Communication method and apparatus

Abstract

A communication method and an apparatus. The method comprises: on the basis of first training data, determining a first mapping relationship, which is used for indicating a correspondence between N pieces of data and M groups, wherein N is a positive integer, M is an integer greater than 1 and less than N, one group among the M groups comprises some pieces of data among the N pieces of data, and the N pieces of data are generated by an AI encoder; and sending first information, which is used for indicating the first mapping relationship. In the method, a first apparatus can obtain the mapping relationship between the N pieces of data and the M groups by means of training, and the mapping relationship is indicated by the first information. On the basis of the mapping relationship, a receiving end can choose to group the N pieces of data compressed by the AI encoder, thereby enabling grouped transmission or grouped processing. Therefore, the method enables an AI compression solution to adapt to a changing transmission bandwidth, thereby improving the transmission performance.

Description

Communication method and apparatus

[0001] This application claims priority to the Chinese Patent Application No. 202410676564.0, filed on May 28, 2024, entitled "Communication method and apparatus", the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present application relates to the field of communication technology, and more particularly, to a communication method and apparatus. BACKGROUND

[0003] With the increasing richness of wireless communication application scenarios, a large amount of data oriented to new scenarios may be generated in future wireless communication systems. For example, massive data and signaling brought by new application scenarios such as Integrated Sensing and Communication (ISAC), artificial intelligence (AI) enabled wireless technology, and terahertz communication. These data have characteristics such as large data volume, high redundancy, and existence of time domain, frequency domain, or spatial domain correlation. These characteristics can be used to compress the data to be transmitted to reduce transmission overhead.

[0004] Exemplarily, by introducing an AI compression scheme, such as AI based channel state information (CSI) compression, sensing data compression, or AI data (such as training, model data, etc.) compression, the transmission overhead can be reduced. However, the current AI compression scheme cannot effectively adapt to the changing transmission bandwidth, thereby reducing the data transmission performance.

[0005] Therefore, how to make the AI compression scheme adapt to the changing transmission bandwidth is a problem to be solved. SUMMARY

[0006] The present application provides a communication method and apparatus, by indicating the mapping relationship between N data and M groups, the execution device of the AI compression scheme can effectively group the intermediate features corresponding to the AI compression, so that the AI compression scheme can effectively adapt to the changing transmission bandwidth, and improve the utilization rate of transmission resources.

[0007] In a first aspect, a communication method is provided. The execution subject of the method provided in the first aspect can be a first device. In the absence of special description, the first device in the present application can refer to the first device itself (for example, a network device or a terminal device), a component (for example, a processor, a chip, or a chip system) in the first device, or a logic module or software capable of realizing all or part of the functions of the first device. For ease of description, the first device is taken as an example in the following description.

[0008] The method comprises: determining, according to first training data, a first mapping relationship, the first mapping relationship being used to indicate a corresponding relationship between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups comprising part of the N data, the N data being generated by an AI encoder; and sending first information, the first information being used to indicate the first mapping relationship.

[0009] Based on the above scheme, the first device can train a mapping relationship between N data and M groups, and indicate the mapping relationship through first information. In this way, the receiving end of the first information can select to group the N data compressed by the AI encoder according to the mapping relationship, so as to be able to transmit or process in groups. Therefore, the above scheme can enable the AI compression scheme to adapt to changing transmission bandwidth, thereby improving transmission performance.

[0010] In some implementations, the first mapping relationship is indicated by the first grouping sequence, and the first grouping sequence comprises N grouping numbers, the N grouping numbers corresponding to the N data one by one.

[0011] In some implementations, the method further comprises: collecting the first training data; or receiving the first training data.

[0012] In some implementations, the first information is used to indicate that the first mapping relationship is added.

[0013] Based on the above scheme, the first information can indicate an added mapping relationship, thereby realizing maintenance of the mapping relationship by the first device.

[0014] In some implementations, determining, according to the first training data, the first mapping relationship comprises: updating, according to the first training data, a second mapping relationship to the first mapping relationship.

[0015] In some implementations, the first information is used to indicate that the second mapping relationship is updated to the first mapping relationship.

[0016] Based on the above scheme, the first information can indicate an updated mapping relationship, thereby realizing maintenance of the mapping relationship by the first device.

[0017] In some implementations, the first information includes first indication information, which is used to indicate P group numbers out of N group numbers in the first group sequence, where P is a positive integer less than or equal to N; wherein the P group numbers out of N group numbers in the first group sequence are different from the P group numbers out of N group numbers in the corresponding second group sequence, the first group sequence indicates the first mapping relationship, and the second group sequence indicates the second mapping relationship.

[0018] Based on the above scheme, the first indication information can indicate the different parts of the first and second block sequences. Thus, the receiving end of the first information can update the parts of the second block sequence that differ from the first block sequence based on the first indication information, thereby updating the mapping relationship with a small amount of signaling.

[0019] In some implementations, updating the second mapping relationship to the first mapping relationship based on the first training data includes: updating the second mapping relationship to the first mapping relationship based on the first training data when a first condition is met; wherein the first condition includes at least one of the following conditions: reaching a predetermined time; the first device receiving an instruction to update the second mapping relationship; or, a first parameter meeting a second condition, wherein the first parameter is used to indicate an evaluation index of the second mapping relationship.

[0020] Based on the above scheme, the first device can retrain the mapping relationship when the first condition is met. On the one hand, when the first condition includes a predetermined time or the first device receives an instruction to update the second mapping relationship, the above scheme can update the mapping relationship in a timely manner. On the other hand, when the first condition includes the first parameter satisfying the second condition, the above scheme can update as needed, reducing unnecessary updates and thus saving the training overhead introduced by updates.

[0021] In some implementations, the first parameter includes at least one of the following: mean square error (MSE), vertical deflection (VD), generalized cosine similarity (GCS), peak signal-to-noise ratio (PSNR), chamfer distance (CD), or downstream task accuracy.

[0022] In some implementations, the method further includes receiving second information, which indicates the predetermined time, and / or the first parameter.

[0023] Based on the above scheme, the second information can be used to configure the predetermined moment, and / or transmit the first parameter. The above scheme realizes the configuration or dynamic update of the first condition.

[0024] In some implementations, the method further includes determining, according to second training data, a second parameter, the second parameter being used to update the AI encoder or the AI decoder.

[0025] Based on the above scheme, the first device can determine, according to training data, a parameter used to update the AI encoder or the AI decoder. The above scheme can make the compression or processing of the AI encoder or the AI decoder more suitable for the mapping relationship. For example, in a scenario where the mapping relationship is used to select data of part of the M groups for transmission, the N data generated by the updated AI encoder can be adapted to the mapping relationship. For another example, in a scenario where the mapping relationship is used to process received data of part of the M groups, the updated AI decoder can process the N data recovered by the mapping relationship, so as to further obtain reconstructed data according to the N data.

[0026] In some implementations, the method further includes collecting the second training data, or receiving the second training data.

[0027] In some implementations, the method further includes transmitting, according to the first mapping relationship, data of M1 groups, the M1 groups belonging to the M groups, and M1 being a positive integer less than or equal to M; or receiving the data of the M1 groups, and processing the data of the M1 groups according to the first mapping relationship.

[0028] Based on the above scheme, by adjusting the parameter of M1 according to the first mapping relationship, the demand for code rate adjustment in the AI compression scenario can be met, so as to ensure the data transmission performance. In addition, the above scheme can match various types of data to be transmitted. For example, the first device can store a large number of mapping relationships with a small amount of storage space, and when the data to be transmitted changes, the first device can use a suitable mapping relationship to process the data to be transmitted, and select M1 groups for transmission.

[0029] In some implementations, the method further includes determining, from at least one mapping relationship, the first mapping relationship according to at least one energy variance corresponding to the at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest in the at least one energy variance.

[0030] Based on the above scheme, providing at least one mapping relationship can adapt to different AI models and / or transmission scenarios, and has higher flexibility and adaptability. In addition, by comparing the energy variances corresponding to the mapping relationships and selecting the first mapping relationship with the maximum energy variance to group the N data, that is, selecting the first mapping relationship with the most uneven energy distribution of the M groups, a better grouping effect can be achieved.

[0031] In some implementations, the priority of the M1 groups is higher than or equal to the priority of M-M1 groups of the M groups other than the M1 groups.

[0032] The above scheme can use limited transmission resources to send more important feature data (i.e., data of the M1 groups), achieve a better signal discarding effect, and thus further meet the demand for code rate adjustment in an AI compression scenario, thereby ensuring data transmission performance.

[0033] In some implementations, after transmitting the data of the M1 groups according to the first mapping relationship, the method further includes transmitting data of M2 groups according to the first mapping relationship, the M2 groups belonging to groups other than the M1 groups among the M groups, and M2 being a positive integer less than or equal to M-M1.

[0034] Based on the above scheme, the data in the M-M1 groups can continue to be transmitted according to the first mapping relationship, thereby improving the accuracy of the receiving end of the data of the M2 groups in obtaining the N data according to the first mapping relationship.

[0035] In some implementations, the priority of the M2 groups is higher than or equal to the priority of M-M1-M2 groups of the M groups other than the M1 groups and the M2 groups.

[0036] Based on the above scheme, the transmitting end can select the data of the most important group in the remaining groups for transmission, thereby helping the receiving end of the data of the M2 groups to recover the N data.

[0037] In some implementations, the method further includes transmitting third information, the third information being used to indicate deletion of the first mapping relationship; or receiving the third information.

[0038] Based on the above scheme, the third information can indicate deletion of the mapping relationship, and the first device can maintain the mapping relationship.

[0039] In some implementations, the third information includes the first index, and the first index indicates the first mapping relationship.

[0040] In a second aspect, a communication method is provided. The execution subject of the method provided in the second aspect can be a second device. In the absence of special description, the second device in the present application can refer to the second device itself (for example, a network device or a terminal device), a component (for example, a processor, a chip, or a chip system) in the second device, or a logic module or software capable of realizing all or part of the functions of the second device. For ease of description, the second device is taken as an example for description hereinafter.

[0041] The method comprises: receiving first information, the first information being used to indicate a first mapping relationship, the first mapping relationship being used to indicate a correspondence between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups comprising part of the N data, the N data being generated by an AI encoder; and determining the first mapping relationship according to the first information.

[0042] In some implementations, the first mapping relationship is indicated by a first grouping sequence, the first grouping sequence comprising N grouping numbers, the N grouping numbers one-to-one corresponding to the N data.

[0043] In some implementations, the method further comprises: sending first training data, the first training data being used to determine the first mapping relationship.

[0044] In some implementations, determining the first mapping relationship according to the first information comprises: adding the first mapping relationship according to the first information.

[0045] In some implementations, determining the first mapping relationship according to the first information comprises: updating a second mapping relationship to the first mapping relationship according to the first information.

[0046] In some implementations, the first information comprises first indication information, the first indication information being used to indicate P grouping numbers of N grouping numbers of a first grouping sequence, P being a positive integer less than or equal to N; wherein the P grouping numbers of the N grouping numbers of the first grouping sequence are different from P grouping numbers of N grouping numbers of a corresponding second grouping sequence, the first grouping sequence indicating the first mapping relationship, and the second grouping sequence indicating a second mapping relationship; and wherein updating the second mapping relationship to the first mapping relationship according to the first information comprises: updating, according to the first indication information, the P grouping numbers of the N grouping numbers in the second grouping sequence to the P grouping numbers of the N grouping numbers of the first grouping sequence.

[0047] In some embodiments, the method further includes: sending second information, the second information being used to indicate a predetermined time and / or a first parameter; wherein the predetermined time is used to indicate a time for determining whether to update a second mapping relationship, the second mapping relationship being different from the first mapping relationship; and the first parameter is used to indicate an evaluation index of the second mapping relationship, the first parameter being used to determine whether to update the second mapping relationship.

[0048] In some embodiments, the first parameter includes at least one of the following: MSE, VD, GCS, PSNR, CD, or at least one of downstream task accuracy.

[0049] In some embodiments, the method further includes: sending indication information of updating a second mapping relationship, the second mapping relationship being different from the first mapping relationship.

[0050] In some embodiments, the method further includes: sending second training data, the second training data being used to determine a second parameter, the second parameter being used to update the AI encoder or the AI decoder.

[0051] In some embodiments, the method further includes: according to the first mapping relationship, sending M1 groups of data, the M1 groups belonging to the M groups, M1 being a positive integer less than or equal to M; or receiving the M1 groups of data, and processing the M1 groups of data according to the first mapping relationship.

[0052] In some embodiments, the method further includes: determining the first mapping relationship from at least one mapping relationship according to at least one energy variance corresponding to the at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest among the at least one energy variance.

[0053] In some embodiments, the priority of the M1 groups is higher than or equal to the priority of M-M1 groups of the M groups other than the M1 groups.

[0054] In some embodiments, after the M1 groups of data are sent according to the first mapping relationship, the method further includes: sending M2 groups of data according to the first mapping relationship, the M2 groups belonging to groups of the M groups other than the M1 groups, M2 being a positive integer less than or equal to M-M1.

[0055] In some embodiments, the priority of the M2 groups is higher than or equal to the priority of M-M1-M2 groups of the M groups other than the M1 groups and the M2 groups.

[0056] In a third aspect, a communication apparatus is provided, which can include a processing circuit (or processor) and an input / output interface (also referred to as interface circuit) for inputting and / or outputting signals. The processing circuit is configured to perform the method of the first aspect and any possible implementation of the first aspect, or the processing circuit is configured to perform the method of the second aspect and any possible implementation of the second aspect.

[0057] In some embodiments, the processing circuit is configured to communicate with other apparatuses via the interface circuit, and perform the method of the first aspect and any possible implementation of the first aspect, or perform the method of the second aspect and any possible implementation of the second aspect.

[0058] In a fourth aspect, a communication apparatus is provided. The communication apparatus can include units, modules, or means for performing the functions of the communication apparatus.

[0059] In some embodiments, the communication apparatus can include modules, units, or means for performing the methods / operations / steps / actions described in the first aspect and any possible implementation of the first aspect, which can be hardware circuit, software, or a combination of hardware circuit and software.

[0060] In some embodiments, the communication apparatus includes a processing unit and a transceiver unit. The processing unit can be configured to determine, according to first training data, a first mapping relationship, the first mapping relationship being used to indicate a correspondence between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups including part of the N data, the N data being generated by an AI encoder; and the transceiver unit can be configured to send first information, the first information being used to indicate the first mapping relationship.

[0061] In some embodiments, the first mapping relationship is indicated by a first grouping sequence, the first grouping sequence including N grouping sequence numbers, the N grouping sequence numbers one-to-one corresponding to the N data.

[0062] In some embodiments, the processing unit is further configured to collect the first training data, or the transceiver unit is further configured to receive the first training data.

[0063] In some embodiments, the first information is used to indicate that the first mapping relationship is added.

[0064] In some embodiments, the processing unit is specifically configured to update a second mapping relationship to the first mapping relationship according to the first training data.

[0065] In some embodiments, the first information is used to indicate that the second mapping relationship is updated to the first mapping relationship.

[0066] In some embodiments, the first information comprises first indication information, the first indication information being used to indicate P packet sequence numbers in the N packet sequence numbers of the first packet sequence, P being a positive integer less than or equal to N; wherein the P packet sequence numbers in the N packet sequence numbers of the first packet sequence are different from the P packet sequence numbers in the N packet sequence numbers of the corresponding second packet sequence, the first packet sequence indicating the first mapping relationship, and the second packet sequence indicating the second mapping relationship.

[0067] In some embodiments, the processing unit is specifically configured to: in a case where a first condition is met, update the second mapping relationship to the first mapping relationship according to the first training data; wherein the first condition comprises at least one of the following conditions: a predetermined time is reached; the first device receives indication information for updating the second mapping relationship; or a first parameter satisfies a second condition, wherein the first parameter is used to indicate an evaluation index of the second mapping relationship.

[0068] In some embodiments, the first parameter comprises at least one of the following: MSE, VD, GCS, PSNR, CD, or at least one of the downstream task accuracies.

[0069] In some embodiments, the transceiving unit is further configured to: receive second information, the second information being used to indicate the predetermined time and / or the first parameter.

[0070] In some embodiments, the processing unit is further configured to: determine a second parameter according to second training data, the second parameter being used to update the AI encoder or the AI decoder.

[0071] In some embodiments, the processing unit is further configured to: collect the second training data; or the transceiving unit is further configured to: receive the second training data.

[0072] In some embodiments, the transceiving unit is further configured to: transmit M1 groups of data according to the first mapping relationship, the M1 groups belonging to the M groups, M1 being a positive integer less than or equal to M. Or, the transceiving unit is further configured to: receive the M1 groups of data; and the processing unit is further configured to: process the M1 groups of data according to the first mapping relationship.

[0073] In some embodiments, the processing unit is further configured to: determine the first mapping relationship from at least one mapping relationship according to at least one energy variance corresponding to the at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest in the at least one energy variance.

[0074] In some embodiments, the priority of the M1 groups is higher than or equal to the priority of M-M1 groups in the M groups other than the M1 groups.

[0075] In some embodiments, the transceiver is further configured to transmit, according to the first mapping relationship, M2 groups of data, the M2 groups belonging to groups of the M groups other than the M1 groups, M2 being a positive integer less than or equal to M-M1.

[0076] In some embodiments, the M2 groups have a priority higher than or equal to a priority of M-M1-M2 groups of the M groups other than the M1 groups and the M2 groups.

[0077] In some embodiments, the transceiver is further configured to transmit third information, the third information being used to indicate deletion of the first mapping relationship, or receive the third information.

[0078] In some embodiments, the third information includes the first index, the first index being used to indicate the first mapping relationship.

[0079] In some embodiments, the communication apparatus can include a module, unit or means corresponding to each of the methods / operations / steps / actions described in the second aspect and any possible implementation manner of the second aspect, which can be hardware circuit, software or a combination of hardware circuit and software.

[0080] In some embodiments, the communication apparatus includes a transceiver and a processing unit. The transceiver can be configured to receive first information, the first information being used to indicate a first mapping relationship, the first mapping relationship being used to indicate a correspondence between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups including part of the N data, the N data being generated by an AI encoder; and the processing unit can be configured to determine the first mapping relationship according to the first information.

[0081] In some embodiments, the first mapping relationship is indicated by a first grouping sequence, the first grouping sequence including N grouping sequence numbers, the N grouping sequence numbers one-to-one corresponding to the N data.

[0082] In some embodiments, the transceiver is further configured to transmit first training data, the first training data being used to determine the first mapping relationship.

[0083] In some embodiments, the processing unit is specifically configured to add the first mapping relationship according to the first information.

[0084] In some embodiments, the processing unit is specifically configured to update a second mapping relationship to the first mapping relationship according to the first information.

[0085] In some embodiments, the first information comprises first indication information, the first indication information being used to indicate P packet sequence numbers in the N packet sequence numbers of the first packet sequence, P being a positive integer less than or equal to N; wherein the P packet sequence numbers in the N packet sequence numbers of the first packet sequence are different from the P packet sequence numbers in the N packet sequence numbers of the corresponding second packet sequence, the first packet sequence indicating the first mapping relationship, and the second packet sequence indicating the second mapping relationship; and the processing unit is specifically configured to: according to the first indication information, update the P packet sequence numbers in the N packet sequence numbers of the second packet sequence to the P packet sequence numbers in the N packet sequence numbers of the first packet sequence.

[0086] In some embodiments, the transceiving unit is further configured to: send second information, the second information being used to indicate a predetermined time and / or a first parameter; wherein the predetermined time is used to indicate a time for determining whether to update a second mapping relationship, the second mapping relationship being different from the first mapping relationship; and the first parameter is used to indicate an evaluation index of the second mapping relationship, the first parameter being used to determine whether to update the second mapping relationship.

[0087] In some embodiments, the first parameter comprises at least one of the following: MSE, VD, GCS, PSNR, CD, or at least one of downstream task accuracies.

[0088] In some embodiments, the transceiving unit is further configured to: send indication information of updating a second mapping relationship, the second mapping relationship being different from the first mapping relationship.

[0089] In some embodiments, the transceiving unit is further configured to: send second training data, the second training data being used to determine a second parameter, the second parameter being used to update the AI encoder or the AI decoder.

[0090] In some embodiments, the transceiving unit is further configured to: according to the first mapping relationship, send M1 groups of data, the M1 groups belonging to the M groups, M1 being a positive integer less than or equal to M. Or, the transceiving unit is further configured to: receive the M1 groups of data; and the processing unit is further configured to: according to the first mapping relationship, process the M1 groups of data.

[0091] In some embodiments, the processing unit is further configured to: determine the first mapping relationship from at least one mapping relationship according to at least one energy variance corresponding to the at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest in the at least one energy variance.

[0092] In some embodiments, the priority of the M1 groups is higher than or equal to the priority of M-M1 groups in the M groups except the M1 groups.

[0093] In some implementations, the transceiver is further configured to transmit, according to the first mapping relationship, M2 groups of data, the M2 groups belonging to groups of the M groups other than the M1 groups, M2 being a positive integer less than or equal to M-M1.

[0094] In some implementations, the M2 groups have a priority higher than or equal to a priority of M-M1-M2 groups of the M groups other than the M1 groups and the M2 groups.

[0095] In a fifth aspect, a computer-readable storage medium is provided, and the computer-readable storage medium has stored thereon a computer program or instructions, which, when executed by a processor, cause the first aspect and any possible method of the first aspect to be performed (or implemented), or cause the second aspect and any possible method of the second aspect to be performed (or implemented).

[0096] In a sixth aspect, a computer program product is provided, and the computer program product contains a computer program or instructions, which, when executed by a processor, cause the first aspect and any possible method of the first aspect to be performed (or implemented), or cause the second aspect and any possible method of the second aspect to be performed (or implemented).

[0097] In a seventh aspect, a communication apparatus is provided, and the communication apparatus includes a processor configured to cause the first aspect and any possible method of the first aspect to be performed (or implemented), or cause the second aspect and any possible method of the second aspect to be performed (or implemented), by executing a computer program (or computer executable instructions) stored in a memory and / or by a logic circuit.

[0098] In a possible implementation, the apparatus further includes a memory. In a possible implementation, the processor and the memory are integrated together. In another possible implementation, the memory is located outside the communication apparatus. The processor can include one or more processors.

[0099] In a possible implementation, the communication apparatus further includes a communication interface configured to enable the communication apparatus to communicate with other devices, such as transmitting or receiving data and / or signals. Exemplarily, the communication interface can be a transceiver, a circuit, a bus, a module, or other types of communication interfaces.

[0100] In an implementation, the communication apparatus of the third aspect, the fourth aspect, or the seventh aspect can be a chip or a chip system.

[0101] In an eighth aspect, a chip is provided, including a processor configured to invoke a computer program or computer instructions in a memory, so that the processor performs or implements any of the implementation manners of the first aspect, or so that the processor performs or implements any of the implementation manners of the second aspect.

[0102] In some implementation manners, the processor is coupled with the memory through an interface.

[0103] In a ninth aspect, a communication system is provided, including a first device configured to perform the first aspect and any possible implementation manner of the first aspect, and a second device configured to perform the second aspect and any possible implementation manner of the second aspect.

[0104] The beneficial effects of any of the second aspect to the ninth aspect can be referred to the beneficial effects of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0105] FIG. 1 is a schematic diagram of a communication system.

[0106] FIG. 2 is a schematic diagram of an AI-based data compression and transmission process.

[0107] FIG. 3 is a schematic diagram of an encoding process of a CNN-based encoder.

[0108] FIG. 4 is a schematic flowchart of a communication method provided by an embodiment of the present application.

[0109] FIG. 5 is a schematic diagram of some packet sequences provided by an embodiment of the present application.

[0110] FIG. 6 is a schematic diagram of an information format provided by an embodiment of the present application.

[0111] FIG. 7 is a schematic diagram of determining a mapping relationship provided by an embodiment of the present application.

[0112] FIG. 8 is a schematic flowchart of a mapping relationship determination method provided by an embodiment of the present application.

[0113] FIG. 9 is a schematic flowchart of two training methods provided by an embodiment of the present application.

[0114] FIG. 10 is a schematic flowchart of another communication method provided by an embodiment of the present application.

[0115] FIG. 11 is a schematic flowchart of another training method provided by an embodiment of the present application.

[0116] FIG. 12 is a schematic flowchart of two other training methods provided by an embodiment of the present application.

[0117] FIG. 13 is a schematic diagram of an implementation manner of determining a first mapping relationship according to an embodiment of the present application.

[0118] FIG. 14 is a schematic diagram of an overall coding flow of an AI compressed signal discarding scheme according to an embodiment of the present application.

[0119] FIG. 15 is a schematic diagram of a processing flow of an encoding unit according to an embodiment of the present application.

[0120] FIG. 16 is a schematic diagram of a processing flow of a decoding unit according to an embodiment of the present application.

[0121] FIG. 17 is a schematic diagram of an AI encoder / decoder based on a residual network (ResNet) network structure according to an embodiment of the present application.

[0122] FIG. 18 is a schematic block diagram of a communication apparatus according to an embodiment of the present application.

[0123] FIG. 19 is a schematic block diagram of another communication apparatus according to an embodiment of the present application. DETAILED DESCRIPTION

[0124] To facilitate understanding of the embodiments of the present application, the following points are explained:

[0125] In the present application, the terms and / or descriptions of different embodiments are consistent and can be mutually referred to if there is no special description and logical conflict. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0126] In the present application, “at least one” means one or more, and “multiple” means two or more. “And / or” describes the association relationship of the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which can represent the following cases: A exists alone, A and B exist together, B exists alone, where A and B can be singular or plural. In the textual description of the present application, the character “ / ” generally represents an “or” relationship between the front and rear associated objects. “At least one of the following” or similar expressions means any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b and c can represent: a, or b, or c, or a and b, or a and c, or b and c, or a, b and c. Where a, b and c can be single or multiple.

[0127] In the present application, "first", "second", and various numerical numbers (e.g., #1, #2, etc.) indicate the differentiation for the convenience of description, and are not used to limit the scope of the embodiments of the present application. For example, different messages are differentiated, rather than used to describe a specific order or sequence. It should be understood that the objects thus described can be interchanged as appropriate to describe solutions other than the embodiments of the present application.

[0128] In the present application, "when", "in the case of", and "if" and the like descriptions all refer to the case that the device will make corresponding processing under certain objective conditions, and are not limited to time, and do not require the device to have a judgment action when implemented, nor does it mean that there are other limitations.

[0129] In the present application, "indicate" or "for indicating" can include direct indication and indirect indication. When describing that certain indication information is used to indicate A, it can include that the indication information directly indicates A or indirectly indicates A, and it does not mean that A must be carried in the indication information.

[0130] The indication manner involved in the embodiments of the present application should be understood as covering various methods that can enable the to-be-indicated party to know the to-be-indicated information. The to-be-indicated information can be sent as a whole, or can be sent separately into multiple sub-information, and the sending period and / or sending occasion of these sub-information can be the same or different, and the present application does not limit the sending method.

[0131] The "indication information" in the embodiments of the present application can be explicit indication, that is, directly indicated through signaling, or obtained according to the parameters indicated by the signaling, in combination with other rules or in combination with other parameters or through derivation. It can also be implicit indication, that is, obtained according to rules or relationships, or according to other parameters, or through derivation. The present application does not make specific limitations on this.

[0132] In the present application, "protocol" can refer to a standard protocol in the communication field, which can include 5G protocol, NR protocol, and related protocols applied in future communication systems, and the present application does not limit this. "Predefined" can include predefinition. For example, protocol definition. "Preconfigured" can be implemented by pre-storing corresponding codes, tables or other methods that can be used to indicate related information in the device, and the present application does not limit the implementation manner thereof.

[0133] In this application, “communication” can also be described as “data transmission”, “information transmission”, “data processing” and the like. “Transmission” includes “sending” and “receiving”. Exemplarily, the transmission can be uplink transmission, for example, the terminal device can send a signal to the network device; the transmission can also be downlink transmission, for example, the network device can send a signal to the terminal device; the transmission can also be sidelink transmission, for example, the terminal device can send a signal to another terminal device. Exemplarily, “transmission” can be air interface level transmission, or can be signal sending at chip input (I) / output (O) port, rather than air interface level transmission.

[0134] In this application, “message”, “information”, “signal” or “information element (IE)” and the like can be used interchangeably, and the name of the message or information is not limited in any way, as long as the corresponding function can be implemented.

[0135] “Sending information to XX (device)” can be understood as that the destination of the information is the device. It can include directly or indirectly sending information to the device. “Receiving information from XX (device), or receiving information from XX (device)” can be understood as that the source of the information is the device, and it can include directly or indirectly receiving information from the device. The information can be processed as necessary between the source and the destination of the information transmission, for example, format change and the like, but the destination can understand the valid information from the source. Similar expressions in this application can be understood similarly, and will not be described here. In addition, “sending” can also be understood as “output” of the chip interface, and “receiving” can also be understood as “input” of the chip interface. In other words, “sending” or “receiving” can be performed between devices, for example, the network device and the terminal device send or receive through the air interface respectively, and “sending” or “receiving” can also be performed within the device, for example, through a bus, a wire or an interface to send or receive between components, modules, chips, software modules or hardware modules within the device.

[0136] In this application, the words “exemplarily”, “for example” and the like are used to represent examples, illustrations or descriptions, and to present concepts in a specific way. Any embodiment or design scheme described as “example” in this application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. In the embodiments of this application, “of”, “corresponding”, “relevant”, “corresponding” and “associated” can be used interchangeably at times, and it should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent.

[0137] In this application, the configuration can be signaling configuration, or can be described as configuration signaling. For example, the signaling configuration includes configuration by signaling sent by the base station, which can be radio resource control (RRC) message, downlink control information (DCI), or system information block (SIB). Alternatively, the signaling configuration can also be configured to the terminal device by pre-configuration, or configured to the terminal device by pre-configuration. Here, the pre-configuration is to define or configure the value of the corresponding parameter in advance in the protocol, and store it in the terminal device when communicating with the terminal device. The pre-configured message can be modified or updated under the condition that the terminal device is connected to the network.

[0138] The present application will present various aspects, embodiments or features around a system that can include a plurality of devices, components, modules, etc. Each system can include devices, components, modules, etc. other than those illustrated and / or can not include all of the devices, components, modules, etc. discussed in connection with the figures.

[0139] The service scenarios described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems as new service scenarios appear.

[0140] In various embodiments of the present application, the size of the serial number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0141] In this specification, the reference to "one embodiment" or "some embodiments" or the like means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, the appearances of the phrases "in one embodiment" or "in some embodiments" or "in other embodiments" or "in other alternative embodiments" or the like, in various places in the specification are not necessarily all referring to the same embodiment, unless otherwise be specifically noted. The terms "comprise", "comprising", "have", "having", "include", "including", "contain", "containing", and the like, mean "including but not limited to", unless otherwise specifically noted.

[0142] The technical solutions of the embodiments of the present application can be applied to various communication systems, including but not limited to: a long term evolution (LTE) system, a new radio (NR) system, and other fifth generation (5G) mobile communication systems, a narrow band internet of things (NB-IoT) system, an enhanced machine-type communication (eMTC) system, an enhanced mobile broadband (eMBB) system, an ultra reliable low latency communications (URLLC) system, a satellite communication system, an LTE-machine-to-machine (LTE-M) system, or a system evolved after 5G, such as a future mobile communication system, and the like. th generation,5G) mobile communication systems, a narrow band internet of things (NB-IoT) system, an enhanced machine-type communication (eMTC) system, an enhanced mobile broadband (eMBB) system, an ultra reliable low latency communications (URLLC) system, a satellite communication system, an LTE-machine-to-machine (LTE-M) system, or a system evolved after 5G, such as a future mobile communication system, and the like.

[0143] The technical solutions in the present application will be described below with reference to the accompanying drawings.

[0144] FIG. 1 is a schematic diagram of a communication system 100. As shown in FIG. 1, the communication system 100 includes a radio access network 110 and a core network 120. Optionally, the communication system 100 can also include an Internet 130. The radio access network 110 can include at least one network device (e.g., 111a and 111b in FIG. 1) and at least one terminal device (e.g., 112a-112j in FIG. 1). The terminal device is connected to the network device in a wireless manner. The network device is connected to the core network 120 in a wireless or wired manner. The core network 120 can include one or more core network devices. The core network device and the network device can be independent and different physical devices, or the functions of the core network device and the logical functions of the network device can be integrated on the same physical device, or a physical device can integrate part of the functions of the core network device and part of the functions of the network device. The terminal device and the terminal device, and the network device and the network device can be connected to each other in a wired or wireless manner. The terminal device and the terminal device, the network device and the network device, and the terminal device and the network device can communicate with each other in a wireless manner through air interface resources. Exemplarily, the air interface resources can include at least one of time domain resources, frequency domain resources, code resources, and space resources. FIG. 1 is only a schematic diagram, and the communication system 100 can also include other network devices, such as a wireless relay device and a wireless backhaul device, which are not shown in FIG. 1.

[0145] The network device can be any device with wireless transceiver function. For example, the network device can be a base station for connecting a terminal device to a radio access network (RAN). The network device can also be referred to as an access network device or an access network node. It can be understood that in systems using different wireless access technologies, the names of devices with network device functions can be different. For ease of description, the apparatuses providing wireless communication access functions for terminal devices in the embodiments of the present application are collectively referred to as base stations. In the embodiments of the present application, the network device includes, but is not limited to, various forms of macro base stations (such as 111a in FIG. 1), micro base stations or indoor stations (such as 111b in FIG. 1), pico base stations, small stations, balloon stations, relay stations, access points, etc. The network device can include an evolved node B (eNB or eNodeB) in LTE, an access point (AP) in a wireless fidelity (WiFi) system, a wireless relay node, a wireless backhaul node, a transmission point (TP), or a transmission reception point (TRP), etc. It can also include a next generation NodeB (gNB) or a transmission point (TRP or TP) in a 5G system, one or a group of (including multiple antenna panels) antenna panels of a base station in a 5G system, a network node constituting a gNB or a transmission point, such as a baseband unit (BBU) or a distributed unit (DU), and can also include network devices, servers or vehicle-mounted devices, etc. in future mobile communication systems and other networks evolved after 5G. The network device can also be a module or unit that completes part of the functions of the base station, for example, it can be a central unit (CU) or a DU.

[0146] In the embodiments of the present application, the apparatus for implementing the functions of the network device can be a network device, or an apparatus capable of supporting the network device to implement the functions, such as a chip system, which can be installed in the network device. The chip system can be composed of a chip, or can include a chip and other discrete devices.

[0147] In another possible scenario, a plurality of network devices cooperates to assist a terminal to implement wireless access, and different network devices respectively implement part of functions of a base station. For example, a network device can be a CU, a DU, a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU), etc. The CU and the DU can be separately arranged, or can be included in the same network element, for example, in a BBU. The RU can be included in a radio frequency device or a radio frequency unit, for example, in a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH).

[0148] In different systems, the CU (or CU-CP and CU-UP), the DU, or the RU can also have different names, but those skilled in the art can understand their meanings. For example, in an open radio access network (O-RAN) system, the CU can also be referred to as an O-CU (open CU), the DU can also be referred to as an O-DU, the CU-CP can also be referred to as an O-CU-CP, the CU-UP can also be referred to as an O-CU-UP, and the RU can also be referred to as an O-RU. For the convenience of description, the CU, the CU-CP, the CU-UP, the DU, and the RU are taken as examples for description in this application. Any one of the CU (or the CU-CP, the CU-UP), the DU, and the RU in this application can be implemented by means of a software module, a hardware module, or a combination of a software module and a hardware module. The embodiments of this application do not limit the specific technology and the specific device form adopted by the network device.

[0149] The terminal device can be a device providing voice and / or data connectivity to users; the terminal device can also be a device having wireless connection function. The terminal device can be deployed on land, including indoor or outdoor, handheld or vehicle-mounted; can also be deployed on water surface (such as ships, etc.); can also be deployed in the air (such as airplanes, balloons and satellites, etc.). The terminal device can also be referred to as user equipment (UE), access terminal, terminal, subscriber unit, subscriber station, mobile station, mobile station (MS), mobile terminal (MT), remote station, remote terminal, mobile device, user terminal, wireless network device, user agent or user apparatus. In the embodiments of the present application, the terminal device includes but is not limited to: cellular phone, mobile phone, wireless data card, wireless modem, pad, laptop computer, notebook computer, palm computer, mobile internet device (MID), computer with wireless transceiver function, cordless phone, session initiation protocol (SIP) phone, smart phone, wireless local loop (WLL) station, personal digital assistant (PDA), handset with wireless communication function, computing device or other device connected to wireless modem, vehicle-mounted device (such as automobile, bicycle, electric vehicle, airplane, ship, train, high-speed rail, etc.), wearable device (such as smart watch, smart bracelet, pedometer, smart glasses, etc.), satellite terminal, terminal device in Internet of Things or Internet of Vehicles, and any form of terminal in future network, relay user equipment or terminal in future evolved public land mobile network (PLMN), etc.The terminal device can also be a virtual reality (VR) device, an augmented reality (AR) device, a smart point of sale (POS) machine, a customer-premises equipment (CPE), a light UE, a reduced capability UE (RedCap UE), a machine type communication (MTC) terminal, a terminal device in industrial control, a terminal device in self driving, a terminal device in remote medical treatment, a terminal device in a smart grid, a wireless terminal in transportation safety, a terminal device in a smart city, a terminal device in a smart home, a haptic terminal device, a smart home device (e.g., a refrigerator, a television, an air conditioner, an electricity meter, etc.), a smart robot, a mechanical arm, a plant device, a wireless terminal in self driving, or a flight device (e.g., a smart robot, a hot air balloon, a drone, an airplane), and the like. The terminal device can also be a vehicle device, such as a whole vehicle device, a vehicle-mounted module, a vehicle-mounted chip, an on board unit (OBU), a telematics box (T-BOX), and the like. The terminal device can also be other devices with terminal functions, for example, the terminal device can also be a device in device to device (D2D) communication. The embodiments of the present application are not limited in this regard.

[0150] In the embodiments of the present application, the device for implementing the function of the terminal device can be a terminal device, or a device capable of supporting the terminal device to implement the function, such as a chip or a chip system, which can be installed in the terminal device. The chip system can be composed of a chip, or can include a chip and other discrete devices. In the technical solutions of the embodiments of the present application, the device for implementing the function of the terminal device is a terminal device, which can also be referred to as a terminal. In the following, the terminal device can be taken as an example of a UE to describe the technical solutions provided in the embodiments of the present application.

[0151] The roles of base stations and terminals can be relative, for example, the helicopter or drone 112i in FIG. 1 can be configured as a mobile base station, and for those terminals 112j accessing the wireless access network 110 through 112i, the terminal 112i is a base station; but for the base station 111a, 112i is a terminal, that is, 111a and 112i communicate through a wireless air interface protocol. Of course, 111a and 112i can also communicate through a base station-to-base station interface protocol, in which case 112i is also a base station relative to 111a. Therefore, both base stations and terminals can be collectively referred to as communication devices, and 111a and 111b in FIG. 1 can be referred to as communication devices with base station functions, and 112a-112j in FIG. 1 can be referred to as communication devices with terminal functions.

[0152] The network device and the terminal device can communicate through a wireless link. The transmission link from the network device to the terminal device can be referred to as a downlink (DL) or a downlink channel, used to transmit a downlink signal. The transmission link from the terminal device to the network device can be referred to as an uplink (UL) or an uplink channel, used to transmit an uplink signal. The transmission link from the terminal device to the terminal device can be referred to as a sidelink (SL) or a sidelink channel.

[0153] Wireless communication application scenarios are increasingly diverse, and in future wireless communication systems, there will be a lot of data oriented to new scenarios, and there are new needs to transmit these new scenario data. For example, massive data and signaling brought by new application scenarios such as ISAC, AI-enabled wireless technology, and terahertz communication. Therefore, in a future radio access network system, for example, there may be multiple data types, and different data types of data need to be transmitted under different scenarios or tasks.

[0154] A large amount of data oriented to new scenarios will be generated in future wireless communication. Among them, data oriented to new scenarios can be data derived from new emerging application scenarios in future wireless communication systems, for example, RAN data that needs to be transmitted over the air, or local data generated within the RAN. The above data can include data of multiple data types (and possibly data subtypes), such as perception data, AI data, or channel data, etc.

[0155] Exemplarily, perception data can include 2-dimensional (D) or 3D imaging data (such as acquired environmental reflection points, environmental patches), environmental reconstruction data, point cloud data, radio frequency maps, or positioning data, etc.

[0156] Exemplarily, the AI data or edge AI data can include AI model data, training data, gradient data, update data of the gradient, inference result, feature information extracted by a neural network, or performance data, etc.

[0157] Exemplarily, the channel data can include a channel matrix, channel information fed back by a device in a multi-antenna system, or channel state information (CSI) data, etc.

[0158] In a possible implementation, the transmitted data or data type is perception data. In another possible implementation, the transmitted data or data type is point cloud data. In still another possible implementation, the transmitted data or data type is positioning data and inference result.

[0159] The above data can have characteristics such as large data volume, more redundancy, and time / frequency / space correlation. For example, imaging data and radar detection data have strong sparsity, and positioning tracking data, environmental imaging / reconstruction data, and AI training data obtained in continuous time have strong time correlation.

[0160] By utilizing the sparsity and correlation, the data to be transmitted can be compressed to reduce transmission overhead. By introducing an AI compression scheme, the data characteristics can be more fully utilized, thereby improving the efficiency of data compression, such as CSI compression based on an AI scheme, perception data, AI data (such as training and model data) compression, etc.

[0161] FIG. 2 is a schematic diagram of an AI-based data compression and transmission process.

[0162] Referring to FIG. 2, the input data can be channel data (such as CSI data), perception data (such as point cloud data), AI data or edge AI data (such as AI training data and AI model data), etc. The input data can pass through an AI encoder to obtain intermediate features. The intermediate features can be quantized to obtain a binary code stream. The obtained binary code stream can be processed by channel coding, modulation, etc., and then transmitted in a wireless channel. After receiving the signal, the receiving end correspondingly performs demodulation, channel decoding, dequantization, AI decoding, etc., to obtain reconstructed data.

[0163] Exemplarily, the AI encoder or AI decoder can be implemented using a convolutional neural network (CNN), a transformer, etc.

[0164] FIG. 3 is a schematic diagram of an encoding process of a CNN-based encoder. FIG. 3 only takes the CNN-based encoder as an example and does not constitute a limitation on the present application. The embodiments of the present application can also be applicable to other encoders, for example, a transformer-based encoder.

[0165] Referring to FIG. 3, the input data can be subjected to multi-layer CNN or ResNet operation to obtain a preliminary result. Then, after the preliminary result is straightened into a vector, it is compressed to an intermediate feature of a specific dimension through FCN operation. For example, the intermediate feature can exist in the form of a floating-point number.

[0166] However, the model of a single AI encoder can only be compressed into an intermediate feature of a fixed length, which cannot adapt to a changing transmission bandwidth. Preparing multiple AI encoder models corresponding to different compression lengths will result in a large storage overhead and an increase in network parameter quantity.

[0167] Therefore, how to enable the AI compression scheme to effectively adapt to a changing transmission bandwidth and improve transmission resource utilization is a problem to be solved.

[0168] FIG. 4 is a schematic flowchart of a communication method 400 provided by an embodiment of the present application. The method 400 enables an execution device of an AI compression scheme to effectively group the intermediate features corresponding to the AI compression by indicating the mapping relationship between N data and M groups, so that the AI compression scheme can adapt to a changing transmission bandwidth and improve transmission resource utilization. The optional operations in the method 400 are indicated by dashed lines in FIG. 4. The method 400 will be described below in conjunction with FIG. 4.

[0169] S440, the first device determines the first mapping relationship according to the first training data.

[0170] In the case where no special description is made, the first device in the present application can refer to the first device itself (for example, a network device or a terminal device), or a component (for example, a processor, a chip, or a chip system, etc.) in the first device, or a logic module or software capable of realizing all or part of the functions of the first device. For the convenience of description, the first device will be described below as an example.

[0171] Exemplarily, the first device can be a network device or a terminal device, but the present application is not limited thereto, for example, the first device can also be a core network element.

[0172] The first training data can include one or more data (for example, N data). The first training data can be used to determine the first mapping relationship. Exemplarily, the first device determines the first mapping relationship according to the first training data and an AI model.

[0173] The present application does not limit the specific form of the first training data. In some examples, the first training data can be related to the task to be performed by the first device. For example, the first device is to compress CSI data, and the N data included in the first training data can be CSI data. For another example, the first device is to compress AI data, and the N data included in the first training data can be AI data. For another example, the first device is to compress perception data, and the N data included in the first training data can be perception data.

[0174] The N data can be an N-dimensional vector including N data elements, a sequence including N data elements with a length (or size) of N, N-dimensional feature data, feature data (or intermediate feature) with a length (or size) of N, or binary data divided into N groups, etc. The present application does not limit the specific form of the N data, and the N data can also be in other forms. Among them, the N data can also be referred to as N feature data or have other names.

[0175] In some examples, the first device can generate the first mapping relationship according to the first training data. For example, the first mapping relationship does not exist in the set of mapping relationships, or in other words, the first mapping relationship does not exist in advance. The first device generates the first mapping relationship from scratch according to the first training data. In other examples, the first device updates the second mapping relationship to the first mapping relationship according to the first training data. For example, the second mapping relationship can be any mapping relationship in the set of mapping relationships, or in other words, the second mapping relationship already exists in advance. The first device can retrain the second mapping relationship according to the first training data to obtain the first mapping relationship. In other words, the first device can obtain the first mapping relationship based on the second mapping relationship according to the first training data.

[0176] The first mapping relationship can be one mapping relationship or multiple mapping relationships, which is not limited by the present application.

[0177] Optionally, the first mapping relationship is used to indicate the correspondence between the N data and the M groups (or M sets). Among them, N can be a positive integer. M can be an integer greater than 1 and less than N.

[0178] In some examples, it is assumed that the N data includes N1 data and N2 data, and the M groups includes M1 groups and M2 groups. Wherein, N1+N2≤N, M1+M2≤M, N1, N2, M1 and M2 are positive integers. For example, the first mapping relationship is used to indicate that the N1 data correspond to the M1 groups; the first mapping relationship is used to indicate that the N2 data correspond to the M2 groups. For another example, the first mapping relationship is used to indicate that the N1 data and the N2 data correspond to the M1 groups. Or, the first mapping relationship is used to indicate that the N1 data correspond to the M1 groups; the first mapping relationship is used to indicate that the N2 data correspond to the M1 groups.

[0179] Optionally, one of the M groups can include part of the N data. For example, it is assumed that N=8 and M=4, then one of the 4 groups can include part of the 8 data, for example, 3 data. In the M groups, different groups can include different data, or the same data, which is not limited in the present application. The total number of data included in each of the M groups can be equal to N, or less than N. In other words, the M groups can include the N data, or part of the N data. For example, preliminary screening is performed on the N data, resulting in a number less than N.

[0180] Optionally, the first mapping relationship is used to indicate the grouping manner of the N data. For example, the first mapping relationship is used to indicate that every n data in the N data is a group. Wherein, n is an integer less than N. As an example, n can be divisible by N. As another example, n is not divisible by N. In this case, the number of data in the last group can be less than n.

[0181] Optionally, the first mapping relationship is used to indicate that the N data is divided into the M groups.

[0182] Optionally, the first mapping relationship is used to indicate the data included in each of the M groups, which belongs to the N data. In other words, the first mapping relationship is used to indicate the N data in the M groups.

[0183] Optionally, the N data is generated by an AI encoder. For example, the N data can be obtained by processing the input data by the AI encoder. Optionally, the N data is used for processing by an AI decoder. For example, the N data is processed by the AI decoder, and the reconstructed data can be obtained.

[0184] S450, the first device sends the first information to the second device. Correspondingly, the second device receives the first information from the first device.

[0185] In the case where no special description is made, the second device in the present application can refer to the second device itself (for example, a network device or a terminal device), can refer to a component (for example, a processor, a chip, or a chip system, etc.) in the second device, or can refer to a logic module or software capable of realizing all or part of the functions of the second device. For the convenience of description, the second device is taken as an example for description hereinafter.

[0186] Exemplarily, the first device can be a network device or a terminal device, but the present application is not limited thereto, for example, the first device can also be a core network element.

[0187] Optionally, the first information is used to indicate the first mapping relationship. The first information can be direct indication information, for example, the first information can include the first mapping relationship. The first information can also be indirect indication information, for example, the second device can determine the first mapping relationship according to the first information.

[0188] In the case where the first information includes the first mapping relationship, it can be understood that the first information includes original information of the first mapping relationship, or it can be understood that the first information includes a compressed code stream of the first mapping relationship, and the present application does not determine this.

[0189] The present application does not limit the specific name of the first information, and the first information can also be referred to as update indication, indication information, mapping information, or other names.

[0190] S460, the second device determines the first mapping relationship according to the first information.

[0191] In some examples, the second device can divide the N data into M groups according to the first mapping relationship, and selectively transmit data of at least one group of the M groups. In other examples, the second device can receive data of part of the groups of the M groups; and the second device can process the received data to obtain reconstructed data according to the first mapping relationship.

[0192] Based on the above scheme, the first device can train to obtain a mapping relationship between the N data and the M groups, and indicate the mapping relationship through the first information. In this way, the receiving end of the first information can group the N data compressed by the AI encoder according to the mapping relationship, so as to be able to transmit or process in groups. Therefore, the above scheme can make the AI compression scheme adapt to the changing transmission bandwidth, thereby improving the transmission performance.

[0193] In some possible implementation manners, the first mapping relationship is indicated by a first bitmap. Optionally, the first mapping relationship includes the first bitmap. For example, S440 includes: determining, by the first device, the first bitmap according to the first training data. For another example, the first information is used to indicate the first bitmap. For yet another example, S460 includes: determining, by the second device, the first bitmap according to the first information.

[0194] For example, the length of the first bitmap can be N. N bits in the first bitmap correspond to N data one by one. For example, "1" in the first bitmap indicates group 1, and "0" in the first bitmap indicates group 2. For example, the first bitmap is {1001}, and the first and fourth data of N=4 data belong to group 1, and the second and third data of the 4 data belong to group 2.

[0195] Based on the above scheme, the first mapping relationship can be indicated by the first bitmap. The overhead of indicating the mapping relationship by the first bitmap is lower.

[0196] In some possible implementation manners, the first mapping relationship is indicated by a first mapping table. Optionally, the first mapping relationship includes the first mapping table. For example, S440 includes: determining, by the first device, the first mapping table according to the first training data. For another example, the first information is used to indicate the first mapping table. For yet another example, S460 includes: determining, by the second device, the first mapping table according to the first information.

[0197] For example, the first mapping table can include two fields or columns. One field or column can indicate one data or the number of data of N data, and the other field or column can indicate one group of M groups. In this way, each record or row can represent the corresponding relationship between one data and one group.

[0198] In some possible implementation manners, the first mapping relationship is indicated by a first grouping sequence. Optionally, the first mapping relationship includes the first grouping sequence. For example, S440 includes: determining, by the first device, the first grouping sequence according to the first training data. For another example, the first information is used to indicate the first grouping sequence. For yet another example, S460 includes: determining, by the second device, the first grouping sequence according to the first information.

[0199] For example, the length of the first bitmap can be N. N bits in the first bitmap correspond to N data one by one. For example, "1" in the first bitmap indicates group 1, and "0" in the first bitmap indicates group 2. For example, the first bitmap is {1001}, and the first and fourth data of N=4 data belong to group 1, and the second and third data of the 4 data belong to group 2.

[0200] FIG. 5 is a schematic diagram of some grouping sequences provided by an embodiment of the present application. FIG. 5 is merely an example and does not constitute a limitation on the present application. The first grouping sequence can adopt the form shown in FIG. 5 or other forms.

[0201] Suppose there are N = 8 data, e.g., data 1, data 2, …, data 8, data 1 to 8 can one-to-one correspond to each group sequence shown in FIG. 5.

[0202] As some examples, the group sequence numbers in the first group sequence can be regularly arranged.

[0203] For example, the group sequence numbers can be arranged in ascending order. The first group sequence can be shown in (a) of FIG. 5, which can be represented as [1, 1, 2, 2, 3, 3, 4, 4]. Among them, 1, 2, 3, and 4 are group sequence numbers of M = 4 groups. For example, group 1 includes data 1 and data 2, group 2 includes data 3 and data 4, group 3 includes data 5 and data 6, and group 4 includes data 7 and data 8. In some examples, the first information can indicate [1, 1, 2, 2, 3, 3, 4, 4]. In other examples, the first information can indicate [1, 2, 3, 4] and indicate or pre-configure that each group sequence number is repeated twice.

[0204] For another example, the group sequence numbers can be arranged in descending order. The first group sequence can be shown in (b) of FIG. 5, which can be represented as [4, 4, 3, 3, 2, 2, 1, 1]. For example, group 1 includes data 7 and data 8, group 2 includes data 5 and data 6, group 3 includes data 3 and data 4, and group 4 includes data 1 and data 2. In some examples, the first information can indicate [4, 4, 3, 3, 2, 2, 1, 1]. In other examples, the first information can indicate [4, 3, 2, 1] and indicate or pre-configure that each group sequence number is repeated twice.

[0205] As other examples, the group sequence numbers in the first group sequence can be irregularly arranged. For example, the first group sequence can be shown in (c) of FIG. 5, which can be represented as [4, 2, 1, 4, 3, 1, 2, 3]. For example, group 1 includes data 3 and data 6, group 2 includes data 2 and data 7, group 3 includes data 5 and data 8, and group 4 includes data 1 and data 4.

[0206] As still other examples, the first group sequence can include regularly arranged sequence numbers and irregularly arranged sequence numbers. For example, the first group sequence can be shown in (d) of FIG. 5, which can be represented as [1, 2, 3, 4, 4, 3, 1, 2]. For example, group 1 includes data 1 and data 7, group 2 includes data 2 and data 8, group 3 includes data 3 and data 6, and group 4 includes data 4 and data 5.

[0207] The name of the first grouping sequence is not limited in the present application. The first grouping sequence can also be referred to as a grouping pattern (or pattern), a dropping pattern (or pattern), a pattern (or pattern), a dropping sequence, a priority sequence, a grouping vector, a dropping vector, a vector, or has other names.

[0208] Based on the above scheme, the first mapping relationship can be indicated by the first grouping sequence. The precision of indicating the mapping relationship by using the first grouping sequence is higher.

[0209] Exemplarily, different mapping relationships can correspond to different numbers. For example, the number of (a) in FIG. 5 can be 0 (or 00); the number of (b) in FIG. 5 can be 1 (or 01); the number of (c) in FIG. 5 can be 2 (or 10); and the number of (d) in FIG. 5 can be 3 (or 11).

[0210] FIG. 6 is a schematic diagram of an information format provided by an embodiment of the present application. FIG. 6 is exemplary and does not constitute a limitation on the present application. For example, the first information in the embodiment of the present application can adopt the format shown in FIG. 6, or can adopt other formats. The exemplary functions of the first information will be introduced below in combination with FIG. 6.

[0211] Referring to (a) in FIG. 6, the first information can include at least one of an update type field, a number field, or a mapping relationship field. The content in the update type field can be used to indicate an update type, for example, at least one of adding a mapping relationship, modifying a mapping relationship, or deleting a mapping relationship. The content in the number field can be used to indicate the number (or index, identifier) of one or more mapping relationships, for example, 00001. The content in the mapping relationship field can be used to indicate a mapping relationship, for example, a first mapping relationship, a compressed first mapping relationship, part of the information of the first mapping relationship, or other information.

[0212] The name of the above field is not limited in the present application. For example, the update type field can also be referred to as a type field or has other names. For another example, the number field can also be referred to as a pattern number field, a grouping sequence index field, or has other names. For another example, the mapping relationship field can also be referred to as a pattern information code stream field, an information field, or has other names.

[0213] The order of the above field is not limited in the present application, and the order of the above field can be adjusted arbitrarily. The first information can include more or fewer fields than those shown in (a) in FIG. 6, and the present application is not limited thereto.

[0214] In some possible implementation, S440 comprises: generating, by the first device, the first mapping relationship according to the first training data. In some possible implementation, the first information is used to indicate adding the first mapping relationship. In some possible implementation, S460 comprises: adding, by the second device, the first mapping relationship according to the first information.

[0215] In some examples, the first information can comprise original information of the first mapping relationship. In some other examples, the first information can comprise compressed first mapping relationship. For example, the first mapping relationship can be compressed by entropy coding to obtain the compressed first mapping relationship.

[0216] For example, the first information can comprise a mapping relationship field. The content in the mapping relationship field is used to indicate the first mapping relationship. For example, the content in the mapping relationship field can comprise original information of the first mapping relationship, or can comprise compressed first mapping relationship. For another example, the first information can comprise an update type field and a mapping relationship field. The update type field is used to indicate adding the mapping relationship, and the content in the mapping relationship field is used to indicate the first mapping relationship.

[0217] Based on the above scheme, the first information can indicate adding the mapping relationship, and the maintenance of the mapping relationship by the first device is realized.

[0218] FIG. 7 is a schematic diagram of determining a mapping relationship according to an embodiment of the present application. FIG. 7 is only an example and does not constitute a limitation on the present application. The first mapping relationship in the present application can be determined in the manner shown in FIG. 7, or can be determined in other manners.

[0219] For example, the first device can perform AI encoding operation on the first training data to generate S training samples. For example, training sample 1, training sample 2, …, and training sample S. S is a positive integer. Each training sample can comprise N data. In other words, each training sample can be feature data (or intermediate feature) with a length of N.

[0220] For example, the S training samples each comprise data 1 to data N. The first device can obtain the average energy of data 1, denoted as average energy 1, the average energy of data 2, denoted as average energy 2, …, and the average energy of data N, denoted as average energy N, by counting S*N data.

[0221] For example, the first device can determine the initial mapping relationship according to the average energy. For example, the first device can sort the average energy, and data with high average energy corresponds to one group, and data with low average energy corresponds to another group.

[0222] Optionally, at least two of the M groups have different priorities. Optionally, a group corresponding to data with a high average energy has a higher priority than a group corresponding to data with a low average energy.

[0223] In some possible implementation, the initial mapping relationship can be the first mapping relationship. In some other possible implementation, the initial mapping relationship can further optimize to generate the first mapping relationship. The optimization scheme will be described later, and will not be described here.

[0224] Optionally, the first information is used to indicate deleting the third mapping relationship. Illustratively, the third mapping relationship can indicate a corresponding relationship between N’ data and M’ groups, N’ being a positive integer, and M’ being an integer greater than 1 and less than N’. Wherein, N’ can be the same as N, or different from N. M’ can be the same as M, or different from M.

[0225] For example, the first information can include an index of the third mapping relationship. In this way, the second device can delete the third mapping relationship corresponding to the index according to the indication of the first information. For example, the first information can include an update type field and a number field. Wherein, the content in the update type field can be used to indicate deleting the mapping relationship. The content in the number field can be used to indicate the index of the third mapping relationship. In this way, the second device can delete the third mapping relationship according to the first information.

[0226] Optionally, the first information is used to indicate adding the first mapping relationship and deleting the third mapping relationship. For example, the first information can include an update type field, a number field and a mapping relationship field. Wherein, the content in the update type field is used to indicate adding the mapping relationship indicated by the mapping relationship field, and deleting the mapping relationship indicated by the number field. Or, the content in the update type field is used to indicate adding the mapping relationship and deleting the mapping relationship. The content in the number field is used to indicate the index of the third mapping relationship. The content in the mapping relationship field is used to indicate the first mapping relationship. In this way, the first information can add the first mapping relationship while deleting the third mapping relationship. The above scheme can realize multiple types of maintenance for multiple mapping relationships in one signaling, reduce the time delay of signaling transmission, and thus reduce the maintenance time delay of the mapping relationship.

[0227] Optionally, the index of the first mapping relationship is the same as the index of the third mapping relationship. In this way, the added first mapping relationship can use the index of the deleted third mapping relationship. The above scheme can also be understood as replacing the third mapping relationship with the first mapping relationship.

[0228] In some possible implementation, the first information is used to indicate updating the second mapping relationship to the first mapping relationship. In some possible implementation, S460 includes: S464, the second device updates the second mapping relationship to the first mapping relationship according to the first information.

[0229] It can be understood that the second mapping relationship can be a mapping relationship that the second device already has. Exemplarily, the second mapping relationship can be used to indicate a correspondence between N" data and M" groups, N" being a positive integer, and M" being an integer greater than 1 and less than N". Wherein, N" can be the same as N, or can be different from N. M" can be the same as M, or can be different from M.

[0230] Optionally, the first mapping relationship and the second mapping relationship are different. For example, the grouping manners indicated by the first mapping relationship and the second mapping relationship are different.

[0231] Optionally, the first information indicates all information of the first mapping relationship. For example, the first information can indicate original information of the first mapping relationship, or indicate the first mapping relationship after compression. Optionally, the first information indicates change information or difference information. The change information or difference information can indicate parts of the first mapping relationship that are different from the second mapping relationship.

[0232] In some examples, P group sequence numbers in the N group sequence numbers of the first group sequence and P group sequence numbers in the N group sequence numbers of the corresponding second group sequence can be different. In other words, there are P group sequence numbers that are different between the first group sequence and the second group sequence. In other words, there are N-P group sequence numbers that are the same between the first group sequence and the second group sequence.

[0233] For example, referring to (b) in FIG. 6, X0 can represent the second group sequence, and X1 can represent the first group sequence. Wherein, 3 group sequence numbers in 6 group sequence numbers of X1 are different from 3 group sequence numbers in 6 group sequence numbers of the corresponding X0. For example, the first, fourth and sixth group sequence numbers between X1 and X0 are different.

[0234] In some possible implementation manners, the first information includes first indication information.

[0235] Wherein, the first indication information can be used to indicate P group sequence numbers in the N group sequence numbers of the first group sequence, P can be a positive integer less than or equal to N. In other words, the first indication information can be used to indicate P group sequence numbers in the first group sequence.

[0236] In some examples, the first indication information indicates P group sequence numbers in the first group sequence in the form of difference information. For example, referring to (b) in FIG. 6, the first indication information can indicate X2, wherein X2=X1-X0. In this way, the first indication information indicates 3 group sequence numbers in X1 that are different from X0 in the form of difference information. The difference information can also be in other forms, which are not limited by the present application.

[0237] In some examples, the first indication information indicates P packet sequence numbers in the first packet sequence in the form of change information. For example, referring to (b) in FIG. 6, the first indication information can indicate X3, where X3 only represents the packet sequence numbers that are different from X0 and X1. In this way, the first indication information indicates, in the form of change information, the 3 packet sequence numbers in X1 that are different from X0. The change information can also be in other forms, which are not limited in the present application. For example, the change information can be in the form of information indicating the changed packet sequence numbers and information indicating the positions where the changes occur. For example, X3 can be in the form of [1, 3, 2] and [1, 0, 0, 1, 0, 1] (or [1, 4, 6]). Where [1, 3, 2] represents the 3 changed packet sequence numbers, and [1, 0, 0, 1, 0, 1] (or [1, 4, 6]) represents that the first, fourth and sixth packet sequence numbers in the first packet sequence are changed.

[0238] (b) in FIG. 6 is exemplary only, and the packet sequence in the embodiments of the present application can include more or fewer packet sequence numbers. For example, the packet sequence in the embodiments of the present application can include N packet sequence numbers, and the N packet sequence numbers have M values in common.

[0239] The present application does not limit the specific name of the first indication information, which can be referred to as change information, difference information, or other names.

[0240] In some possible implementations, S464 includes: according to the first indication information, the second device updates P packet sequence numbers in the N packet sequence numbers in the second packet sequence to P packet sequence numbers in the N packet sequence numbers in the first packet sequence.

[0241] In some examples, the second device determines the first packet sequence according to the first indication information and the second packet sequence. For example, the first indication information can indicate X2. In this way, the second device can add X2 to X0 to obtain X1, thereby updating X0 to X1.

[0242] In some examples, the second device determines the first packet sequence according to the first indication information and the second packet sequence. For example, the first indication information can indicate X2. In this way, the second device can add X2 to X0 to obtain X1, thereby updating X0 to X1.

[0243] For example, the first information can include an update type field, a number field and a mapping relationship field. The content in the update type field can indicate that the mapping relationship is modified (or updated). The content in the number field can indicate the index of the second mapping relationship. The content in the mapping relationship field can indicate the first mapping relationship.

[0244] Based on the above scheme, the first indication information can indicate the part of the first packet sequence that is different from the second packet sequence. In this way, the receiving end of the first information can update the part of the second packet sequence that is different from the first packet sequence according to the first indication information, thereby realizing the update of the mapping relationship with a small amount of signaling.

[0245] In some possible implementation manners, S440 includes: S444, the first device updates the second mapping relationship to the first mapping relationship according to the first training data. In other words, the first device determines the first mapping relationship according to the first training data and the second mapping relationship.

[0246] FIG. 8 is a schematic flowchart of a mapping relationship determination method 800 provided by an embodiment of the present application. FIG. 8 is merely an example and does not constitute a limitation on the present application. For example, the first mapping relationship in the present application can be determined by applying the method 800 or by other methods. In some other possible implementation manners, the first mapping relationship is newly generated and is not obtained based on an existing mapping relationship update. The method 800 will be introduced below in combination with FIG. 8.

[0247] S810, the first device selects the second mapping relationship.

[0248] In some other possible implementation manners, the method 800 can be combined with the scheme shown in FIG. 7. S810 can be replaced by: the first device determines an initial mapping relationship. In the case where there are multiple initial mapping relationships, the first device can select an initial mapping relationship with the best performance from the multiple initial mapping relationships.

[0249] S820, the first device fine tunes the second mapping relationship.

[0250] For example, the first device can adjust some factors in the second mapping relationship according to certain rules. For example, the value of the packet sequence number, the number of the packet sequence number, or the position of the packet sequence number, etc.

[0251] In some other possible implementation manners, the method 800 can be combined with the scheme shown in FIG. 7. S820 can be replaced by: the first device fine tunes the initial mapping relationship.

[0252] S830, the first device determines whether the fine-tuned second mapping relationship has a performance increase.

[0253] Optionally, in the case where the performance increases, S840 is performed. Optionally, in the case where the performance does not increase, S850 is performed.

[0254] In some possible implementation manners, the method 800 can be combined with the scheme shown in FIG. 7. S830 can be replaced by: the first device determining whether the performance of the fine-tuned initial mapping relationship increases.

[0255] S840, the first device updating the second mapping relationship.

[0256] For example, the first device can randomly partially exchange different grouping elements in each iteration based on an algorithm such as a genetic algorithm or reinforcement learning, and update the mapping relationship when performance increase is detected.

[0257] In some possible implementation manners, the method 800 can be combined with the scheme shown in FIG. 7. S840 can be replaced by: the first device updating the initial mapping relationship.

[0258] S850, the first device determining whether the iteration converges.

[0259] Optionally, in the case of convergence of the iteration, S860 is performed. Optionally, in the case of non-convergence of the iteration, S820 is performed.

[0260] S860, the first device determining the first mapping relationship.

[0261] For example, when the maximum number of iterations is reached or performance no longer increases, it is determined that convergence has been achieved, and the first mapping relationship updated based on the second mapping relationship is output.

[0262] For example, when the maximum number of iterations is reached or performance no longer increases, it is determined that convergence has been achieved, and the first mapping relationship updated based on the initial mapping relationship is output.

[0263] Next, possible sources of the first training data in the method 400 are described in combination with FIG. 4.

[0264] In some possible implementation manners, the method 400 further includes: S430, the first device collecting the first training data.

[0265] The first device collecting the first training data can be that the first device obtains the training data locally, or that the first device receives the training data from another device, or that the first device obtains the training data locally and receives the training data from another device.

[0266] The first device can be a device with data collection capability. For example, the first device can be applied to a terminal device.

[0267] In some possible implementation manners, S430 includes: the first device collecting the first training data according to a third parameter.

[0268] Optionally, the third parameter can indicate a number of mapping relationships included in the first mapping relationship. For example, the first mapping relationship includes Q mapping relationships, where Q can be a positive integer. The third parameter can be Q. In this way, the first training data collected by the first device can include Q pieces of training data. Exemplarily, the Q pieces of training data can be collected in different scenarios, so as to improve the training effect. However, the present application is not limited thereto, and the Q pieces of training data can also be collected in the same scenario.

[0269] Optionally, the third parameter can indicate a size of the mapping relationship set. For example, the mapping relationship set includes Q mapping relationships. In this way, the first training data collected by the first device can include Q pieces of training data. The first device can determine the Q mapping relationships from the Q pieces of training data in the first training data, where the Q mapping relationships can include the first mapping relationship. For example, the first mapping relationship can include one or more mapping relationships of the Q mapping relationships.

[0270] In some possible implementation manners, the method 400 further includes: S420, the first device receives the first training data.

[0271] In some examples, S420 includes: the first device receives the first training data from the second device. Correspondingly, the second device sends the first training data to the first device. For example, in the case where the second device is a device with data collection capability (for example, a terminal device), the first device can receive the first training data from the second device. Optionally, the second device collects the first training data.

[0272] In other examples, S420 includes: the first device receives the first training data from the third device. Correspondingly, the third device sends the first training data to the first device. Wherein, the third device can be another device different from the first device and the second device. For example, in the case where the third device is a device with data collection capability (for example, a terminal device), the first device can receive the first training data from the third device. Optionally, the third device collects the first training data.

[0273] FIG. 9 is a schematic flowchart of a training method 900 and another training method 950 provided by the embodiments of the present application. FIG. 9 is described by taking the interaction between a UE and a BS as an example. Those skilled in the art can understand that FIG. 9 is only an example and does not constitute a limitation on the present application.

[0274] (a) of FIG. 9 shows a schematic flowchart of the method 900. In the method 900, the data is collected by the UE, which can save the overhead of transmitting the training data set. The optional operations in the method 900 are shown in dashed lines in the figure. The method 900 is described below in conjunction with (a) of FIG. 9.

[0275] S902, the UE and the BS configure a size of an initial mapping relationship set.

[0276] For example, in a case where the mapping relationship is indicated by a packet sequence (or referred to as a dropping pattern), the initial mapping relationship set can also be referred to as an initial pattern set.

[0277] The size of the initial mapping relationship set can be indicated by the third parameter Q described above. For ease of description, the size of the initial mapping relationship set is denoted as Y below, Y being a positive integer. Y can represent that the initial mapping relationship set trained includes Y mapping relationships.

[0278] Optionally, the size of the initial mapping relationship set can also be pre-configured or pre-defined.

[0279] S904, the UE collects first training data.

[0280] For example, the size of the initial mapping relationship set is Y, and the UE can collect Y portions of training data. In some examples, the first training data is the Y portions of training data. In other examples, the first training data is one or more portions of the Y portions of training data. One portion of training data can include one or more data.

[0281] In some possible implementation manners, S904 can include S430. In this way, the first device can be the UE in the method 900.

[0282] S906, the UE determines an initial mapping relationship set according to the first training data.

[0283] For example, the UE can use the configured AI encoder and AI decoder to train the mapping relationship according to the Y portions of training data. Illustratively, the initial mapping relationship set determined by the UE can include Y mapping relationships. In some examples, the first mapping relationship is the Y mapping relationships. In other examples, the first mapping relationship is one or more mapping relationships of the Y mapping relationships.

[0284] In some possible implementation manners, S906 can include S420. In this way, the first device can be the UE in the method 900.

[0285] S908, the UE sends the initial mapping relationship set to the BS. Correspondingly, the BS receives the initial mapping relationship set from the UE.

[0286] In some possible implementations, S908 can include S450. In this way, the first device can be the UE in method 900, and the second device can be the BS in method 900. The first mapping relationship can be the initial mapping relationship set. The first mapping relationship can also be one or more mapping relationships in the initial mapping set. The initial mapping set can be indicated by the first information.

[0287] In some possible implementations, the UE can send the original information of the initial mapping relationship set to the BS. In some other possible implementations, the UE can send the compressed code stream of the initial mapping relationship set to the BS.

[0288] Figure 9(b) shows a schematic flowchart of method 950. In method 950, the BS performs the training, and the UE can be avoided to perform the training, thereby reducing the power consumption of the UE. The optional operations in method 950 are shown in dashed lines in the figure. Method 950 is described below in conjunction with Figure 9(b).

[0289] S952, the UE and the BS configure the size of the initial mapping relationship set.

[0290] The description of S952 above can be referred to the previous description, for example, the description of S902, which is not repeated here.

[0291] S954, the UE collects the first training data.

[0292] The description of S954 above can be referred to the previous description, for example, the description of S904, which is not repeated here.

[0293] S956, the UE sends the first training data to the BS. Correspondingly, the BS receives the first training data from the UE.

[0294] In some examples, the UE can send the original information of the first training data to the BS. In some other examples, the UE can send the compressed first training data to the BS. For example, the UE can compress the original information of the first training data. After receiving the compressed first training data, the BS can decompress to obtain the original information of the first training data.

[0295] In some possible implementations, S956 can include S420. In this way, the first device can be the BS in method 900. Optionally, the second device can be the BS in method 900. Optionally, the third device can be the BS in method 900.

[0296] S958, the BS determines the initial mapping relationship set according to the first training data.

[0297] For example, the BS can train the mapping relationship according to Y portions of training data, using the configured AI encoder and AI decoder. Illustratively, the initial mapping relationship set determined by the BS can include Y mapping relationships. In some examples, the first mapping relationship is one of the Y mapping relationships. In other examples, the first mapping relationship is one or more of the Y mapping relationships.

[0298] In some possible implementation, S958 can include S420. In this way, the first device can be the BS in the method 900.

[0299] S960, the BS sends the initial mapping relationship set to the UE. Correspondingly, the UE receives the initial mapping relationship set from the BS.

[0300] In some possible implementation, S960 can include S450. In this way, the first device can be the BS in the method 900, and the second device can be the UE in the method 900.

[0301] In some possible implementation, the BS can send the original information of the initial mapping relationship set to the UE. In other possible implementation, the BS can send the compressed code stream of the initial mapping relationship set to the UE.

[0302] In some possible implementation, S444 includes: in a case where the first condition is met, the first device updates the second mapping relationship to the first mapping relationship according to the first training data. In other words, the first device performs S444 and any example implementation of S444 in a case where the first condition is met.

[0303] Optionally, the first condition includes at least one of the following conditions: a predetermined time is reached; the first device receives indication information for updating the second mapping relationship; or the first parameter meets a second condition.

[0304] As an example, the first device performs S444 in a case where a predetermined time is reached. That is, the predetermined time can indicate a time for updating the second mapping relationship. As another example, the first device performs S444 in a case where the first parameter meets a second condition, and / or the first device receives indication information for updating the second mapping relationship, and a predetermined time is reached. That is, reaching the predetermined time alone is not enough to trigger the first device to perform S444. For example, the first device can determine whether other conditions in the first condition are met at the predetermined time. In this way, the predetermined time can indicate a time for determining whether to update the second mapping relationship.

[0305] The application does not limit the configuration form of the predetermined time. In one aspect, the predetermined time can be pre-configured or pre-defined, or can be configured or indicated by the network device, and the application does not limit this. In another aspect, the predetermined time can be configured in a periodic form or dynamically indicated, and the application does not limit this.

[0306] In some examples, the predetermined time can be configured in a periodic form. For example, the first device can pre-configure, pre-define or update the period configured by the network device. In this way, the predetermined time is reached every time the updated period is passed. Illustratively, the network device can achieve the configuration of the updated period by sending radio resource control (RRC) signaling. For example, the second device can be the network device.

[0307] In other examples, the predetermined time can be configured in a semi-static periodic form. For example, the first device can pre-configure, pre-define or update the period configured by the network device. On this basis, after the first device receives the trigger information, the updated period can be activated. In turn, the predetermined time is reached every time the updated period is passed. In other words, in the case where the first device does not receive the trigger information, the updated period can not take effect. Illustratively, the trigger information can be medium access control (MAC) signaling. Illustratively, the MAC signaling can be sent by the network device. For example, the second device can be the network device.

[0308] In still other examples, the predetermined time can be configured in a dynamic form. For example, the first device can receive second indication information, which can be used to indicate the predetermined time, or to indicate a predetermined time length, and the time after the predetermined time length is passed is the predetermined time. Illustratively, the second indication information can be carried in a physical downlink control channel (PDCCH). Illustratively, the second indication information can be downlink control information (DCI). However, the application does not limit this, and the second indication information can also be carried in uplink or sidelink signaling or channels.

[0309] For ease of description, the above-mentioned indication information for updating the second mapping relationship is referred to as third indication information.

[0310] As an example, the first device performs S444 in a case that the first device receives third indication information. That is, the third indication information can indicate the first device to update the second mapping relationship immediately (or within a certain time delay). As another example, the first device performs S444 in a case that the first parameter satisfies the second condition and the first device receives the third indication information. That is, the first device receiving the third indication information is not enough to trigger the first device to perform S444. For example, the first device can determine whether other conditions in the first condition are satisfied in a case that the first device receives the third indication information. In this way, the third indication information is used to determine whether to update the second mapping relationship.

[0311] Optionally, the third indication information is carried in the signaling configured in the dynamic manner. For example, the third indication information can be carried in a PDCCH. As another example, the third indication information can be a DCI. The third indication information can dynamically indicate a predetermined time point.

[0312] Optionally, the first parameter is used to indicate an evaluation index of the second mapping relationship. For example, the first parameter includes at least one of the following: MSE, VD, GCS, PSNR, CD, or at least one of downstream task accuracy.

[0313] In some examples, the first parameter can be determined by the first device. In other examples, the first device can receive information indicating the first parameter. For example, RRC signaling can be used to indicate the first parameter. Optionally, the information indicating the first parameter can be periodically sent to the first device.

[0314] The present application does not limit the specific name of the first parameter, and the first parameter can also be referred to as an evaluation parameter, evaluation data, or other names.

[0315] Optionally, the second condition can indicate a relationship between the first parameter and a preset threshold. For example, the first parameter is greater than or equal to the preset threshold. As another example, the first parameter is less than or equal to the preset threshold. As another example, the first parameter is greater than the preset threshold; the first parameter is equal to the preset threshold; or, the first parameter is less than the preset threshold.

[0316] The present application does not limit the specific name of the preset threshold. For example, the preset threshold can be referred to as a decision threshold, a condition threshold, or other names.

[0317] For example, the first parameter can include MSE, GCS, or CD, etc. for the second mapping relationship used for compressive sensing related data. For example, MSE can be used as an evaluation index of electromagnetic signal data, etc. As another example, GCS can be used as an evaluation index of channel response related sensing data, etc. As another example, CD can be used as an evaluation index of patch data, or point cloud data, etc.

[0318] Exemplarily, for the second mapping relationship for compressing AI-related data, the first parameter can include MSE, or downstream task accuracy, etc. For example, the MSE can be used to evaluate the accuracy of the AI-related data itself. For another example, the downstream task accuracy can be used to evaluate the performance of the AI task.

[0319] Exemplarily, for the second mapping relationship for compressing channel-related data, the first parameter can include MSE, or GCS, etc.

[0320] The second mapping relationship update in the case where the first parameter satisfies the second condition can be understood as updating the mapping relationship in the case where the performance of a mapping relationship does not satisfy the condition. Therefore, the above scheme can update on demand, reduce unnecessary updates, and thus save the training overhead introduced by updating.

[0321] In some possible implementation ways, the method 400 further includes: S410, the first device receives second information. Optionally, S410 includes: the first device receives the second information from the second device. Correspondingly, the second device sends the second information to the first device. Optionally, S410 includes: the first device receives the second information from the fourth device. Correspondingly, the fourth device sends the second information to the first device.

[0322] The fourth device can be another device different from the first device and the second device. For example, the fourth device can be a network device.

[0323] Optionally, the second information is used to indicate the predetermined moment and / or the first parameter. Optionally, the second information is further used to indicate the second condition. For example, the second information can be used to indicate a preset threshold. Optionally, the second information is used to indicate the first condition. Exemplarily, the second information can be carried in RRC signaling.

[0324] Optionally, the second device sends third indication information to the first device. Optionally, the fourth device sends the third indication information to the first device.

[0325] In some possible implementation ways, the method 400 further includes: S470, the first device sends or receives third information.

[0326] Optionally, the third information is used to indicate deleting the first mapping relationship.

[0327] For example, the first device can send the third information to the second device. For another example, the first device can receive the third information from the second device; correspondingly, the second device can send the third information to the first device. For another example, the first device can receive the third information from the fourth device; correspondingly, the fourth device can send the third information to the first device.

[0328] The specific name of the third information is not limited in the present application. The third information can also be referred to as update information, indication information or other names.

[0329] Based on the above scheme, the third information can indicate to delete the mapping relationship, thereby realizing the maintenance of the mapping relationship by the first device.

[0330] In some possible implementation manners, the third information includes the first index, and the first index indicates the first mapping relationship.

[0331] The first index can be an index of the first mapping relationship. The specific name of the first index is not limited in the present application, and the first index can also be referred to as a first number, a first identifier or other names.

[0332] The receiving end of the third information can determine the first mapping relationship according to the first index in the third information. Then the first mapping relationship can be deleted.

[0333] Exemplarily, the third information can include an update type field and a number field. The content in the update type field can indicate deletion. The content in the number field can indicate the first index. Exemplarily, the third information can include an update type field and a mapping relationship field. The content in the update type field can indicate deletion. The content in the mapping relationship field can indicate the original information of the first mapping relationship or the compressed first mapping relationship.

[0334] FIG. 10 is a schematic flowchart of another communication method 1000 provided by an embodiment of the present application. The method 1000 enables the AI encoder or the AI decoder to be more suitable for the mapping relationship. The method 1000 can be combined with any embodiment of the method 400. The optional operations in the method 1000 are indicated by dashed lines in FIG. 10. The method 1000 will be described below in combination with FIG. 10.

[0335] In S1030, the first device determines a second parameter according to second training data.

[0336] The second training data can be the same as the first training data. The second training data can also be different from the first training data. Exemplarily, the second training data can include initial mapping relationship, feature data, first training data, gradient information, decoder training data, encoder training data or other data.

[0337] Optionally, the second parameter is used to update the AI encoder or the AI decoder. For example, the AI encoder can perform data compression according to the second parameter to obtain N data. For another example, the AI decoder can process the N data according to the second parameter to obtain reconstructed data. The second parameter can also be referred to as a parameter of the AI encoder, a parameter of the AI decoder or other names, which are not limited in the present application.

[0338] S1040, the first device sends fourth information to the second device. Correspondingly, the second device receives the fourth information from the first device.

[0339] Optionally, the fourth information is used to indicate the second parameter.

[0340] The fourth information can be direct indication information, for example, the fourth information can include the second parameter. The fourth information can be indirect indication information. For example, the second device can determine the second parameter according to the fourth information.

[0341] Based on the above scheme, the first device can determine the parameter for updating the AI encoder or the AI decoder according to the training data. The above scheme can enable the AI encoder or the AI decoder to compress or process more suitable for the above mapping relationship. For example, in the scenario where the data of part of the groups in the M groups is selected for transmission by using the mapping relationship, the N data generated by the updated AI encoder can be adapted to the mapping relationship. For example, in the scenario where the data of part of the groups in the M groups received is processed by using the mapping relationship, the updated AI decoder can process the N data recovered by the mapping relationship, so as to further obtain the reconstructed data according to the N data.

[0342] In some possible implementation manners, the method 1000 further includes: S1020, the first device collects second training data. Alternatively, the first device determines the second training data.

[0343] The first device collects the second training data can be that the first device obtains the training data locally from the first device, or that the first device receives the training data from other devices, or that the first device obtains the training data locally from the first device and receives the training data from other devices.

[0344] The first device can be a device with data collection capability. Illustratively, the first device can be applied to a terminal device.

[0345] In some possible implementation manners, S1020 includes: the first device collects the second training data according to a third parameter. For details, refer to the related description of S430, which will not be described here.

[0346] In some possible implementation manners, the method 1000 further includes: S1010, the first device receives second training data.

[0347] Optionally, S1010 includes that the first device receives second training data from a second device. Correspondingly, the second device sends the second training data to the first device. As an example, the second device can be a device with data collection capability. For example, the second device can be a terminal device. As another example, the second device can be a device with training capability, for example, the second device can be a terminal device or a network device. Illustratively, the second training data can be gradient information.

[0348] Optionally, S1010 includes that the first device receives second training data from a fifth device. Correspondingly, the fifth device sends the second training data to the first device.

[0349] The fifth device can be another device different from the first device and the second device. For example, the fifth device can be a network device or a terminal device.

[0350] In some possible implementation manners, S1030 includes that, in a case where the first condition is met, the first device determines the second parameter according to the second training data. That is, the first condition can be used to trigger retraining of the mapping relationship, or can be used to trigger retraining of the second parameter.

[0351] FIG. 11 is a schematic flowchart of another training method 1100 provided by an embodiment of the present application. The method 1100 can enable the communication devices of multiple sides to perform parameter adjustment through multi-side joint training, and achieve joint optimization. The method 1100 can be combined with any embodiment of the method 400 or the method 1000. The optional operations in the method 1100 are represented by dashed lines in FIG. 11. The method 1100 will be described below in combination with FIG. 11.

[0352] S1110, the UE and the BS configure a size of the initial mapping relationship set.

[0353] The description of S1110 above can be referred to the foregoing description, for example, the description of S902, which will not be described herein again.

[0354] S1120, the UE collects first training data.

[0355] The description of S954 above can be referred to the foregoing description, for example, the description of S904, which will not be described herein again.

[0356] S1130, the UE sends the first training data to the BS. Correspondingly, the BS receives the first training data from the UE.

[0357] The description of S1130 above can be referred to the foregoing description, for example, the description of S956, which will not be described herein again.

[0358] S1140, the UE determines an initial mapping relationship set according to the first training data.

[0359] The description of S1140 can be referred to the foregoing description, for example, the description of S906, which is not repeated here.

[0360] S1150, the UE sends the initial mapping relationship set and the feature data to the BS. Correspondingly, the UE receives the initial mapping relationship set and the feature data from the BS.

[0361] Exemplarily, the first device can be the BS in S1150, so that the initial mapping relationship set and the feature data can be regarded as second training data.

[0362] The initial mapping relationship set and the feature data can send original information or compressed information, which is not limited in the present application.

[0363] The initial mapping relationship set can include at least one mapping relationship. Each mapping relationship in the at least one mapping relationship can correspond to a set of feature data. Wherein, a set of feature data can be understood as the aforementioned N data (or N-dimensional data). Exemplarily, the initial mapping relationship set can include 10 mapping relationships, so that the UE can correspondingly send 10 sets of feature data to the BS. Wherein, each set of feature data can include N-dimensional data.

[0364] In some possible implementation manners, the BS can send the original information of the initial mapping relationship set to the UE. In other possible implementation manners, the BS can send the compressed code stream of the initial mapping relationship set to the UE.

[0365] The present application does not limit the execution order of the above actions. For example, S1130 can also be executed after S1150. For another example, S1130 and S1150 can be executed simultaneously.

[0366] S1160, the BS updates the AI decoder according to the initial mapping relationship set, the feature data and the first training data.

[0367] Exemplarily, the first device can be the BS in S1160, so that S1160 can include: the first device determines a second parameter according to the second training data, that is, S1030. Wherein, the second parameter is used to update the AI decoder. Wherein, the second training data includes the initial mapping relationship set, the feature data and the first training data.

[0368] Exemplarily, S1030 can include that the first device determines the second parameter according to the second training data and the first training data. S1160 can include that the first device determines the second parameter according to the second training data and the first training data. The second training data includes the initial mapping relationship set and the feature data.

[0369] S1170, the BS sends gradient information to the UE. Correspondingly, the UE receives the gradient information from the BS.

[0370] The gradient information can also be called reverse gradient information or other names. In the process of training the neural network or AI model, the above-mentioned BS can take the partial derivative of each model parameter to obtain the gradient information. The above-mentioned scheme can also be called gradient back propagation.

[0371] Exemplarily, the first device can be the UE in S1170, so that the second training data can include the gradient information.

[0372] S1180, the UE updates the AI encoder according to the gradient information.

[0373] Exemplarily, the first device can be the UE in S1180, so that S1180 can include that the first device determines the second parameter according to the second training data, that is, S1030. The second parameter is used to update the AI encoder. The second training data includes the gradient information.

[0374] In some possible implementation manners, S1140 to S1180 can be iterated for multiple times.

[0375] In the method 1100, the UE trains first and the BS trains later. However, the present application is not limited thereto, and the BS can train first and the UE trains later. Details are not described herein again.

[0376] FIG. 12 is a schematic flowchart of a training method 1200 and a training method 1250 provided by an embodiment of the present application. The method 1200 and the method 1250 can realize joint training of devices of different manufacturers through multi-side joint training. The method 1200 or the method 1250 can be combined with any embodiment of the method 400, the method 1000 or the method 1100. The optional operations in the method 1200 and the method 1250 are represented by dashed lines in FIG. 12. The method 1200 and the method 1250 are described below in combination with FIG. 12.

[0377] S1202, the UE and the BS configure the size of the initial mapping relationship set.

[0378] The description of S1202 can be referred to the foregoing description, for example, the description of S902, which is not described herein again.

[0379] S1204, the UE collects first training data.

[0380] The description of S1204 above can be referred to the previous description, for example, the description of S904, which will not be repeated here.

[0381] S1206, the UE determines an initial mapping relationship set according to the first training data.

[0382] The description of S1206 above can be referred to the previous description, for example, the description of S906, which will not be repeated here.

[0383] S1208, the UE updates the AI encoder according to the first training data.

[0384] Exemplarily, the first device can be the UE in S1208, so that S1208 can include: the first device determines a second parameter according to second training data, i.e., S1030. Wherein, the second parameter is used to update the AI encoder. Wherein, the second training data includes the first training data.

[0385] S1206 and S1208 above can be performed as one operation. For example, the UE can perform joint training on the initial mapping relationship set and the AI encoder according to the first training data.

[0386] In some possible implementations, S1206 and S1208 can be iterated for multiple times.

[0387] S1210, the UE sends the initial mapping relationship set and decoder training data to the BS. Correspondingly, the UE receives the initial mapping relationship set and decoder training data from the BS.

[0388] Exemplarily, the first device can be the BS in S1210, so that the initial mapping relationship set and the decoder training data above can be regarded as second training data.

[0389] Exemplarily, the decoder training data can include feature data and original data. Wherein, the feature data can be used as the input of the AI decoder training. The original data can be used as the output of the AI decoder training. In an ideal state, the AI decoder can obtain the original data according to the feature data.

[0390] The initial mapping relationship set can include at least one mapping relationship. Each mapping relationship in the at least one mapping relationship can correspond to a set of feature data and a set of original data. The set of feature data can be understood as the aforementioned N data (or N-dimensional data). For example, the initial mapping relationship set can include 10 mapping relationships, so that the UE can correspondingly send 10 sets of feature data and 10 sets of original data to the BS. Each set of feature data can include N-dimensional data.

[0391] In some possible implementation manners, the UE can send original information of the decoder training data to the BS. In other possible implementation manners, the UE can send a compressed code stream of the decoder training data to the BS.

[0392] In S1212, the BS updates the AI decoder according to the initial mapping relationship set and the decoder training data.

[0393] For example, the first device can be the BS in S1212, so that S1212 can include: the first device determines a second parameter according to the second training data, that is, S1030. The second parameter is used to update the AI decoder. The second training data includes the initial mapping relationship set and the decoder training data.

[0394] Next, method 1250 is described according to (b) in FIG. 12.

[0395] In S1252, the UE and the BS configure the size of the initial mapping relationship set.

[0396] The description of S1252 can be referred to the foregoing description, for example, the description of S902, which is not described here again.

[0397] In S1254, the UE collects first training data.

[0398] The description of S1254 can be referred to the foregoing description, for example, the description of S904, which is not described here again.

[0399] In S1256, the UE sends the first training data to the BS. Correspondingly, the BS receives the first training data from the UE.

[0400] The description of S1256 can be referred to the foregoing description, for example, the description of S956, which is not described here again.

[0401] In S1258, the BS determines the initial mapping relationship set according to the first training data.

[0402] The description of S1258 can be referred to the foregoing description, for example, the description of S958, which is not described here again.

[0403] S1260, the BS updates the AI decoder according to the first training data.

[0404] Exemplarily, the first device can be the BS in S1260, and thus S1260 can include that the first device determines the second parameter according to the second training data, i.e., S1030. The second parameter is used to update the AI decoder. The second training data includes the first training data.

[0405] S1258 and S1260 can be performed as one operation. For example, the BS can perform joint training on the initial mapping relationship set and the AI decoder according to the first training data.

[0406] S1262, the BS sends the initial mapping relationship set and the encoder training data to the UE. Correspondingly, the BS receives the initial mapping relationship set and the encoder training data from the UE.

[0407] Exemplarily, the first device can be the UE in S1262, and thus the initial mapping relationship set and the encoder training data can be regarded as the second training data.

[0408] Exemplarily, the encoder training data can include feature data and original data. The feature data can be used as the output of the AI encoder training. The original data can be used as the input of the AI encoder training. In an ideal state, the AI encoder can obtain the feature data according to the original data.

[0409] The initial mapping relationship set can include at least one mapping relationship. Each mapping relationship in the at least one mapping relationship can correspond to a set of feature data and a set of original data. The set of feature data can be understood as the aforementioned N data (or N-dimensional data). Exemplarily, the initial mapping relationship set can include 10 mapping relationships, and thus the BS can correspondingly send 10 sets of feature data and 10 sets of original data to the UE. Each set of feature data can include N-dimensional data.

[0410] In some possible implementation manners, the BS can send the original information of the decoder training data to the UE. In other possible implementation manners, the BS can send the compressed code stream of the decoder training data to the UE.

[0411] S1264, the UE updates the AI encoder according to the initial mapping relationship set and the encoder training data.

[0412] Exemplarily, the first device can be the UE in S1264, and thus S1264 can include that the first device determines the second parameter according to the second training data, i.e., S1030. The second parameter is used to update the AI encoder. The second training data includes the initial mapping relationship set and the encoder training data.

[0413] In some possible implementation manners, the first condition can be used to trigger retraining of the mapping relationship and retraining of the second parameter.

[0414] Optionally, the method 900 further includes that the UE performs S904 in a case where a predetermined time arrives or the third indication information is received. Optionally, the method 900 further includes that the UE collects the first parameter in a case where a predetermined time arrives or the third indication information is received. Optionally, the method 900 further includes that S906 and S908 are performed in a case where the first parameter satisfies the second condition. In S906, the second mapping relationship is updated to the first mapping relationship according to the first training data. In S908, the UE sends the first information to the BS. The first information is used to indicate that the second mapping relationship is updated to the first mapping relationship.

[0415] Optionally, the method 950 further includes that the BS sends a first request to the UE in a case where a predetermined time arrives or the third indication information is received. The first request is used to request training data. Illustratively, the method 400 further includes that the first device sends a first request to the second device. The first request is used to request training data. Alternatively, the first request is used to request the first training data.

[0416] Optionally, in the method 950, the UE performs S954 in response to the first request. Optionally, the method 950 further includes that the UE collects the first parameter in response to the first request. Optionally, the method 950 further includes that the UE sends the first parameter to the BS. Correspondingly, the BS receives the first parameter from the UE. Optionally, the method 950 further includes that S958 and S960 are performed in a case where the first parameter satisfies the second condition. In S958, the second mapping relationship is updated to the first mapping relationship according to the first training data. In S960, the BS sends the first information to the UE. The first information is used to indicate that the second mapping relationship is updated to the first mapping relationship.

[0417] Optionally, the method 1200 further includes that the UE performs S1204 in a case where a predetermined time arrives or the third indication information is received. Optionally, the method 1200 further includes that the UE collects the first parameter in a case where a predetermined time arrives or the third indication information is received. Optionally, the method 1200 further includes that S1206, S1208 and S1210 are performed in a case where the first parameter satisfies the second condition. In S1206, the second mapping relationship is updated to the first mapping relationship according to the first training data. In S1210, the UE sends the first information and the decoder training data to the BS. The first information is used to indicate that the second mapping relationship is updated to the first mapping relationship.

[0418] Optionally, the method 1250 further includes: in case of reaching a predetermined time or receiving a third indication information, the BS sends a first request to the UE, the first request being used for requesting the training data. Optionally, the UE performs S1254 in response to the first request. Optionally, the method 1250 further includes: the UE collects the first parameter in response to the first request. Optionally, the method 1250 further includes: the UE sends the first parameter to the BS. Correspondingly, the BS receives the first parameter from the UE. Optionally, the method 1250 further includes: in case that the first parameter satisfies a second condition, performing S1258, S1260 and S1262. Wherein, S1258 is replaced by: updating the second mapping relationship to the first mapping relationship according to the first training data. Wherein, S1262 is replaced by: the BS sends the first information and the encoder training data to the UE. Wherein, the first information is used for indicating updating the second mapping relationship to the first mapping relationship.

[0419] In some possible implementation manners, the method 400 further includes: the first device (or the second device) transmits the M1 groups of data according to the first mapping relationship.

[0420] Wherein, the M1 groups belong to the M groups, and M1 is a positive integer less than or equal to M. The value of M1 can be determined by resource constraints. For example, when the transmission resource is small, the value of M1 can be small. This scheme can match the resource constraints. For another example, when the transmission resource is large, the value of M1 can be large. This scheme can improve the accuracy of the receiving end of the M1 groups of data recovering the N data.

[0421] Exemplarily, the first device (or the second device) can group the N data according to the first mapping relationship to obtain M groups. Further, the first device (or the second device) can select M1 groups from the M groups and output the data of the M1 groups. Optionally, the data of M-M1 groups is discarded.

[0422] Wherein, the data of the M1 groups can be output after quantization. Optionally, the number of quantization bits can be determined according to resource constraints. Optionally, the quantization configuration can be predefined by a protocol, or dynamic quantization (for example, optimizing the quantization parameter according to the distribution characteristics of the data of the M1 groups) is introduced to improve the quantization precision.

[0423] In some possible implementation, the first device (or the second device) sends fourth indication information. The fourth indication information is used to indicate that the data of the M1 groups is determined according to the first mapping relationship. For example, the fourth indication information can include an index of the first mapping relationship. The fourth indication information can be sent at the same time as the data of the M1 groups. For example, the fourth indication information and the data of the M1 groups can be carried in the same message. However, the application is not limited thereto, and the fourth indication information can be sent at different times from the data of the M1 groups.

[0424] Based on the above scheme, according to the first mapping relationship, by adjusting the parameters of M1, the demand for code rate adjustment in the AI compression scene can be met, so as to ensure the data transmission performance. In addition, the above scheme can match various types of data to be transmitted. For example, the first device can store a large number of mapping relationships with a small amount of storage space, and when the data to be transmitted changes, the first device can use a suitable mapping relationship to process the data to be transmitted, and select M1 groups for transmission.

[0425] For example, the first device can send the data of the M1 groups to the second device. However, the application is not limited thereto, and the first device can also send the data of the M1 groups to other devices. For example, the second device can send the data of the M1 groups to the first device. However, the application is not limited thereto, and the second device can also send the data of the M1 groups to other devices.

[0426] In some possible implementation, the priority of the M1 groups is higher than or equal to the priority of M-M1 groups in the M groups except the M1 groups.

[0427] For example, the smaller the group number is, the higher the priority is. For another example, the larger the group number is, the higher the priority is. In some examples, M1 = 2. Referring to FIG. 5, the data corresponding to the group numbers 1 and 2 can be used for transmission, and the data corresponding to the group numbers 3 and 4 can be discarded, i.e., not used for transmission.

[0428] As an example, the priority can be determined according to the power (or absolute value of the power) of the N data. For example, the larger the power of one or more data in the N data is, the higher the priority of the group corresponding to the data or the data is. It can be understood that the larger the power is, the higher the contribution to the decoder network is, and therefore the priority can also be higher.

[0429] As another example, the priority can be determined according to the reconstruction accuracy of the N data. For example, if one or more data in the N data have a higher degree of influence on the reconstruction accuracy, the priority of the group corresponding to the data is higher. For example, if the accuracy of the reconstructed data determined by the N data is greatly reduced after the data is set to zero, the degree of influence of the data on the reconstruction accuracy is higher.

[0430] The above scheme can use limited transmission resources to transmit more important feature data (i.e., data of the M1 groups), achieve a better signal discarding effect, and further meet the demand for code rate adjustment in an AI compression scenario, thereby ensuring data transmission performance.

[0431] In some possible implementations, the method 400 further includes: the first device (or the second device) receiving data of the M1 groups according to the first mapping relationship; and the first device (or the second device) processing the data of the M1 groups according to the first mapping relationship.

[0432] For example, the first device (or the second device) can perform dequantization on the data stream of the M1 groups to obtain the data of the M1 groups. For example, the first device (or the second device) can perform dequantization on the data stream of the M1 groups according to the indicated quantization configuration or a fixed quantization configuration (e.g., predefined by a protocol) to obtain the data of the M1 groups.

[0433] For example, the first device (or the second device) can perform dequantization on the data stream of the M1 groups to obtain the data of the M1 groups. For example, the first device (or the second device) can perform dequantization on the data stream of the M1 groups according to the indicated quantization configuration or a fixed quantization configuration (e.g., predefined by a protocol) to obtain the data of the M1 groups.

[0434] In some possible implementations, the method 400 further includes: the first device (or the second device) determining the first mapping relationship from the at least one mapping relationship according to at least one energy variance corresponding to the at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest in the at least one energy variance.

[0435] For example, the first device (or the second device) can determine the first mapping relationship from the at least one mapping relationship according to at least one energy standard deviation corresponding to the at least one mapping relationship, wherein the energy standard deviation corresponding to the first mapping relationship is the largest in the at least one energy standard deviation.

[0436] FIG. 13 is a schematic diagram of an implementation manner of determining the first mapping relationship provided in the embodiments of the present application. Assuming that there are N = 8 data, for example, data 1, data 2, …, data 8, and the data 1-8 are equally divided into M = 4 groups, then the length of each group is N / M = 2, that is, each group includes 2 data. As shown in the mapping relationship of (a) of FIG. 13, group 1 corresponds to energy 1 (which can be understood as the total energy corresponding to data 1 and data 2), group 2 corresponds to energy 2 (which can be understood as the total energy corresponding to data 3 and data 4), group 3 corresponds to energy 3 (which can be understood as the total energy corresponding to data 5 and data 6), and group 4 corresponds to energy 4 (which can be understood as the total energy corresponding to data 7 and data 8). The energy variance of energy 1-energy 4 is calculated, that is, energy variance 1 is obtained. As shown in the mapping relationship of (b) of FIG. 13, group 1 corresponds to energy 1, group 2 corresponds to energy 2, group 3 corresponds to energy 3, and group 4 corresponds to energy 4. The energy variance of energy 1-energy 4 is calculated, that is, energy variance 2 is obtained. As shown in the mapping relationship of (c) of FIG. 13, group 1 corresponds to energy 1, group 2 corresponds to energy 2, group 3 corresponds to energy 3, and group 4 corresponds to energy 4. The energy variance of energy 1-energy 4 is calculated, that is, energy variance 3 is obtained. Finally, the first device can select the mapping relationship with the largest energy variance by comparing the energy variances 1-3, and select the mapping relationship with the largest energy variance as the first mapping relationship. Assuming that the mapping relationship shown in (a) of FIG. 13 has the largest energy variance, the first device can determine to use the mapping relationship shown in (a) of FIG. 13 as the first mapping relationship, and group the N = 8 data to obtain M = 4 groups.

[0437] The above FIG. 13 is only an example given for ease of understanding, and other schemes are not excluded. Alternatively, the energy variance in the above FIG. 13 can also be replaced by the energy standard deviation, and the above calculation of the energy corresponding to the grouping can also be replaced by the calculation of the L1 norm corresponding to the grouping to implement. Alternatively, if the number of data included in each group is not equal, the energy calculated in FIG. 13 can be the average energy of the data included in each group, which is not limited in the present application.

[0438] Based on the above scheme, providing at least one mapping relationship can adapt to different AI models and / or transmission scenarios, and has higher flexibility and adaptability. In addition, by comparing the energy variances corresponding to the mapping relationships and selecting the first mapping relationship with the largest energy variance to group the N data, that is, selecting the first mapping relationship with the most uneven energy distribution of the M groups, a better grouping effect can be achieved.

[0439] Figure 14 is a schematic diagram of the overall encoding and decoding process of the AI ​​compression signal discarding scheme provided in this application embodiment. As shown in Figure 14, at the transmitting end, the input data is processed by the AI ​​encoder to obtain feature data of length N. Then, the feature data is discarded to obtain the signal to be transmitted. Finally, the signal to be transmitted is output through a wireless channel after channel coding, modulation, and other operations. The input data can be channel data, sensing data, AI training / model data, etc. Correspondingly, at the receiving end, the received data is first demodulated and channel decoded to obtain the signal to be reconstructed. Then, the signal to be reconstructed is reconstructed to obtain reconstructed feature data of length N. Finally, the reconstructed data is output through the AI ​​decoder. The AI ​​encoder or AI decoder can include, but is not limited to, CNN, recurrent neural network (RNN), or transformer models.

[0440] Figure 15 is a schematic diagram of the processing flow of the encoding unit provided in an embodiment of this application. As shown in Figure 15, the processing flow of the encoding unit includes, but is not limited to, the following steps.

[0441] S1: Recombination feature data.

[0442] For example, after receiving N data points, the encoding unit then processes the feature data. That is, the encoding unit groups the N data points, for example, dividing the N data points into M groups at equal intervals according to the mapping relationship, namely group 1, group 2, group 3, ..., group M.

[0443] Optionally, there may be one or more ways to reorganize (or group) the N data, that is, including one or more mapping relationships.

[0444] S2: Filter M1 groups.

[0445] It is understandable that the M1 groups belong to the M groups, and the priority (or importance) of the M1 groups is greater than that of the M-M1 groups, so as to reduce transmission overhead while ensuring data transmission accuracy.

[0446] For example, the encoding unit sorts the M groups in step S1 according to their importance and selects M1 groups from them. Optionally, the importance of each group can be determined by calculating the energy or L1 norm corresponding to that group. Optionally, the importance of each group can be determined by the group number.

[0447] S3: Quantization operation.

[0448] For example, for the M1 groups obtained by group filtering, the encoding unit can determine the quantization bit length according to the scheduled resources, and then perform quantization processing on the M1 groups based on the quantization bit length to obtain the data bit stream.

[0449] S4: output the signal to be sent.

[0450] Exemplarily, the signal to be sent includes a data stream and configuration information, wherein the configuration information includes fourth indication information, etc. Optionally, the configuration information further includes quantization configuration.

[0451] The present application does not limit the specific name of the encoding unit, for example, the encoding unit can also be referred to as a signal discarding module or other names.

[0452] FIG. 16 is a schematic diagram of the processing flow of the decoding unit provided by the embodiment of the present application. As shown in FIG. 16, the processing flow of the decoding unit includes the following steps.

[0453] S5: inverse quantization.

[0454] Exemplarily, the decoding unit performs inverse quantization operation on the data stream according to the quantization configuration indicated by the configuration information, to obtain the data of M1 groups.

[0455] S6: feature data recovery.

[0456] Exemplarily, the decoding unit performs zero padding operation on the M-M1 groups screened out by the first device, and then recovers the order of N data according to the fourth indication information carried in the configuration information, to obtain the reconstructed data.

[0457] The present application does not limit the specific name of the decoding unit, for example, the decoding unit can also be referred to as a feature reconstruction module or other names.

[0458] Next, in combination with FIG. 17, the specific implementation mode of the AI encoder and AI decoder involved in FIG. 14 will be described.

[0459] FIG. 17 is a structural schematic diagram of an AI encoder / decoder based on a ResNet network structure provided by the embodiment of the present application.

[0460] As shown in (a) of FIG. 17, the ResNet network structure includes 3 CNNs, wherein CNN1 and CNN2 are respectively followed by a normalization or batch normalization (BN) process and a parametric rectified linear unit (PReLU) activation function, the convolution kernel size is configured to be 3x3, the channel number is set to be 4, and the strides are set to be 2 and 1 respectively, so that the dimension reduction effect can be achieved, for example, the first and second dimensions are each reduced by half. CNN3 processes the input signal of the shortcut, which is mainly used to align the dimensions and the channel number, the convolution kernel size is 1x1, the channel number is the same as that of CNN1 or CNN2, and the stride is 2. Alternatively, the above channel number can also be set to be an integer value such as 2, 8, or 16.

[0461] As shown in (c) of FIG. 17, the AI encoder, after the input data (for example, the dimension is n1x n2x n3) is processed by two layers of ResNet network (for example, ResNet1 and ResNet2), the feature with the dimension of (n1 / 4)x(n2 / 4)x4 is obtained, then the length of the vector is n1n2 / 4 after being straightened, and then the vector is sent to the full connection network FCN with the dimension of (n1n2 / 4)xN to obtain the vector with the dimension of N, and finally the feature data with the length of N is obtained after the Sigmoid activation function, and the feature data is output. Generally, the feature data can be normalized to 0-1. channel

[0462] As shown in (b) of FIG. 17, the ResNet up-sampling network structure includes 3 deconvolutional neural networks (DCNNs), wherein DCNN1 and DCNN2 are respectively followed by a BN process and a PReLU activation function, the convolution kernel size is configured to be 3x3, the channel number is set to be nchannel, and the up-sampling multiples are set to be 2 and 1 respectively, so that the dimension increase effect can be achieved, for example, the first and second dimensions are each increased by one time. DCNN3 processes the input signal of the shortcut, which is mainly used to align the dimensions and the channel number, the convolution kernel size is 1x1, the channel number is the same as that of DCNN1 or DCNN2, and the up-sampling multiple is 2.

[0463] ​The AI decoder shown in Fig. 17(d) receives feature data with a length of N, and then performs FCN processing with a length of N x (n1n2 / 4) to obtain a vector with a length of n1n2 / 4; the vector is reshaped to obtain a signal with a dimension of (n1 / 4) x (n2 / 4) x 4, and then processed by two layers of ResNet up-sampling network (for example, ResNet up-sampling 1 and ResNet up-sampling 2) to obtain reconstructed data (with a dimension of n1x n2x n channel ) after processing.

[0464] In some possible implementation manners, after the first device (or the second device) transmits the M1 groups of data according to the first mapping relationship, the method 400 further includes: the first device (or the second device) transmits M2 groups of data according to the first mapping relationship, the M2 groups belong to groups other than the M1 groups in the M groups, and M2 is a positive integer less than or equal to M-M1.

[0465] That is, after the M1 groups are transmitted, the data of the remaining M-M1 groups can not be discarded, but the data of part or all of the M-M1 groups is transmitted in the second round.

[0466] In some possible implementation manners, the method 400 further includes: the first device (or the second device) receives the M1 groups of data; the first device (or the second device) receives the M2 groups of data; and the first device (or the second device) processes the M1 groups of data and the M2 groups of data according to the first mapping relationship.

[0467] For example, after receiving the M1 groups of data, the receiving end can not be able to recover the received data into N data according to the first mapping relationship due to channel quality or its own capability. Optionally, the receiving end sends fifth indication information to the sending end of the M1 groups of data, and the fifth indication information is used to indicate data recovery failure or to indicate that the sending end sends data related to the M1 groups.

[0468] Optionally, in response to the fifth indication information, the sending end sends the M1 groups of data to the receiving end again, that is, retransmits the M1 groups of data. Optionally, in response to the fifth indication information, the sending end transmits the M2 groups of data according to the first mapping relationship.

[0469] For example, the receiving end described above can be the first device, the second device or another device. The sending end described above can be the first device, the second device or another device.

[0470] Based on the above scheme, the data in the M-M1 groups can continue to be transmitted according to the first mapping relationship, thereby improving the accuracy of the receiving end obtaining N data according to the first mapping relationship.

[0471] In some possible implementation manners, the priority of the M2 groups is higher than or equal to the priority of M-M1-M2 groups in the M groups, except for the M1 groups and the M2 groups.

[0472] That is, the priority of the M2 groups can be the highest priority in the M-M1 groups remaining after the last transmission. In some examples, M2=2. Referring to FIG. 5, after the data corresponding to packet numbers 1 and 2 (i.e., data of the M1 groups) is transmitted, the data corresponding to packet numbers 3 and 4 (i.e., data of the M2 groups) can be used for transmission.

[0473] Based on the above scheme, the sending end can select the data of the most important group in the remaining packets for transmission, thereby helping the receiving end to recover the N data.

[0474] The receiving end described above can be the receiving end of the data (for example, data of the M1 groups). The sending end described above can be the sending end of the data (for example, data of the M1 groups).

[0475] The above scheme can also be referred to as incremental transmission. The above scheme shows an example of two rounds of transmission. In this example, the first round transmits data of the M1 groups, and the second round transmits data of the M2 groups. The present application does not limit the scheme of incremental transmission to only two rounds. For example, three or more rounds of incremental transmission can be performed. For example, M2=2. Referring to FIG. 5, after the data corresponding to packet numbers 1 and 2 (i.e., data of the M1 groups) is transmitted, the data corresponding to packet number 3 (i.e., data of the M2 groups) can be used for transmission. Further, the data corresponding to packet number 4 can be used for transmission.

[0476] The following describes a device embodiment corresponding to the method embodiment of the present application. The following only briefly describes the device, and the specific implementation steps and details of the scheme can be referred to the foregoing method embodiments.

[0477] To implement each function in the method provided by the present application, the communication device can include a hardware structure and / or a software module to implement the above functions in the form of hardware structure, software module, or hardware structure plus software module. Whether a certain function in the above functions is executed in the form of hardware structure, software module, or hardware structure plus software module depends on the specific application of the technical solution and the design constraint conditions.

[0478] FIG. 18 is a schematic block diagram of a communication device 10 according to an embodiment of the present application. The communication device 10 includes a processor 11 and a communication interface 12. Optionally, the processor 11 and the communication interface 12 can be connected to each other through a bus. The communication device 10 can be a first device or a second device. Exemplarily, the first device can be a terminal device or a network device; and the second device can be a terminal device or a network device.

[0479] Optionally, the communication apparatus 10 can further comprise a memory 14. The memory 14 can include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, an erasable programmable read only memory (EPROM), a synchronous dynamic random access memory (SDRAM), a hard disk drive (HDD), a solid-state drive (SSD), or a compact disc read-only memory (CD-ROM). The memory 14 is configured to store relevant instructions and / or data. The memory 14 can be integrated with the processor 11 or separately arranged.

[0480] The processor 11 can be one or more central processing units (CPUs). In the case where the processor 11 is a CPU, the CPU can be a single-core CPU or a multi-core CPU. The processor 11 can be a signal processor, a chip, or other integrated circuits that can implement the method of the present application, or a part of the foregoing processor, chip, or integrated circuit for processing functions. In addition, the communication interface 12 can also be an input / output interface for input or output of signals or data, or an input / output circuit.

[0481] Exemplarily, the communication apparatus 10 is a first apparatus, and the processor 11 is configured to: determine, according to first training data, a first mapping relationship, the first mapping relationship being used to indicate a correspondence between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups comprising part of the N data, the N data being generated by an AI encoder; and send first information, the first information being used to indicate the first mapping relationship.

[0482] Exemplarily, the communication apparatus 10 is a second apparatus, and the processor 11 is configured to: receive first information, the first information being used to indicate a first mapping relationship, the first mapping relationship being used to indicate a correspondence between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups comprising part of the N data, the N data being generated by an AI encoder; and determine, according to the first information, the first mapping relationship.

[0483] The foregoing merely serves as an exemplary description. The communication apparatus 10 is responsible for performing the method or steps related to the first device or the second device in the foregoing method embodiments.

[0484] In a possible implementation, the communication interface 12 can be a transceiver. The transceiver can include a transmitter and a receiver, the transmitter being configured to perform a transmitting operation, and the receiver being configured to perform a receiving operation. For example, the processor 11 is configured to control the transceiver to receive and / or transmit a signal.

[0485] In a possible implementation, the communication interface 12 can also be a communication circuit, a pin, an input / output interface, a bus, or the like.

[0486] The communication apparatus 10 can include a transmitter but not a receiver. Alternatively, the communication apparatus 10 can include a receiver but not a transmitter. Specifically, whether the communication apparatus 10 includes a transmitting action and a receiving action can depend on whether the communication apparatus 10 performs the scheme described above.

[0487] The foregoing merely serves as an exemplary description. Specific contents can be referred to the contents shown in the foregoing method embodiments. The implementation of each operation in FIG. 18 can also correspond to the description of the corresponding method embodiments shown in FIGS. 4 to 17.

[0488] For example, the communication apparatus 10 can be configured to perform the schemes shown in FIGS. 4 to 17.

[0489] Exemplarily, the communication apparatus 10 is the first device, and the communication interface 12 can be configured to transmit the first information.

[0490] Exemplarily, the communication apparatus 10 is the second device, and the communication interface 12 can be configured to receive the first information.

[0491] For other implementations, refer to the detailed description of the embodiments shown in FIGS. 4 to 17. It should be understood that the specific processes in which each component performs the corresponding processes have been described in the foregoing method embodiments, and thus will not be described here for brevity.

[0492] FIG. 19 is a schematic block diagram of another communication apparatus 20 according to an embodiment of the present application. The communication apparatus 20 can be the first device or the second device, or a chip or a module of the first device or the second device, and is configured to implement the method related to the embodiments shown in FIGS. 4 to 17. For details, refer to the related description in the foregoing method embodiments.

[0493] The communication apparatus 20 includes a transceiving unit 21. The transceiving unit 21 is exemplarily described as follows.

[0494] The transceiver unit 21 can include a sending unit and a receiving unit. The sending unit is configured to perform the sending action of the communication apparatus, and the receiving unit is configured to perform the receiving action of the communication apparatus. For ease of description, the sending unit and the receiving unit are combined into one transceiver unit in the embodiments of the present application. This is uniformly described herein, and will not be described below. The transceiver unit 21 can implement the corresponding communication function. The transceiver unit 21 can also be referred to as a communication interface or a communication module.

[0495] The communication apparatus 20 can include a sending unit and not include a receiving unit. Alternatively, the communication apparatus 20 can include a receiving unit and not include a sending unit. Specifically, whether the sending action and the receiving action are included in the above-mentioned schemes performed by the communication apparatus 20 can be determined.

[0496] Exemplarily, the transceiver unit 21 is configured to send the first information and the like.

[0497] Optionally, the communication apparatus 20 further includes a processing unit 22, which is configured to perform the content related to processing, coordination and the like of the communication apparatus 20.

[0498] Exemplarily, the transceiver unit 21 is configured to receive the first information and the like.

[0499] Optionally, the communication apparatus 20 further includes a processing unit 22, which is configured to perform the content related to processing, coordination and the like of the communication apparatus 20.

[0500] The above-mentioned content is only exemplary description. The communication apparatus 20 will be responsible for performing the related methods or steps in the foregoing method embodiments.

[0501] Optionally, the communication apparatus 20 further includes a storage unit 23, which is configured to store the program or code for performing the foregoing method. Alternatively, the storage unit 23 can be configured to store instructions and / or data, and the processing unit 22 can read the instructions and / or data in the storage unit 23, so that the communication apparatus 20 implements the foregoing method embodiments. For example, the communication apparatus 20 can be configured to perform the schemes shown in FIGS. 4 to 17.

[0502] Exemplarily, the processing unit 22 can be configured to determine a first mapping relationship according to first training data, the first mapping relationship being used to indicate a corresponding relationship between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one group of the M groups including part of the N data, the N data being generated by an AI encoder; and send first information, the first information being used to indicate the first mapping relationship.

[0503] Exemplarily, the transceiving unit 21 can be configured to receive first information, the first information being used to indicate a first mapping relationship, the first mapping relationship being used to indicate a correspondence between N data and M groups, N being a positive integer, M being an integer greater than 1 and less than N, one of the M groups including part of the N data, the N data being generated by an AI encoder; and the processing unit 22 can be configured to determine the first mapping relationship according to the first information.

[0504] For other implementations, refer to the detailed description of the embodiments shown in FIGS. 4-17, which will not be repeated here. It should be understood that the specific processes of the components performing the corresponding processes described above have been described in detail in the method embodiments, and will not be repeated here for the sake of brevity.

[0505] When the communication apparatus 10 in FIG. 18 is a chip, the communication interface 12 can be a transceiver, an input / output circuit or a communication interface of the chip. The processor 11 can be an integrated processor on the chip, or a microprocessor, or an integrated circuit. The sending operation of the first apparatus or the second apparatus in the method embodiments described above can be understood as the output of the chip, and the receiving operation of the first apparatus or the second apparatus in the method embodiments described above can be understood as the input of the chip.

[0506] When the communication apparatus 20 in FIG. 19 is a chip, the transceiving unit 21 can be a transceiver, an input / output circuit or a communication interface of the chip. The processing unit 22 can be an integrated processor on the chip, or a microprocessor, or an integrated circuit. The sending operation of the first apparatus or the second apparatus in the method embodiments described above can be understood as the output of the chip, and the receiving operation of the first apparatus or the second apparatus in the method embodiments described above can be understood as the input of the chip.

[0507] The present application also provides a chip, comprising a processor, configured to invoke and run instructions stored in a memory, so that a communication apparatus installed with the chip performs the method in any of the examples described above.

[0508] The present application also provides another chip, comprising an input interface, an output interface and a processor, the input interface, the output interface and the processor being connected through internal connection paths, the processor being configured to execute code in a memory, when the code is executed, the processor is configured to perform the method in any of the examples described above. Optionally, the chip further comprises a memory, configured to store a computer program or code.

[0509] The present application also provides a processor, configured to be coupled with a memory, configured to perform the method and function related to the sensing apparatus or the communication apparatus in any of the embodiments described above, or configured to perform the method and function related to the first apparatus or the second apparatus in any of the embodiments described above.

[0510] In another embodiment of the present application, a computer program product containing computer programs or instructions is provided, when the computer program product is run on a computer, the method of the foregoing embodiments is implemented.

[0511] The present application also provides a computer program, when the computer program is run on a computer, the method of the foregoing embodiments is implemented.

[0512] In another embodiment of the present application, a computer readable storage medium is provided, the computer readable storage medium stores a computer program, when the computer program is executed by a computer, the method of the foregoing embodiments is implemented.

[0513] The present application also provides a communication system, the communication system comprises a first device and a second device. The first device and the second device are respectively used to execute the method executed by the first device and the second device in the foregoing embodiments.

[0514] Those skilled in the art can understand that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware or in combination of computer software and electronic hardware. Whether the functions are realized in hardware or software mode depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0515] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0516] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0517] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0518] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0519] The functions, if realized in the form of software functional units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the parts of the present application that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of software products. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

Claims

1. A communication method, characterized in that, The method is applied to a first device, and the method includes: Based on the first training data, a first mapping relationship is determined. The first mapping relationship is used to indicate the correspondence between N data and M groups, where N is a positive integer and M is an integer greater than 1 and less than N. One of the M groups includes a portion of the N data. The N data is generated by an artificial intelligence (AI) encoder. Send a first message, which is used to indicate the first mapping relationship.

2. The method according to claim 1, characterized in that, The first mapping relationship is indicated by the first grouping sequence, which includes N grouping numbers, each of which corresponds to one of the N data.

3. The method according to claim 1 or 2, characterized in that, The method further includes: Collect the first training data; or, Receive the first training data.

4. The method according to any one of claims 1 to 3, characterized in that, The first information is used to indicate the addition of the first mapping relationship.

5. The method according to any one of claims 1 to 3, characterized in that, Determining the first mapping relationship based on the first training data includes: Based on the first training data, the second mapping relationship is updated to the first mapping relationship.

6. The method according to claim 5, characterized in that, The first information is used to indicate that the second mapping relationship should be updated to the first mapping relationship.

7. The method according to claim 6, characterized in that, The first information includes first indication information, which is used to indicate P group numbers out of N group numbers in the first grouping sequence, where P is a positive integer less than or equal to N; wherein, P group numbers in the N group numbers of the first group sequence are different from P group numbers in the corresponding second group sequence. The first group sequence indicates the first mapping relationship, and the second group sequence indicates the second mapping relationship.

8. The method according to any one of claims 5 to 7, characterized in that, The step of updating the second mapping relationship to the first mapping relationship based on the first training data includes: If the first condition is met, the second mapping relationship is updated to the first mapping relationship based on the first training data; The first condition includes at least one of the following conditions: The scheduled time has been reached; The first device receives an instruction to update the second mapping relationship; or, The first parameter satisfies the second condition, wherein the first parameter is used to indicate the evaluation index of the second mapping relationship.

9. The method according to claim 8, characterized in that, The first parameter includes at least one of the following: Mean squared error (MSE), vertical deviation (VD), generalized cosine similarity (GCS), peak signal-to-noise ratio (PSNR), chamfer distance (CD), or downstream task accuracy, or at least one of these.

10. The method according to claim 8 or 9, characterized in that, The method further includes: Receive second information, which indicates the predetermined time, and / or the first parameter.

11. The method according to any one of claims 1 to 10, characterized in that, The method further includes: Based on the second training data, a second parameter is determined, which is used to update the AI ​​encoder or AI decoder.

12. The method according to claim 11, characterized in that, The method further includes: Collect the second training data; or receive the second training data.

13. The method according to any one of claims 1 to 12, characterized in that, The method further includes: According to the first mapping relationship, send M1 groups of data, where the M1 groups belong to the M groups, and M1 is a positive integer less than or equal to M; or, Receive the data from the M1 groups and process the data from the M1 groups according to the first mapping relationship.

14. The method according to claim 13, characterized in that, The method further includes: The first mapping relationship is determined from the at least one mapping relationship based on at least one energy variance corresponding to at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest among the at least one energy variance.

15. The method according to claim 13 or 14, characterized in that, The priority of the M1 groups is higher than or equal to the priority of the M-M1 groups other than the M1 group.

16. The method according to any one of claims 13 to 15, characterized in that, After sending M1 groups of data according to the first mapping relationship, the method further includes: According to the first mapping relationship, data in M2 groups are sent. The M2 groups belong to the groups other than the M1 groups among the M groups, and M2 is a positive integer less than or equal to M-M1.

17. The method according to claim 16, characterized in that, The priority of the M2 groups is higher than or equal to the priority of the M-M1-M2 groups other than the M1 groups and the M2 groups.

18. A communication method, characterized in that, The method is applied to a second device, and the method includes: Receive first information, the first information is used to indicate a first mapping relationship, the first mapping relationship is used to indicate the correspondence between N data and M groups, N is a positive integer, M is an integer greater than 1 and less than N, one of the M groups includes a portion of the N data, the N data is generated by an artificial intelligence AI encoder; Based on the first information, the first mapping relationship is determined.

19. The method according to claim 18, characterized in that, The first mapping relationship is indicated by the first grouping sequence, which includes N grouping numbers, each of which corresponds to one of the N data.

20. The method according to claim 18 or 19, characterized in that, The method further includes: Send first training data, which is used to determine the first mapping relationship.

21. The method according to any one of claims 18 to 20, characterized in that, Determining the first mapping relationship based on the first information includes: Based on the first information, add the first mapping relationship.

22. The method according to any one of claims 18 to 20, characterized in that, Determining the first mapping relationship based on the first information includes: Based on the first information, the second mapping relationship is updated to the first mapping relationship.

23. The method according to claim 22, characterized in that, The first information includes first indication information, which is used to indicate P group numbers out of N group numbers in the first grouping sequence, where P is a positive integer less than or equal to N; wherein, P group numbers out of the N group numbers in the first grouping sequence are different from P group numbers out of the corresponding N group numbers in the second grouping sequence. The first grouping sequence indicates the first mapping relationship, and the second grouping sequence indicates the second mapping relationship; wherein... The step of updating the second mapping relationship to the first mapping relationship based on the first information includes: According to the first instruction information, P group numbers out of N group numbers in the second group sequence are updated to P group numbers out of N group numbers in the first group sequence.

24. The method according to any one of claims 18 to 23, characterized in that, The method further includes: Send a second message, which indicates a predetermined time, and / or a first parameter; wherein, The predetermined time is used to indicate the time when to determine whether to update the second mapping relationship, which is different from the first mapping relationship; The first parameter is used to indicate the evaluation index of the second mapping relationship, and the first parameter is used to determine whether to update the second mapping relationship.

25. The method according to claim 24, characterized in that, The first parameter includes at least one of the following: Mean squared error (MSE), vertical deviation (VD), generalized cosine similarity (GCS), peak signal-to-noise ratio (PSNR), chamfer distance (CD), or downstream task accuracy, or at least one of these.

26. The method according to any one of claims 18 to 25, characterized in that, The method further includes: Send an instruction to update a second mapping relationship, which is different from the first mapping relationship.

27. The method according to any one of claims 18 to 26, characterized in that, The method further includes: Send second training data, which is used to determine a second parameter, and the second parameter is used to update the AI ​​encoder or AI decoder.

28. The method according to any one of claims 18 to 27, characterized in that, The method further includes: According to the first mapping relationship, send M1 groups of data, where the M1 groups belong to the M groups, and M1 is a positive integer less than or equal to M; or, Receive the data from the M1 groups and process the data from the M1 groups according to the first mapping relationship.

29. The method according to claim 28, characterized in that, The method further includes: The first mapping relationship is determined from the at least one mapping relationship based on at least one energy variance corresponding to at least one mapping relationship, wherein the energy variance corresponding to the first mapping relationship is the largest among the at least one energy variance.

30. The method according to claim 28 or 29, characterized in that, The priority of the M1 groups is higher than or equal to the priority of the M-M1 groups other than the M1 group.

31. The method according to any one of claims 28 to 30, characterized in that, After sending M1 groups of data according to the first mapping relationship, the method further includes: According to the first mapping relationship, data in M2 groups are sent. The M2 groups belong to the groups other than the M1 groups among the M groups, and M2 is a positive integer less than or equal to M-M1.

32. The method according to claim 31, characterized in that, The priority of the M2 groups is higher than or equal to the priority of the M-M1-M2 groups other than the M1 groups and the M2 groups.

33. A communication device, characterized in that, include: A processor configured to, by executing a computer program or instructions, cause the method of any one of claims 1 to 17 to be performed, or cause the method of any one of claims 18 to 32 to be performed.

34. The communication device according to claim 33, characterized in that, The communication device further includes a memory for storing the computer program or the instructions.

35. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions that, when executed on a computer, cause the method of any one of claims 1 to 17 to be performed, or cause the method of any one of claims 18 to 32 to be performed.

36. A computer program product, characterized in that, It includes a computer program or instructions that, when run, implement the method as described in any one of claims 1 to 17, or implement the method as described in any one of claims 18 to 32.

Images