiver.

[0341] As shown in FIG. 5B, FIG. 5B is a structural schematic diagram of a model inference apparatus provided by an embodiment of the present disclosure. The structure of the model inference apparatus 5200 can be as shown in FIG. 5B. The model inference apparatus 5200 can be a second node. The model inference apparatus 5200 includes a second transceiver module 5201 and a second processing module 5202. In some embodiments, the second transceiver module 5201 is configured to obtain second to-be-demodulated data from a sending end; and the second processing module 5202 is configured to input the second to-be-demodulated data into a trained first model for demodulation to obtain third demodulated data; wherein the first model is trained according to a plurality of groups of sample data obtained based on different modulation modes, each group of sample data including first to-be-demodulated data and first demodulated data, the first demodulated data corresponding to the first to-be-demodulated data, and the data lengths of the first demodulated data in different groups of sample data being the same.

[0342] In some embodiments, the second transceiver module 5201 is configured to perform at least one of the communication steps (e.g., steps S3201 and S3202) performed by the second node in any of the methods described above, which will not be described herein again. In some embodiments, the second processing module 5202 is configured to perform at least one of the processing steps (e.g., steps S3203 to S3205) performed by the second node in any of the methods described above, which will not be described herein again.

[0343] In some embodiments, the second transceiver module 5201 can include a transmitting module and / or a receiving module, which can be separate or integrated together. Alternatively, the second transceiver module 5201 can be replaced by a transceiver.

[0344] FIG. 6A is a structural schematic diagram of a communication device provided by an embodiment of the present disclosure. The communication device 6100 can be a first node (e.g., a base station, a terminal, etc.), a second node (e.g., a base station, a terminal, etc.), a sending end (e.g., a base station, a terminal, etc.), a chip, a chip system, or a processor supporting the communication device to implement any of the methods described above, or a chip, a chip system, or a processor supporting the first node or the second node to implement any of the methods described above. The communication device 6100 can be used to implement the methods described in the above method embodiments, and details can be referred to the descriptions in the above method embodiments.

[0345] As shown in FIG. 6A, the communication device 6100 includes one or more processors 6101. The processor 6101 can be a general processor or a special-purpose processor, etc., for example, a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, the central processing unit can be used to control a network node (e.g., a base station, a baseband chip, a terminal device, a terminal device chip, a DU or a CU, etc.), execute programs, and process data of the programs. Optionally, the communication device 6100 is configured to perform any of the above methods. Optionally, the one or more processors 6101 are configured to invoke instructions to cause the communication device 6100 to perform any of the above methods.

[0346] In some embodiments, the communication device 6100 further includes one or more transceivers 6102. When the communication device 6100 includes one or more transceivers 6102, the transceiver 6102 performs at least one of the communication steps (e.g., steps S2103, S2104, S2203, S2301, S2302, but not limited to) in the above methods, and the processor 6101 performs at least one of the other steps (e.g., steps S2101, S2102, S2105-S2107, S2201, S2202, S2303-S2305, but not limited to) in the above methods. In optional embodiments, the transceiver can include a receiver and / or a transmitter, which can be separate or integrated together. Optionally, the terms transceiver, transceiving unit, transceiver, transceiving circuit, interface circuit, interface, etc. can be replaced by each other, the terms transmitter, transmitting unit, transmitter, transmitting circuit, etc. can be replaced by each other, and the terms receiver, receiving unit, receiver, receiving circuit, etc. can be replaced by each other.

[0347] In some embodiments, the communication device 6100 further includes one or more memories 6103 for storing data. Optionally, all or part of the memory 6103 can also be outside the communication device 6100. In optional embodiments, the communication device 6100 can include one or more interface circuits 6104. Optionally, the interface circuit 6104 is connected to the memory 6103, and the interface circuit 6104 can be used to receive data from the memory 6103 or other devices, and can be used to send data to the memory 6103 or other devices. For example, the interface circuit 6104 can read data stored in the memory 6103 and send the data to the processor 6101.

[0348] The communication device 6100 in the above embodiments can be the first node, the second node, or the sending end, but the scope of the communication device 6100 described in the present disclosure is not limited to this, and the structure of the communication device 6100 can not be limited by FIG. 6A. The access network device can be a standalone device or can be part of a larger device. For example, the terminal can be: (1) a standalone integrated circuit (IC), or a chip, or a chip system or subsystem; (2) a set of one or more ICs, which can optionally also include a storage component for storing data, programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, intelligent terminal device, cellular phone, wireless device, handset, mobile unit, vehicle-mounted device, network device, cloud device, artificial intelligence device, and the like; (6) other devices, and the like.

[0349] FIG. 6B is a schematic diagram of a structure of a chip according to an embodiment of the present disclosure. For the case where the communication device 6100 can be a chip or a chip system, refer to the schematic diagram of the structure of the chip 6200 shown in FIG. 6B, but the present disclosure is not limited to this.

[0350] The chip 6200 includes one or more processors 6201. The chip 6200 is configured to perform any of the above methods.

[0351] In some embodiments, the chip 6200 further includes one or more interface circuits 6202. Optionally, the terms interface circuit, interface, transceiver pin, and the like can be replaced by each other. In some embodiments, the chip 6200 further includes one or more memories 6203 for storing data. Optionally, all or part of the memory 6203 can be outside the chip 6200. Optionally, the interface circuit 6202 is connected to the memory 6203, and the interface circuit 6202 can be configured to receive data from the memory 6203 or other devices, and the interface circuit 6202 can be configured to send data to the memory 6203 or other devices. For example, the interface circuit 6202 can read the data stored in the memory 6203 and send the data to the processor 6201.

[0352] In some embodiments, the interface circuit 6202 performs at least one of the communication steps such as sending and / or receiving in the above methods. The interface circuit 6202 performing the communication steps such as sending and / or receiving in the above methods means, for example, that the interface circuit 6202 performs data interaction between the processor 6201, the chip 6200, the memory 6203, or the transceiver device.

[0353] The modules and / or devices described in each of the embodiments of the virtual device, the physical device, the chip, and the like can be combined or separated as appropriate. Optionally, part or all of the steps can also be performed by a plurality of modules and / or devices in cooperation, which is not limited herein.

[0354] The embodiment of the disclosure further provides a storage medium, and instructions are stored on the storage medium. When the instructions are executed on the communication device 6100, the communication device 6100 performs any one of the above methods. Alternatively, the storage medium is an electronic storage medium. Alternatively, the storage medium is a computer readable storage medium, but is not limited to this, and can also be a storage medium readable by other devices. Alternatively, the storage medium can be a non-transitory storage medium, but is not limited to this, and can also be a transitory storage medium.

[0355] The embodiment of the disclosure further provides a computer program product, and the computer program product is executed by the communication device 6100, so that the communication device 6100 performs any one of the above methods.

[0356] The embodiment of the disclosure further provides a computer program, and when the computer program is executed on a computer, the computer performs any one of the above methods.

[0357] Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the embodiments disclosed herein. The disclosure is intended to cover any variations, uses or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure that come within known

[0358] It should be understood that the disclosure is not limited to the precise structures as herein described and illustrated in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the disclosure is limited only by the claims that follow.

Claims

1. A model training method, performed by a first node, comprising: obtaining a plurality of groups of sample data based on different modulation manners, each group of sample data comprising first undemodulated data and first demodulated data corresponding to the first undemodulated data, and the first demodulated data in different groups of sample data having a same data length; and training a first model based on the first undemodulated data and the first demodulated data, the first model being used for data demodulation. The data length of the first demodulated data is a preset value. The preset value is a maximum value in modulation orders corresponding to the different modulation manners.

2. The method of claim 1, wherein, The obtaining of the plurality of groups of sample data based on the different modulation manners comprises: obtaining a plurality of first undemodulated data obtained in different modulation manners; demodulating each first undemodulated data based on a modulation manner corresponding to the first undemodulated data to obtain second demodulated data corresponding to the first undemodulated data; performing data padding on the second demodulated data having a data length smaller than a preset value to obtain the first demodulated data; and taking the first undemodulated data and the first demodulated data obtained based on a same modulation manner as a group of sample data to obtain the plurality of groups of sample data.

3. The method of claim 2, wherein, The data padded in the data padding is different from a value at each position in the second demodulated data.

4. The method according to any one of claims 1 to 3, wherein, The value at each position in the second demodulated data comprises one of: a first value greater than 0 and less than 1. The first model is an artificial intelligence (AI) model or a machine learning (ML) model. The first node is deployed in at least one of: a terminal; an access network device; and a first core network element configured to train the first model. 9.A model inference method, performed by a second node, comprising: obtaining second undemodulated data from a sending end; inputting the second undemodulated data into a trained first model for demodulation to obtain third demodulated data; wherein the first model is trained based on a plurality of groups of sample data obtained based on different modulation manners, each group of sample data comprising first undemodulated data and first demodulated data corresponding to the first undemodulated data, and the first demodulated data in different groups of sample data having a same data length. The method further comprises: dividing the third demodulated data into a first N positions and a second M positions, wherein a value at each position in the first N positions is different from a value at each position in the second M positions, or a value range at each position in the first N positions is different from a value range at each position in the second M positions; and retaining the first N positions to obtain fourth demodulated data.

5. The method of claim 4, wherein, The value at each position in the first N positions comprises one of: a first value greater than 0 and less than 1.

6. The method of claim 4 or 5, wherein, The second undemodulated data is obtained based on a first modulation manner, the fourth demodulated data is the same as fifth demodulated data corresponding to the second undemodulated data, and the fifth demodulated data is data obtained by demodulating the second undemodulated data based on the first modulation manner. 0； 1； ​ 7. The method according to any one of claims 1 to 6, wherein, ​ 8. The method according to any one of claims 1 to 7, wherein, ​ ​ ​ ​ ​ ​ ​ ​ 10. The method of claim 9, wherein, ​ ​ ​ 11. The method of claim 10, wherein, ​ 0； 1； ​ 12. The method of claim 10, wherein, ​ 13. The method according to any one of claims 9 to 12, wherein, A data length of the third demodulation data is a preset value.

14. The method of claim 13, wherein, The preset value is a maximum value in modulation orders corresponding to the different modulation modes.

15. The method according to any one of claims 9 to 14, wherein, The first model is an artificial intelligence (AI) model or a machine learning (ML) model.

16. The method according to any one of claims 9 to 15, wherein, The second node is deployed in at least one of the following: a terminal; an access network device; a second core network element, which is configured to perform data demodulation based on the trained first model.

17. A model training apparatus, comprising: a first transceiver module configured to obtain a plurality of groups of sample data obtained based on different modulation modes; each group of sample data comprises first to-be-demodulated data and first demodulation data, the first demodulation data and the first to-be-demodulated data correspond to each other, and data lengths of the first demodulation data in different groups of sample data are the same; a first processing module configured to train a first model based on the first to-be-demodulated data and the first demodulation data, the first model being used for data demodulation.

18. A model inference apparatus, comprising: a second transceiver module configured to obtain second to-be-demodulated data from a sending end; a second processing module configured to input the second to-be-demodulated data to a trained first model for demodulation to obtain third demodulation data; wherein the first model is trained based on a plurality of groups of sample data obtained based on different modulation modes, each group of sample data comprises first to-be-demodulated data and first demodulation data, the first demodulation data and the first to-be-demodulated data correspond to each other, and data lengths of the first demodulation data in different groups of sample data are the same.

19. A first node, comprising: at least one processor; a memory storing instructions; wherein the instructions, when executed by the first node, cause the first node to implement the model training method according to any one of claims 1 to 8.

20. A second node, comprising: at least one processor; a memory storing instructions; wherein the instructions, when executed by the second node, cause the second node to implement the model inference method according to any one of claims 9 to 16.

21. A communication system, comprising: a first node configured to implement the model training method according to any one of claims 1 to 8; a second node configured to implement the model inference method according to any one of claims 9 to 16.

22. A storage medium storing instructions, wherein, The instructions, when executed on a communication device, cause the communication device to perform the method according to any one of claims 1 to 16.

23. A computer program product comprising a computer program which, when executed on a communication device, causes the communication device to perform the method according to any one of claims 1 to 16.

Images