Communication control method, communication device, communication system, and storage medium

Abstract

The present disclosure relates to a communication control method, a communication device, a communication system, and a storage medium. The method comprises: receiving first decoded data, wherein the first decoded data is determined by a second device on the basis of semantic data; training a first decoding unit at least once on the basis of the first decoded data and the semantic data until training of the first decoding unit is completed, wherein in this training, parameter adjustment is performed on the first decoding unit on the basis of a fine-tuning layer, the adjusted first decoding unit is used for retraining, and the fine-tuning layer belongs to the second device; receiving first information, wherein the first information is used for indicating parameters of the fine-tuning layer; and generating an encoder on the basis of the first information. Therefore, the training efficiency can be effectively improved, and the training costs can be reduced, thereby enabling effective application in mobile scenarios.

Description

Communication control methods, communication equipment, communication systems, and storage media Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a communication control method, communication equipment, communication system, and storage medium. Background Technology

[0002] The 6th generation mobile communication system (6G) will support many challenging applications, such as intelligent transportation and vehicle networks. Machine learning-based semantic communication will significantly advance next-generation 6G wireless network systems. In practice, effective semantic communication requires online training on unknown semantic content. However, in highly mobile vehicle networks, reliable and efficient model training becomes extremely challenging.

[0003] Summary of the Invention

[0004] This disclosure provides a communication control method, a first device, a second device, a chip system, a storage medium, a computer program, and a computer program product, which can be applied in the field of communication technology to solve the technical problem that "in related technologies, training efficiency is low and training cost is high, thus making them unsuitable for mobile scenarios".

[0005] This disclosure proposes a communication control method, communication equipment, communication system, and storage medium.

[0006] According to a first aspect of the present disclosure, a communication control method is proposed, executed by a first device, the first device including at least: a first decoding unit; wherein the method includes: receiving first decoded data, wherein the first decoded data is determined by a second device based on semantic data; training the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit training is completed, wherein in this training, the parameters of the first decoding unit are adjusted based on a fine-tuning layer, the adjusted first decoding unit is used for retraining, and the fine-tuning layer belongs to the second device; receiving first information, wherein the first information is used to indicate the parameters of the fine-tuning layer; and generating an encoder based on the first information.

[0007] According to a second aspect of the present disclosure, a communication control method is proposed, executed by a second device, the second device comprising: a fine-tuning layer and a second decoding unit; wherein the method comprises: processing semantic data through the second decoding unit to obtain first decoded data, wherein the first decoded data is used to determine a loss value; sending the first decoded data; training the second decoding unit at least once based on the loss value until the second decoding unit is trained, wherein in this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining; and sending first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0008] According to a third aspect of the present disclosure, a communication control method is proposed, comprising: a second device processing semantic data through a second decoding unit to obtain first decoded data, and sending the first decoded data, wherein the first decoded data is used to determine a loss value; the first device receiving the first decoded data, and training the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained, wherein in this training, the parameters of the first decoding unit are adjusted based on a fine-tuning layer, and the adjusted first decoding unit is used for retraining, the fine-tuning layer belonging to the second device; the second device training the second decoding unit at least once based on the loss value until the second decoding unit is trained, wherein in this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining; and sending first information, wherein the first information is used to indicate the parameters of the fine-tuning layer; the first device receiving the first information, and generating an encoder based on the first information.

[0009] According to a fourth aspect of the present disclosure, a first device is provided, comprising: a transceiver module for receiving first decoded data, wherein the first decoded data is determined by a second device based on semantic data; and receiving first information, wherein the first information is used to indicate parameters of a fine-tuning layer; a processing module for training a first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit training is completed, wherein in this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer, and the adjusted first decoding unit is used for retraining, the fine-tuning layer belonging to the second device; and also generating an encoder based on the first information.

[0010] According to a fifth aspect of the present disclosure, a second device is provided, comprising: a processing module, configured to process semantic data through a second decoding unit to obtain first decoded data, wherein the first decoded data is used to determine a loss value; and to train the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed, wherein, in this training, the parameters of the second decoding unit are adjusted based on a fine-tuning layer, and the adjusted second decoding unit is used for retraining;

[0011] The transceiver module is used to send first decoded data and first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0012] According to a sixth aspect of the present disclosure, a communication device is provided, comprising: one or more processors; wherein the processors are configured to invoke instructions to cause the communication device to execute a communication control method of any one of the first, second, and third aspects.

[0013] According to a seventh aspect of the present disclosure, a communication system is provided, characterized in that it includes a first device and a second device, wherein the first device is configured to implement the communication control method of the first aspect, and the second device is configured to implement the communication control method of the second aspect.

[0014] According to an eighth aspect of the present disclosure, a storage medium is provided, the storage medium storing instructions, characterized in that, when the instructions are executed on a communication device, the communication device causes the communication device to perform a communication control method as described in any one of the first, second, and third aspects.

[0015] According to a ninth aspect of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements a communication control method as described in any of the first, second, and third aspects. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments or background art of this disclosure, the accompanying drawings used in the embodiments or background art of this disclosure will be described below.

[0017] Figure 1A is a schematic diagram of the architecture of a communication system according to an embodiment of the present disclosure;

[0018] Figure 1B is a schematic diagram of a mobile scene in an embodiment of this disclosure;

[0019] Figure 2 is an interactive schematic diagram of a communication control method according to an embodiment of the present disclosure;

[0020] Figure 3A is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure;

[0021] Figure 3B is an interactive schematic diagram of a communication control method according to yet another embodiment of the present disclosure;

[0022] Figure 4A is an interactive schematic diagram of a communication control method according to yet another embodiment of the present disclosure;

[0023] Figure 4B is an interactive schematic diagram of a communication control method according to yet another embodiment of the present disclosure;

[0024] Figure 5 is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure;

[0025] Figure 6 is a schematic diagram of the framework of mobility-aware separation federated transfer learning in an embodiment of this disclosure;

[0026] Figure 7A is a schematic diagram of the structure of the first device proposed in an embodiment of this disclosure;

[0027] Figure 7B is a schematic diagram of the structure of the second device proposed in an embodiment of this disclosure;

[0028] Figure 8A is a schematic diagram of the structure of the communication device proposed in an embodiment of this disclosure;

[0029] Figure 8B is a schematic diagram of the chip structure proposed in an embodiment of this disclosure. Detailed Implementation

[0030] This disclosure presents a communication control method, communication device, communication system, and storage medium.

[0031] In a first aspect, embodiments of this disclosure propose a communication control method, executed by a first device, the first device including at least: a first decoding unit; wherein the method includes:

[0032] Receive first decoded data, wherein the first decoded data is determined by the second device based on semantic data;

[0033] The first decoding unit is trained at least once based on the first decoded data and semantic data until the first decoding unit is trained. During this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device.

[0034] Receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer; and

[0035] An encoder is generated based on the first information.

[0036] In the above embodiments, the first device receives first decoded data, which is determined by the second device based on semantic data; the first device trains the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit training is complete, wherein, in this training, the parameters of the first decoding unit are adjusted based on a fine-tuning layer, and the adjusted first decoding unit is used for retraining, the fine-tuning layer belonging to the second device; the first device receives first information, which is used to indicate the parameters of the fine-tuning layer; and generates an encoder based on the first information. Therefore, training efficiency can be effectively improved and training costs reduced, making it effectively applicable to mobile scenarios.

[0037] In conjunction with some embodiments of the first aspect, in some embodiments, the first device further includes: a first encoding unit and a second encoding unit; wherein generating the encoder based on the first information includes:

[0038] Configure the first encoding unit according to the first information;

[0039] An encoder is generated based on the configured first and second encoding units.

[0040] In the above embodiments, since the fine-tuning layer is trained in the second device, after training is completed, the first device can use the parameters of the fine-tuning layer and the second coding unit to synthesize the encoder in the first device, thereby effectively reducing the storage load in the first device. As a result, it can further support improved training efficiency and reduced training costs, thus effectively making it suitable for high-speed mobile scenarios.

[0041] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes:

[0042] Semantic features of semantic data are extracted through the second encoding unit;

[0043] Send semantic features, wherein the first decoded data is obtained by the second device decoding the semantic features.

[0044] In the above embodiments, after the first device extracts semantic features from the semantic data through the second encoding unit, it can send the semantic features to the second device so that the second device can decode the semantic features to obtain the first decoded data. This enables the extraction of semantic data features and the decoding of semantic data to be completed in different devices, effectively reducing the resource consumption required for the first device to process the semantic data, thereby reducing training costs and supporting improved training efficiency.

[0045] In conjunction with some embodiments of the first aspect, in some embodiments, training the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained is completed, including:

[0046] Based on the first decoded data and semantic data, the first decoding unit is trained this time, and the loss value is obtained;

[0047] If the loss value meets the loss condition, the training of the first decoding unit is considered complete.

[0048] If the loss value does not meet the loss condition, it is determined that the first decoding unit has not been trained completely.

[0049] The first decoding unit is then retrained based on the loss value until the first decoding unit is fully trained.

[0050] In the above embodiments, training efficiency is effectively improved while training accuracy is ensured, making it suitable for mobile scenarios.

[0051] In conjunction with some embodiments of the first aspect, in some embodiments, the first decoding unit is trained based on the first decoded data and semantic data to obtain a loss value, including:

[0052] The first decoded data is processed by the first decoding unit to obtain the second decoded data;

[0053] The loss value is determined based on the semantic data and the second decoded data.

[0054] In the above embodiments, the first decoding unit processes the first decoded data to obtain the second decoded data, and determines the loss value based on the semantic data and the second decoded data. This allows different decoding parts to be completed on different devices, reducing the decoding resource consumption required for training the first device and further improving training efficiency.

[0055] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes:

[0056] Send a second message, wherein the second message is used to indicate the parameters of the first decoding unit;

[0057] Receive target information, wherein the target information is obtained by aggregating multiple second information, different second information corresponds to different first decoding units, and different first decoding units belong to different first devices.

[0058] In the above embodiments, the first device can send second information to the second device, wherein the second information is used to indicate the parameters of the first decoding unit. The first device can then receive target information, wherein the target information is obtained by aggregating multiple pieces of second information, with different pieces of second information corresponding to different first decoding units, and different first decoding units belonging to different first devices. This ensures the consistency of the first decoding units in different first devices, guaranteeing the semantic communication accuracy and effectiveness of the overall semantic communication system.

[0059] In conjunction with some embodiments of the first aspect, in some embodiments, the first decoding unit is retrained based on the loss value, including:

[0060] Based on the loss value, determine the first adjustment information;

[0061] The parameters of the first decoding unit are adjusted based on the first adjustment information and the target information;

[0062] The first decoding unit is obtained through retraining and adjustment.

[0063] In the above embodiments, the process of retraining the first decoding unit based on the loss value can involve determining first adjustment information based on the loss value, adjusting the parameters of the first decoding unit based on the first adjustment information and the target information, and then retraining the adjusted first decoding unit. This effectively improves the accuracy of parameter adjustment and supports improved training precision.

[0064] In conjunction with some embodiments of the first aspect, in some embodiments, determining the first adjustment information based on the loss value includes:

[0065] Send the loss value, whereby the loss value is used by the fine-tuning layer to determine the first adjustment information;

[0066] Receive the first adjustment information.

[0067] In the above embodiments, during the process of determining the first adjustment information based on the loss value, the first device can send the loss value to the second device, wherein the loss value is used by the fine-tuning layer to determine the first adjustment information. The first device can receive the first adjustment information. This enables accurate acquisition of the first adjustment information used to adjust the parameters of the first decoding unit in this training, supporting improvements in training effectiveness and efficiency.

[0068] In conjunction with some embodiments of the first aspect, in some embodiments, the first device is a vehicle.

[0069] In the above embodiments, the first device is, for example, a vehicle. This significantly improves the training efficiency of the encoder in the vehicle, reduces the training cost of the encoder in the vehicle, and supports scenarios suitable for high-speed vehicle movement.

[0070] Secondly, embodiments of this disclosure propose a communication control method, executed by a second device, the second device comprising: a fine-tuning layer and a second decoding unit; wherein the method includes:

[0071] The semantic data is processed by the second decoding unit to obtain the first decoded data, wherein the first decoded data is used to determine the loss value;

[0072] Send the first decoded data;

[0073] The second decoding unit is trained at least once based on the loss value until the training of the second decoding unit is completed. During this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for further training.

[0074] Send a first message, which indicates the parameters of the fine-tuning layer.

[0075] In conjunction with some embodiments of the second aspect, in some embodiments, semantic data is processed by a second decoding unit to obtain first decoded data, including:

[0076] Semantic features of received semantic data;

[0077] The semantic features are decoded by the second decoding unit to obtain the first decoded data.

[0078] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes:

[0079] Receive multiple pieces of second information, wherein different pieces of second information correspond to different first decoding units, and different first decoding units belong to different first devices;

[0080] By aggregating multiple pieces of secondary information, the target information is obtained;

[0081] Send the target information.

[0082] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes:

[0083] Receive loss values, wherein the loss values ​​are determined based on semantic data and second decoded data, and the second decoded data is obtained by processing the first decoded data.

[0084] In conjunction with some embodiments of the second aspect, in some embodiments, the second decoding unit is trained at least once based on the loss value until the second decoding unit is fully trained, including:

[0085] If the loss value meets the loss condition, the training of the second decoding unit is complete.

[0086] The loss value does not meet the loss condition, indicating that the second decoding unit has not been fully trained.

[0087] The second decoding unit is then retrained based on the loss value until the second decoding unit is fully trained.

[0088] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes:

[0089] The loss value is processed by the fine-tuning layer to obtain the first adjustment information, wherein the first adjustment information is used to adjust the parameters of the first decoding unit in the first device;

[0090] Send the first adjustment message.

[0091] In conjunction with some embodiments of the second aspect, in some embodiments, parameter adjustment of the second decoding unit is performed based on the fine-tuning layer, including:

[0092] The loss value is processed by a fine-tuning layer to obtain the second adjustment information;

[0093] The parameters of the second decoding unit are adjusted according to the second adjustment information.

[0094] In conjunction with some embodiments of the second aspect, in some embodiments, the second device further includes: a third decoding unit; wherein the method further includes:

[0095] Configure the third decoding unit according to the target information;

[0096] The decoder is generated based on the configuration of the third and second decoding units.

[0097] In conjunction with some embodiments of the second aspect, in some embodiments, the second device is an edge cloud device.

[0098] Thirdly, embodiments of this disclosure propose a communication control method, the method comprising:

[0099] The second device processes semantic data through the second decoding unit to obtain first decoded data and sends the first decoded data, wherein the first decoded data is used to determine the loss value;

[0100] The first device trains the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained. During this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device.

[0101] The second device trains the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed. During this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining. The device also sends first information, which is used to indicate the parameters of the fine-tuning layer.

[0102] The first device generates an encoder based on the first information.

[0103] Fourthly, embodiments of this disclosure provide a first device, characterized in that the first device comprises:

[0104] The transceiver module is used to receive first decoded data, wherein the first decoded data is determined by the second device based on semantic data; and to receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0105] The processing module is used to train the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained. In this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer, and the adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device. The module also generates an encoder based on the first information.

[0106] Fifthly, embodiments of this disclosure provide a second device, characterized in that the second device comprises:

[0107] The processing module is used to process semantic data through the second decoding unit to obtain first decoded data, wherein the first decoded data is used to determine the loss value; and to train the second decoding unit at least once based on the loss value until the second decoding unit is trained. In this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining.

[0108] The transceiver module is used to send first decoded data and first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0109] Sixthly, embodiments of this disclosure provide a communication device, comprising:

[0110] One or more processors;

[0111] The processor is used to execute the communication control method of any one of the first, second, and third aspects.

[0112] In a seventh aspect, embodiments of this disclosure provide a communication system including a first device and a second device, wherein the first device is configured to implement the communication control method of the first aspect, and the second device is configured to implement the communication control method of the second aspect.

[0113] Eighthly, embodiments of this disclosure provide a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform a communication control method as described in the first, second, or third aspect.

[0114] Ninthly, embodiments of this disclosure provide a computer program product, including a computer program that, when executed by a processor, implements a communication control method as described in any of the first, second, and third aspects.

[0115] It is understood that the aforementioned communication control method, first device, second device, communication device, chip system, storage medium, computer program, and computer program product are all used to execute the methods proposed in the embodiments of this disclosure. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0116] This disclosure provides a communication control method and apparatus, a communication device, a communication system, and a storage medium. In some embodiments, the terms "communication control method" and "information processing method" or "communication method" can be used interchangeably; the terms "communication control apparatus" and "information processing apparatus" or "communication apparatus" can be used interchangeably; and the terms "information processing system" or "communication system" can be used interchangeably.

[0117] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0118] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0119] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0120] In this embodiment of the disclosure, unless otherwise stated, elements expressed in the singular form, such as "a," "an," "the," "the," "the," "the," "the," "the," "this," etc., can mean "one and only one," or "one or more," "at least one," etc. For example, when using articles such as "a," "an," "the," etc. in translation, the noun following the article can be understood as either a singular expression or a plural expression.

[0121] In the embodiments disclosed herein, "multiple" refers to two or more.

[0122] In some embodiments, the terms “at least one of”, “one or more”, “a plurality of”, “multiple”, etc., may be used interchangeably.

[0123] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "in response to one case A, in response to another case B", etc., may include the following technical solutions depending on the situation: in some embodiments, A (execute A regardless of B); in some embodiments, B (execute B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). The same applies when there are more branches such as A, B, C, etc.

[0124] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execution of A regardless of B); in some embodiments, B (execution of B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, C, etc.

[0125] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not impose restrictions on the position, order, priority, quantity, or content of the descriptive objects. The description of the descriptive objects is found in the claims or the context of the embodiments, and the use of prefixes should not constitute unnecessary restrictions. For example, if the descriptive object is a "field," the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is a "level," the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the number of descriptive objects is not limited by ordinal numbers and can be one or more. For example, in "first device," the number of "devices" can be one or more. Furthermore, the objects modified by different prefixes can be the same or different. For example, if the object being described is "device", then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. Similarly, if the object being described is "information", then "first information" and "second information" can be the same information or different information, and their content can be the same or different.

[0126] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0127] In some embodiments, the terms “in response to…”, “in response to determining…”, “in the case of…”, “when…”, “if…”, “if…”, etc., can be used interchangeably.

[0128] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0129] In some embodiments, the apparatus and device may be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. In some cases, they may also be understood as "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "body", etc.

[0130] In some embodiments, "network" can be interpreted as devices included in the network, such as access network devices, core network devices, etc.

[0131] In some embodiments, "access network device (AN device)" may also be referred to as "radio access network device (RAN device)," "base station (BS)," "radio base station," or "fixed station." In some embodiments, it may also be understood as "node," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "antenna panel," "antenna array," "cell," "macro cell," "small cell," "femto cell," "pico cell," "sector," "cell group," "serving cell," "carrier," "component carrier," or "bandwidth part (BWP)."

[0132] In some embodiments, "terminal" or "terminal device" may be referred to as "user equipment (UE)," "user terminal," "mobile station (MS)," "mobile terminal (MT)," "subscriber station," "mobile unit," "subscriber unit," "wireless unit," "remote unit," "mobile device," "wireless device," "wireless communication device," "remote device," "mobile subscriber station," "access terminal," "mobile terminal," "wireless terminal," "remote terminal," "handset," "user agent," "mobile client," "client," etc.

[0133] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0134] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0135] Figure 1A is a schematic diagram of the architecture of a communication system according to an embodiment of the present disclosure. As shown in Figure 1A, the communication system 100 may include a terminal 101 and a network device 102. The network device 102 may include at least one of an access network device and a core network device.

[0136] In some embodiments, terminal 101 may be an optional example of a first device. Terminal 101 may be deployed in a vehicle. Alternatively, a second device may be deployed in network device 102, such as an edge cloud device, without limitation.

[0137] In some embodiments, terminal 101 includes, but is not limited to, at least one of the following: mobile phone, wearable device, Internet of Things device, car with communication function, smart car, tablet computer, computer with wireless transceiver function, virtual reality (VR) terminal device, augmented reality (AR) terminal device, wireless terminal device in industrial control, wireless terminal device in self-driving, wireless terminal device in remote medical surgery, wireless terminal device in smart grid, wireless terminal device in transportation safety, wireless terminal device in smart city, and wireless terminal device in smart home.

[0138] In some embodiments, the access network device is, for example, a node or device that connects a terminal to a wireless network. The access network device may include, but is not limited to, at least one of the following in a 5G communication system: evolved Node B (eNB), next-generation eNB (ng-eNB), next-generation Node B (gNB), node B (NB), home node B (HNB), home evolved node B (HeNB), radio backhaul device, radio network controller (RNC), base station controller (BSC), base transceiver station (BTS), base band unit (BBU), mobile switching center, base station in a 6G communication system, open RAN, cloud RAN, base station in other communication systems, and access node in a WiFi system.

[0139] In some embodiments, the technical solutions of this disclosure can be applied to the Open RAN architecture. In this case, the interfaces between or within access network devices involved in the embodiments of this disclosure can be transformed into internal interfaces of Open RAN. The processes and information interactions between these internal interfaces can be implemented by software or programs.

[0140] In some embodiments, the access network device may be composed of a central unit (CU) and a distributed unit (DU). The CU may also be called a control unit. The CU-DU structure can separate the protocol layer of the access network device. Some of the protocol layer functions are centrally controlled by the CU, while the remaining part or all of the protocol layer functions are distributed in the DU and centrally controlled by the CU. However, this is not the only possibility.

[0141] In some embodiments, a core network device may be a single device comprising one or more network elements, or it may be multiple devices or a group of devices, each comprising all or part of one or more network elements. Network elements may be virtual or physical. The core network may include, for example, at least one of an Evolved Packet Core (EPC), a 5G Core Network (5GCN), or a Next Generation Core (NGC).

[0142] It is understood that the communication system described in this disclosure is for the purpose of more clearly illustrating the technical solutions of this disclosure, and does not constitute a limitation on the technical solutions proposed in this disclosure. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions proposed in this disclosure are also applicable to similar technical problems.

[0143] The following embodiments of this disclosure can be applied to the communication system 100 shown in FIG1A, or to some of the main bodies, but are not limited thereto. The main bodies shown in FIG1A are illustrative. The communication system may include all or some of the main bodies in FIG1A, or it may include other main bodies outside of FIG1A. The number and form of each main body are arbitrary. The connection relationship between the main bodies is illustrative. The main bodies may not be connected or may be connected. The connection can be in any way, it can be a direct connection or an indirect connection, it can be a wired connection or a wireless connection.

[0144] The embodiments disclosed herein can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G new radio (NR), 6th generation mobile communication system (6G), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New Radio Access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20, Ultra-Wideband (UWB), Bluetooth (a registered trademark), Public Land Mobile Network (PLMN) networks, Device-to-Device (D2D) systems, Machine-to-Machine (M2M) systems, Internet of Things (IoT) systems, Vehicle-to-Everything (V2X) systems, systems utilizing other communication methods, and next-generation systems built upon them, etc. Furthermore, multiple systems can be combined (e.g., a combination of LTE or LTE-A with 5G).

[0145] Optionally, 6G will support many challenging applications, such as intelligent transportation and vehicular networks. Machine learning-based semantic communication will significantly advance next-generation 6G wireless network systems. In practice, effective semantic communication requires online training on unknown semantic content. However, reliable and efficient model training becomes extremely challenging in highly mobile vehicular networks. Semantic communication, a new intelligent communication paradigm in 6G, offers a promising solution to these challenges.

[0146] Optionally, vehicular networks can generate and transmit large amounts of data. Therefore, edge cloud (EC) facilities can be introduced to provide vehicles with additional computing and caching resources. In this process, vehicles can offload tasks such as object and / or image recognition and spatial computing to the EC in real time via communication links, thereby accessing these resources. Consequently, the communication load between the vehicle and the edge cloud increases significantly. How to effectively improve the communication efficiency, reliability, and Quality of Service (QoS) of future vehicular networks has become a core issue supporting the continuous development of 6G vehicular networks. However, as the capacity of the wireless physical layer approaches the Shannon limit, wireless technology is increasingly unable to meet the complex data traffic and diverse offloading requirements of future 6G vehicular networks.

[0147] Alternatively, edge cloud EC facilities may also be referred to as edge cloud, EC, or edge cloud equipment, without limitation. Edge cloud EC facilities can be specifically deployed on roadside units (RSUs) or network equipment (such as base stations).

[0148] Optionally, the semantic communication system may include a deep learning-based joint source-channel (DLJSC) encoder. In this approach, the DLJSC encoder and decoder are deployed separately but jointly trained for specific transmission content to ensure that the DLJSC's transmission input and output are identical. This demonstrates the superior performance of the semantic communication system in improving communication efficiency and transmission accuracy. Therefore, the DLJSC encoder deployed on the vehicle and the DLJSC decoder deployed on the EC (Electronic Control Unit) are expected to improve communication efficiency during vehicle task unloading.

[0149] Alternatively, in practice, the DLJSC encoder can learn and update previously untrained semantic data (such as images, speech, and video) to ensure consistent QoS. However, related techniques suffer from low training efficiency and high training costs, making them unsuitable for mobile scenarios.

[0150] Optionally, as shown in Figure 1B, which is a schematic diagram of a mobile scenario in an embodiment of this disclosure, a set of edge cloud ECs is assumed, including {1, 2, ..., m, ..., M}, where 1, 2, ..., m, ..., M are used to represent the indices of each edge cloud EC. Each edge cloud EC can be deployed on a roadside unit (RSU) or a base station (BSs). A group of vehicles {1, 2, ..., n, ..., N} are within the service range (m) of the edge cloud (EC) (this range can be represented as EC m). Within the range of EC m, 1m vehicles participate in the training of the DLJSC encoder. Different vehicles transmit offloaded content (an optional example of training samples) to the edge cloud (EC) through different models of DLJSC encoders, and the edge cloud (EC) receives it through the DLJSC decoder. When the semantic knowledge base of vehicles or ECs is scarce, vehicles participating in training can be selected based on their speed. Semantic communication focuses on the semantic information effectively transmitted per second.

[0151] Figure 2 is an interactive schematic diagram of a communication control method according to an embodiment of the present disclosure. As shown in Figure 2, the embodiment of the present disclosure relates to a communication control method, which can be used in a communication system 100, wherein the communication system is, for example, a 6G communication network system, and is not limited thereto. The above method includes:

[0152] Step S2101: The first device extracts the semantic features of the semantic data.

[0153] Semantic data can also be called semantic communication data or training samples. Semantic data includes, for example, images, speech, and video data, without any restrictions.

[0154] Semantic features refer to the content features contained in semantic data. These content features can be used to characterize the semantic meaning of the semantic data. Semantic features can also be called fuzzy features, without limitation.

[0155] Optionally, the first device is, for example, a vehicle. This significantly improves the training efficiency of the encoder in the vehicle, reduces the training cost of the encoder in the vehicle, and supports scenarios suitable for high-speed vehicle movement.

[0156] Optionally, the number of first devices can be one or more, such as one or more vehicles cooperating with the edge cloud (EC) to complete the joint training process of the encoder on the vehicle side and the decoder on the edge cloud (EC) side. The interaction mode between each first device and the second device can be the same or similar, and can refer to the interaction mode between the first device and the second device described in the embodiments of this disclosure, without limitation.

[0157] Optionally, in some embodiments, the first device may include at least a first decoding unit, which can be considered a component or part of the decoder. The first decoding unit is used to perform part of the decoding function in the decoder. Specifically, the first decoding unit is, for example, the last layer decoder (P4). The training process of the first decoding unit can be completed on the first device. After the training of the first decoding unit is completed on the first device, the first device can send the parameters of the first decoding unit to the second device. That is to say, the second device can pre-deploy a decoding framework corresponding to the first decoding unit. This decoding framework can have the same decoding function as the first decoding unit. The second device can configure the decoding framework corresponding to the first decoding unit deployed in its own device based on the parameters of the first decoding unit, thereby realizing the training of the first decoding unit on the first device and the deployment of the trained first decoding unit on the second device. There are no restrictions on this.

[0158] Optionally, in some embodiments, the first device may further include a first encoding unit and a second encoding unit. The first encoding unit can be considered a component or part of the encoder, and may have the same architecture as the fine-tuning layer in the second device. The first encoding unit can specifically be used to form the encoder in the first device, as detailed in the following embodiments. Additionally, the second encoding unit can also be considered a component or part of the encoder, and is used to implement some encoding functions. Specifically, the second encoding unit may be, for example, a pre-trained model (P1). The first and second encoding units can be used together to form the encoder in the first device.

[0159] Optionally, in some embodiments, the first device may use a second coding unit to process the semantic data to extract semantic features from the semantic data. For example, the vehicle may use a pre-trained model (P1) to process the semantic data to extract semantic features. As another example, the vehicle may use a pre-trained model (P1) to process training samples to extract fuzzy features.

[0160] Step S2102: The first device sends semantic features.

[0161] Optionally, in some embodiments, after extracting the semantic features of the semantic data, the first device may send the semantic features to the second device so that the second device can perform preliminary decoding processing on the semantic features. See subsequent embodiments for details.

[0162] Optionally, in some embodiments, after the first device extracts semantic features from the semantic data through the second encoding unit, it can send the semantic features to the second device so that the second device can decode the semantic features to obtain the first decoded data. This allows for the completion of semantic data feature extraction and semantic data decoding in different devices, effectively reducing the resource consumption required for the first device to process the semantic data, thereby reducing training costs and supporting improved training efficiency.

[0163] Optionally, in some embodiments, the first device may carry semantic features in any possible communication message to send semantic features to the second device, without limitation.

[0164] In step S2103, the second device decodes the semantic features to obtain the first decoded data.

[0165] Optionally, in some embodiments, the second device may be, for example, an edge cloud EC device. The second device may be specifically deployed on a roadside unit (RSU) or a base station. The number of second devices may also be one or more, for example, one or more vehicles cooperating with each edge cloud (EC) to complete the joint training process of one or more vehicle-side encoders and each edge cloud (EC)-side decoder. The interaction method between the first device and each second device may be the same or similar, and can refer to the interaction method between the first device and the second device described in the embodiments of this disclosure, without limitation.

[0166] The first decoded data is used to determine the loss value.

[0167] Optionally, in some embodiments, the second device may receive semantic features sent by the first device and decode the semantic features to obtain first decoded data.

[0168] Optionally, in some embodiments, the second device may include at least a fine-tuning layer (P2) and a second decoding unit, wherein the second decoding unit can be considered a component or part of the decoder. The second decoding unit is used to perform part of the decoding function in the decoder. Specifically, the second decoding unit is, for example, an EC dedicated decoder (P3). The training process of the second decoding unit can be completed on the second device, and the second device can use at least the second decoding unit to form the decoder in the device, the process of which can be specifically described in subsequent embodiments.

[0169] Optionally, in some embodiments, after receiving the semantic features, the second device can decode the received semantic features through a second decoding unit, and the decoded data can be referred to as the first decoded data. This enables feature extraction and semantic feature-based decoding of semantic data to be completed in different devices, further reducing the resource consumption required for decoding semantic data on the first device side, thereby reducing training costs and supporting improved training efficiency.

[0170] Optionally, in some embodiments, the second device may also work with the fine-tuning layer and the second decoding unit to decode the semantic features to obtain the first decoded data, thereby accurately determining the first decoded data corresponding to the semantic data.

[0171] Step S2104: The second device sends the first decoded data.

[0172] Optionally, in some embodiments, after decoding the semantic features to obtain the first decoded data, the second device may also send the first decoded data to the first device. The first decoded data may support the first device in training a first decoding unit (e.g., the last layer decoder (P4)) within the device. For example, the second device may carry the first decoded data in any possible communication message to achieve the sending of the first decoded data to the first device.

[0173] In step S2105, the first device trains the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained. During this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer, and the adjusted first decoding unit is used for retraining.

[0174] Optionally, in some embodiments, the first device may receive the first decoded data sent by the second device, and train the first decoding unit (the last layer decoder (P4)) at least once by combining the first decoded data and semantic data until the first decoding unit training is completed.

[0175] Optionally, in some embodiments, the first device may train the first decoding unit once or multiple times based on the first decoding data and semantic data. After each training, a loss value can be determined. The first device can use the loss value to determine whether the first decoding unit obtained in this training is trained successfully. If it is determined that the first decoding unit is trained successfully, the method steps for generating the encoder in the device can be triggered. If it is determined that the first decoding unit is not trained successfully, the next training of the first decoding unit obtained in this training can be triggered (the method steps for the next training can be referred to the method steps for this training. In each training, at least one interaction is performed with the second device to send or obtain various data required for training, such as the first decoding data, semantic features, loss value, etc.), until the first decoding unit is trained successfully.

[0176] Optionally, in the process of training the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained, the training can be performed on the first decoding unit based on the first decoded data and semantic data to obtain a loss value. If the loss value meets the loss condition, the training of the first decoding unit is determined to be complete. If the loss value does not meet the loss condition, the training of the first decoding unit is determined to be incomplete, and the first decoding unit is trained again based on the loss value until the training of the first decoding unit is complete. This effectively improves training efficiency while ensuring training accuracy, making it suitable for mobile scenarios.

[0177] Optionally, in some embodiments, during the process of training the first decoding unit based on the first decoded data and semantic data to obtain the loss value, the first decoding unit can process the first decoded data to obtain second decoded data, and then determine the loss value based on the semantic data and the second decoded data. This allows different decoding parts to be completed on different devices, reduces the decoding resource consumption required for training the first device, and further improves training efficiency.

[0178] In other words, the first decoding unit configured in the first device can perform part of the decoding function, and the second decoding unit in the second device can perform the other part of the decoding function. The second device can decode to obtain the first decoded data corresponding to the semantic data and send the first decoded data to the first device. The first device can then decode the first decoded data again based on the first decoding unit to obtain the second decoded data. Subsequently, a function calculation can be performed on the semantic data and the second decoded data based on a predefined loss function to determine the loss value.

[0179] It is understood that, since the training method described in this embodiment is a joint training method of the first device and the second device, when the first decoding unit is trained once in the first device, the second decoding unit in the second device will also be trained once in the same way. During the same iterative training process, the first device and the second device can exchange some training data (such as first decoding data, semantic features, loss values, etc.) to support the iterative training of the corresponding units in each device.

[0180] Optionally, in some embodiments, after obtaining the loss value in the current training of the first decoding unit, the first device can send the loss value to the second device. After receiving the loss value, the second device can determine the first adjustment information based on the fine-tuning layer (P2) configured in the device. The first adjustment information refers to the parameter adjustment of the first decoding unit obtained in this training. The first decoding unit obtained by parameter adjustment is used for the next training until it is determined that the training of the first decoding unit obtained in a certain training is completed.

[0181] Optionally, in some embodiments, after training the first decoding unit, the first device may further determine the parameters of the first decoding unit, which are the parameters used by the first decoding unit to perform decoding.

[0182] Optionally, in some embodiments, the first device may send second information to the second device, wherein the second information is used to indicate the parameters of the first decoding unit. The first device may then receive target information, wherein the target information is obtained by aggregating multiple pieces of second information, with different pieces of second information corresponding to different first decoding units, and different first decoding units belonging to different first devices. This ensures the consistency of the first decoding units in different first devices, guaranteeing the semantic communication accuracy and effectiveness of the overall semantic communication system.

[0183] It is understandable that when there are multiple first devices, each first device can send the parameters of its first decoding unit to the second device via the second information. Therefore, the second device can obtain the second information sent by multiple first devices.

[0184] Optionally, in some embodiments, the second device can receive multiple pieces of second information, wherein different pieces of second information correspond to different first decoding units, and different first decoding units belong to different first devices. The second device can also aggregate the multiple pieces of second information to obtain target information and send the target information. This ensures the consistency of the first decoding units in different first devices, and ensures the semantic communication accuracy and effectiveness of the overall semantic communication system.

[0185] Optionally, in some embodiments, the process of retraining the first decoding unit based on the loss value may involve determining first adjustment information based on the loss value, adjusting the parameters of the first decoding unit based on the first adjustment information and the target information, and then retraining the adjusted first decoding unit. This can effectively improve the accuracy of parameter adjustment and support improved training precision.

[0186] In other words, during each training iteration, the first device can send the loss value and the parameters of the first decoding unit to the second device. The loss value can be used by the second device to determine the first adjustment information. Furthermore, the second device can aggregate the parameters of multiple second decoding units and use target information to describe the aggregation result. The second device can then feed back the first adjustment information and target information to the first device. Since the target information ensures the consistency of the first decoding units across different first devices, and the first adjustment information is determined based on the loss value obtained during this training, the first device can use the received first adjustment information and target information to adjust the parameters of the first decoding unit and retrain the adjusted first decoding unit. Therefore, while effectively maintaining the consistency of the first decoding units across different first devices, the training effect and efficiency can also be improved.

[0187] Optionally, in some embodiments, during the process of determining the first adjustment information based on the loss value, the first device may send the loss value to the second device, wherein the loss value is used by the fine-tuning layer to determine the first adjustment information. The first device may receive the first adjustment information. This enables accurate acquisition of the first adjustment information used to adjust the parameters of the first decoding unit in this training, supporting improvements in training effectiveness and efficiency.

[0188] Optionally, in some embodiments, after receiving the loss value sent by the first device, the second device can process the loss value through a fine-tuning layer in its own device to obtain first adjustment information. This first adjustment information is used to adjust the parameters of the first decoding unit in the first device. The second device can then send the first adjustment information back to the first device. For example, the fine-tuning layer can select some features from the semantic features with reference to the loss value, and use these features, along with the parameters and weights corresponding to the first decoding unit, to determine an adjustment amount corresponding to at least one parameter of the first decoding unit. This adjustment amount corresponding to at least one parameter of the first decoding unit is then used as second adjustment information.

[0189] In step S2106, during each training session, the first device determines the loss value based on the first decoded data and semantic data, and sends the loss value.

[0190] It should be noted that the execution order of steps S2106 and the above steps is not limited. Step S2106 can be executed once in each training process in step S2105. In step S2106, after the first device determines the loss value using the first decoded data and semantic data, it can send the loss value to the second device. In the above description, the method steps for the second device to determine the first adjustment information using the loss value have been described.

[0191] Furthermore, since it is a process of joint training of the encoder and decoder by different devices, the second device can also use the loss value to perform at least one iterative training on the second decoding unit deployed in this device (e.g., the EC dedicated decoder (P3)). That is, each time the iterative training of the first decoding unit in the first device is completed, the iterative training of the second decoding unit in the second device is completed accordingly, until the training of the second decoding unit in the second device is completed.

[0192] Optionally, in some embodiments, after the first decoding unit is trained and the loss value is obtained, the first device may send the loss value to the second device.

[0193] In step S2107, the second device trains the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed. During this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining.

[0194] It should be noted that the execution order of steps S2107 and the above steps is not limited. Step S2107 can be executed once in each training process in step S2105. In step S2107, the second device can perform a corresponding iterative training on the second decoding unit based on the loss value until the training of the second decoding unit is completed.

[0195] Optionally, in some embodiments, the second device may receive a loss value during each training process. The second device can then train the second decoding unit at least once based on the loss value until the second decoding unit is fully trained.

[0196] Optionally, in some embodiments, during this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining. This ensures that the parameters of the second decoding unit during training are adjusted in a timely manner, supporting improvements in the training effect and efficiency of the second decoding unit in the second device.

[0197] Optionally, in some embodiments, during the process of training the second decoding unit at least once based on the loss value until the second decoding unit is fully trained, the second device may determine that the second decoding unit is fully trained if the loss value meets the loss condition. If the loss value does not meet the loss condition, the second decoding unit is determined not to be fully trained, and the second decoding unit is trained again based on the loss value until the second decoding unit is fully trained. This ensures the training effect and efficiency of the second decoding unit in the second device, effectively applicable to mobile scenarios.

[0198] In other words, during each training session, the second device can receive the loss value and determine whether the second decoding unit obtained in this training session has been successfully trained based on the loss value. For example, a loss threshold can be preset. If the loss value is less than the loss threshold, it is determined that the second decoding unit obtained in this training session has been successfully trained. If the loss value is greater than or equal to the loss threshold, it is determined that the second decoding unit obtained in this training session has not been successfully trained, and the next training session is initiated. Before initiating the next training session, the parameters of the second decoding unit obtained in this training session are adjusted based on the loss value, and the adjusted second decoding unit is used for retraining.

[0199] Optionally, in some embodiments, during the process of adjusting the parameters of the second decoding unit based on the fine-tuning layer, the second device can process the loss value through the fine-tuning layer to obtain second adjustment information, and adjust the parameters of the second decoding unit according to the second adjustment information. For example, the fine-tuning layer can select some features from the semantic features with reference to the loss value, and use the partial features and the parameters and weights corresponding to the second decoding unit to determine the adjustment amount corresponding to at least one parameter of the second decoding unit, and use the adjustment amount corresponding to at least one parameter of the second decoding unit as the second adjustment information.

[0200] Step S2108: The second device sends the first information.

[0201] The first piece of information is used to indicate the parameters of the fine-tuning layer. The parameters of the fine-tuning layer refer to the parameters used by the fine-tuning layer when determining the adjustment information of each unit, and there are no restrictions on them.

[0202] Optionally, in some embodiments, if the training of the second decoding unit in the second device is completed, it indicates that the joint training process of the first device and the second device is completed. At this time, it indicates that the trained first decoding unit and the second decoding unit meet the training expectations, and then the method steps of synthesizing the encoder in the first device and the decoder in the second device can be executed.

[0203] Optionally, in some embodiments, after the second decoding unit in the second device has been trained, the second device can obtain the parameters of the fine-tuning layer and indicate the parameters of the fine-tuning layer to the first device.

[0204] Step S2109: The first device generates an encoder based on the first information.

[0205] Optionally, in some embodiments, the first device may learn the parameters of the fine-tuning layer based on the first information and generate an encoder according to the parameters of the fine-tuning layer.

[0206] Optionally, in some embodiments, the first device may further include a first encoding unit and a second encoding unit. The first encoding unit may have the same architecture as the fine-tuning layer in the second device, and the first encoding unit may specifically be used to form the encoder in the first device. The second encoding unit is used to implement additional encoding functions. The second encoding unit is, for example, a pre-trained model (P1). The first encoding unit and the second encoding unit may be used together to form the encoder in the first device.

[0207] Optionally, in some embodiments, the first device can configure the first encoding unit in its own device according to the parameters of the fine-tuning layer indicated in the first information (here, since the first encoding unit can have the same fine-tuning function as the fine-tuning layer in the second device, the parameters of the fine-tuning layer in the second device can be used to configure the first encoding unit in the first device, so that the configured first encoding unit can have the superior functional characteristics of the fine-tuning layer after training in the second device). Then, the first device can synthesize the encoder according to the configured first encoding unit and the second encoding unit. Since the fine-tuning layer is trained in the second device, after training, the first device can use the parameters of the fine-tuning layer and the second encoding unit to synthesize the encoder in its own device, thereby effectively reducing the storage load in the first device. This further supports improving training efficiency and reducing training costs, thus effectively making it suitable for high-speed mobile scenarios.

[0208] Optionally, in some embodiments, the second device further includes a third decoding unit. The second device can also configure the third decoding unit according to target information and generate a decoder based on the configured third decoding unit and the second decoding unit. That is, a decoding framework corresponding to the first decoding unit can be pre-deployed in the second device. This decoding framework can have the same decoding function as the first decoding unit. After receiving the second information sent by the first device, and obtaining the parameters of the first decoding unit based on the second information, and aggregating the parameters of multiple first decoding units to obtain target information (i.e., the aggregated parameters of the first decoding unit), the second device can configure the decoding framework deployed in this device corresponding to the first decoding unit based on the target information. This enables training the first decoding unit on the first device and deploying the trained first decoding unit on the second device, without limitation.

[0209] The communication control method disclosed in this embodiment may include at least one of steps S2101 to S2109. For example, step S2101 may be implemented as a standalone embodiment, step S2102 may be implemented as a standalone embodiment, and so on, but is not limited thereto. Steps S2101+S2102 may be implemented as standalone embodiments, and steps S2101+S2102+S2103 may be implemented as standalone embodiments, but is not limited thereto.

[0210] In this implementation or embodiment, unless there is contradiction, each step can be independent, arbitrarily combined or exchanged in order, optional methods or optional examples can be arbitrarily combined, and can be arbitrarily combined with any steps of other implementations or other embodiments.

[0211] In this embodiment, the first device extracts and sends semantic features from the semantic data, the second device decodes the semantic features to obtain first decoded data, the second device sends the first decoded data, and the first device trains the first decoding unit at least once based on the first decoded data and the semantic data until the first decoding unit is trained. In each training iteration, the first device determines a loss value based on the first decoded data and the semantic data and sends the loss value. The second device trains the second decoding unit at least once based on the loss value until the second decoding unit is trained. The second device sends first information, and the first device generates an encoder based on the first information. This effectively improves training efficiency and reduces training costs, making it suitable for mobile scenarios. Furthermore, since the fine-tuning layer is trained in the second device, after training, the first device can use the parameters of the fine-tuning layer and the second encoding unit to synthesize the encoder in its own device, effectively reducing the storage load in the first device. This further supports improved training efficiency and reduced training costs, making it suitable for high-speed mobile scenarios.

[0212] Figure 3A is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure. As shown in Figure 3A, the embodiment of the present disclosure relates to a communication control method, which can be used in a first device, the first device including at least: a first decoding unit. The above method includes:

[0213] Step S3101: Receive first decoded data, wherein the first decoded data is determined by the second device based on semantic data.

[0214] Step S3102: Train the first decoding unit at least once based on the first decoding data and semantic data until the first decoding unit is trained. In this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device.

[0215] Step S3103: Receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0216] Step S3104: Generate an encoder based on the first information.

[0217] The communication control method disclosed in this embodiment may include at least one of steps S3101 to S3104. For example, step S3101 may be implemented as a standalone embodiment, step S3102 may be implemented as a standalone embodiment, and so on, but is not limited thereto. Steps S3101+S3102 may be implemented as standalone embodiments, but are not limited thereto.

[0218] In this implementation or embodiment, unless there is contradiction, each step can be independent, arbitrarily combined or exchanged in order, optional methods or optional examples can be arbitrarily combined, and can be arbitrarily combined with any steps of other implementations or other embodiments.

[0219] Figure 3B is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure. As shown in Figure 3B, the embodiment of the present disclosure relates to a communication control method, which can be used in a first device. The first device includes: a first decoding unit, a first encoding unit, and a second encoding unit. The above method includes:

[0220] Step S3201: Extract semantic features of semantic data through the second encoding unit.

[0221] Step S3202: Send semantic features, wherein the first decoded data is obtained by the second device decoding the semantic features.

[0222] Step S3203: Receive the first decoded data.

[0223] Step S3204: Train the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained. In this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device.

[0224] Step S3205: Receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0225] Step S3206: Configure the first encoding unit according to the first information.

[0226] Step S3207: Generate an encoder based on the configured first and second encoding units.

[0227] The communication control method disclosed in this embodiment may include at least one of steps S3201 to S3207. For example, step S3201 may be implemented as a standalone embodiment, step S3202 may be implemented as a standalone embodiment, and so on, but is not limited thereto. Steps S3201+S3202 may be implemented as standalone embodiments, but are not limited thereto.

[0228] In this implementation or embodiment, unless there is contradiction, each step can be independent, arbitrarily combined or exchanged in order, optional methods or optional examples can be arbitrarily combined, and can be arbitrarily combined with any steps of other implementations or other embodiments.

[0229] In some embodiments of this disclosure, training a first decoding unit at least once based on first decoded data and semantic data until the first decoding unit is trained is completed, including:

[0230] Based on the first decoded data and semantic data, the first decoding unit is trained this time, and the loss value is obtained;

[0231] If the loss value meets the loss condition, the training of the first decoding unit is considered complete.

[0232] If the loss value does not meet the loss condition, it is determined that the first decoding unit has not been trained completely.

[0233] The first decoding unit is then retrained based on the loss value until the first decoding unit is fully trained.

[0234] In some embodiments of this disclosure, the first decoding unit is trained based on the first decoded data and semantic data to obtain a loss value, including:

[0235] The first decoded data is processed by the first decoding unit to obtain the second decoded data;

[0236] The loss value is determined based on the semantic data and the second decoded data.

[0237] In some embodiments of this disclosure, the method further includes:

[0238] Send a second message, wherein the second message is used to indicate the parameters of the first decoding unit;

[0239] Receive target information, wherein the target information is obtained by aggregating multiple second information, different second information corresponds to different first decoding units, and different first decoding units belong to different first devices.

[0240] In some embodiments of this disclosure, the first decoding unit is retrained based on the loss value, including:

[0241] Based on the loss value, determine the first adjustment information;

[0242] The parameters of the first decoding unit are adjusted based on the first adjustment information and the target information;

[0243] The first decoding unit is obtained through retraining and adjustment.

[0244] In some embodiments of this disclosure, determining first adjustment information based on the loss value includes:

[0245] Send the loss value, whereby the loss value is used by the fine-tuning layer to determine the first adjustment information;

[0246] Receive the first adjustment information.

[0247] In some embodiments of this disclosure, the first device is a vehicle.

[0248] Figure 4A is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure. As shown in Figure 4A, the embodiment of the present disclosure relates to a communication control method that can be used in a second device, the second device including: a fine-tuning layer and a second decoding unit. The above method includes:

[0249] Step S4101: The semantic data is processed by the second decoding unit to obtain the first decoded data, wherein the first decoded data is used to determine the loss value.

[0250] Step S4102: Send the first decoded data.

[0251] Step S4103: Train the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed. In this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining.

[0252] Step S4104: Send first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0253] The communication control method disclosed in this embodiment may include at least one of steps S4101 to S4104. For example, step S4101 may be implemented as a standalone embodiment, step S4102 may be implemented as a standalone embodiment, and so on, but is not limited thereto. Steps S4101+S4102 may be implemented as standalone embodiments, but are not limited thereto.

[0254] In this implementation or embodiment, unless there is contradiction, each step can be independent, arbitrarily combined or exchanged in order, optional methods or optional examples can be arbitrarily combined, and can be arbitrarily combined with any steps of other implementations or other embodiments.

[0255] Figure 4B is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure. As shown in Figure 4B, the embodiments of the present disclosure relate to a communication control method that can be used in a second device, the second device including: a fine-tuning layer, a second decoding unit, and a third decoding unit. The above method includes:

[0256] Step S4201: Receive the semantic features of the semantic data.

[0257] Step S4202: The semantic features are decoded by the second decoding unit to obtain the first decoded data.

[0258] Step S4203: Send the first decoded data.

[0259] Step S4204: Receive the loss value, wherein the loss value is determined based on semantic data and second decoded data, and the second decoded data is obtained by processing the first decoded data.

[0260] Step S4205: Train the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed. In this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining.

[0261] Step S4206: Send first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0262] The communication control method disclosed in this embodiment may include at least one of steps S4201 to S4206. For example, step S4201 may be implemented as a standalone embodiment, step S4202 may be implemented as a standalone embodiment, and so on, but is not limited thereto. Steps S4201+S4202 may be implemented as standalone embodiments, but are not limited thereto.

[0263] In this implementation or embodiment, unless there is contradiction, each step can be independent, arbitrarily combined or exchanged in order, optional methods or optional examples can be arbitrarily combined, and can be arbitrarily combined with any steps of other implementations or other embodiments.

[0264] In some embodiments of this disclosure, the method further includes:

[0265] Receive multiple pieces of second information, wherein different pieces of second information correspond to different first decoding units, and different first decoding units belong to different first devices;

[0266] By aggregating multiple pieces of secondary information, the target information is obtained;

[0267] Send the target information.

[0268] In some embodiments of this disclosure, the second decoding unit is trained at least once based on the loss value until the second decoding unit is fully trained, including:

[0269] If the loss value meets the loss condition, the training of the second decoding unit is complete.

[0270] The loss value does not meet the loss condition, indicating that the second decoding unit has not been fully trained.

[0271] The second decoding unit is then retrained based on the loss value until the second decoding unit is fully trained.

[0272] In some embodiments of this disclosure, the method further includes:

[0273] The loss value is processed by the fine-tuning layer to obtain the first adjustment information, wherein the first adjustment information is used to adjust the parameters of the first decoding unit in the first device;

[0274] Send the first adjustment message.

[0275] In some embodiments of this disclosure, parameter adjustment of the second decoding unit is performed based on the fine-tuning layer, including:

[0276] The loss value is processed by a fine-tuning layer to obtain the second adjustment information;

[0277] The parameters of the second decoding unit are adjusted according to the second adjustment information.

[0278] In some embodiments of this disclosure, the second device further includes: a third decoding unit; wherein the method further includes:

[0279] Configure the third decoding unit according to the target information;

[0280] The decoder is generated based on the configuration of the third and second decoding units.

[0281] In some embodiments of this disclosure, the second device is an edge cloud device.

[0282] Figure 5 is an interactive schematic diagram of a communication control method according to another embodiment of the present disclosure.

[0283] As shown in Figure 5, this method is applied to a communication system. The above method includes:

[0284] In step S5101, the second device processes the semantic data through the second decoding unit to obtain the first decoded data and sends the first decoded data, wherein the first decoded data is used to determine the loss value.

[0285] In step S5102, the first device trains the first decoding unit at least once based on the first decoding data and semantic data until the first decoding unit is trained. During this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device.

[0286] Step S5103: The second device trains the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed. In this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining. The first information is sent to indicate the parameters of the fine-tuning layer.

[0287] Step S5104: The first device generates an encoder based on the first information.

[0288] The communication control method involved in the embodiments of this disclosure may include at least one of steps S5101 to S5104. For example, step S5101 may be implemented as a standalone embodiment, step S5102 may be implemented as a standalone embodiment, and so on, but is not limited thereto. Steps S5101 + S5102 may be implemented as standalone embodiments, but are not limited thereto.

[0289] In this implementation or embodiment, unless there is contradiction, each step can be independent, arbitrarily combined or exchanged in order, optional methods or optional examples can be arbitrarily combined, and can be arbitrarily combined with any steps of other implementations or other embodiments.

[0290] The following is an exemplary description of the above method.

[0291] Optionally, the following embodiments are available:

[0292] Optionally, the training framework described above in this disclosure embodiment can be referred to as a mobility-aware separation federated transfer learning framework or a vehicle semantic communication framework. This mobility-aware separation federated transfer learning framework may include an encoder (e.g., an autoencoder). The encoder can extract features from the input (e.g., training samples, semantic data, etc.) through downscaling, and then recover the corresponding content (e.g., an image) through a decoder. During the encoder training process, the encoder can convert the input 'a' into an intermediate feature variable 'b'. Therefore, 'b' is converted back into 'a' by the decoder. ～ Finally, compare the input a and the output a~ to ensure they are infinitely close.

[0293] Optionally, the following description uses the first device as a vehicle and the second device as an edge cloud device as an example, without limitation.

[0294] Optionally, in this embodiment of the disclosure, the encoder in the training vehicle can be divided into a pre-trained model (P1) and a fine-tuning layer (P2), with each vehicle allowed to have different types of pre-trained models (P1). The pre-trained model (P1) is part of the encoder, while to alleviate the problem of small sample size, the fine-tuning layer is trained on an edge cloud device, as described below. Therefore, the vehicle only needs to ensure that the last few layers of the encoder have the same model. This reduces the storage resources and training costs required for different tasks.

[0295] Figure 6 illustrates the framework of mobility-aware separation federated transfer learning in this embodiment. It includes vehicles 1, ..., n, and an edge cloud device EC m, where m represents the service range of the edge cloud device. Training can be divided into four parts: a pre-trained model (P1), a fine-tuning layer (P2), an EC-specific decoder (P3), and a final decoder layer (P4). First, the trainable vehicles and training data are determined. During encoder training, the pre-trained model (P1) and the final decoder layer (P4) are trained on the vehicles, while the fine-tuning layer (P2) and the EC-specific decoder (P3) are trained on the EC. For trainable vehicles n within the EC m range, training samples x are first... n,m (An optional example of semantic data, also referred to as the source message) Extracting fuzzy features (An optional example of a semantic feature). Feature By freezing the pre-trained model P1 n,m (P1 n,m This represents the pre-trained model (P1) on the nth vehicle, and ECm is transmitted. ECm then blurs the features. As input, and begin training. During a single training session, EC m can be used For fine-tuning layer P2 m (P2 m (This refers to the fine-tuning layer trained in EC m) and the EC m dedicated decoder P3. m Conduct forward propagation training. From P3 m The result of forward propagation, namely The data is sent to the corresponding vehicle n. Then, the last layer decoder P4 is trained on the corresponding vehicle n. n,m (P4 n,m This represents the last decoder layer (P4) on the nth vehicle, which produces the output. Then, the vehicle compares the source message x. n,m With forward propagation output The difference between them yields the loss value L. n,m Then proceed with L-based... n,m The backpropagation process is repeated, and it returns along the same path until the fine-tuning layer P2.m Finally, due to the last layer decoder P4 n,m Training was performed only for a single vehicle, therefore, federated aggregation can be used to ensure that the decoders in different vehicles are identical. The vehicles participating in training will use the final layer decoder P4. n,m The parameters are sent to EC m for aggregation, and then EC m sends the aggregation result P4. m The results are returned to each sending vehicle. All participating vehicles complete one training iteration after executing this process once. After completing the entire training, P1... n,m and P2 m The DLJSC encoders that make up vehicle n. Similarly, P3. m and P4 m This constitutes the DLJSC decoder for EC m. Throughout the training process, due to P1... n,m and x n,m There is no leakage, thus effectively protecting the privacy of vehicle-side data from being leaked, thereby improving training security. Vehicles only need to change the fine-tuning layer for different transmitted content, thereby reducing the vehicle's storage load.

[0296] This disclosure also provides embodiments of an apparatus for implementing any of the above methods. For example, an apparatus is provided that includes units or modules for implementing the steps performed by the terminal in any of the above methods. Alternatively, another apparatus is provided that includes units or modules for implementing the steps performed by a network device (e.g., a RAN) in any of the above methods.

[0297] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC), and the functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD), such as a field-programmable gate array (FPGA), which can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0298] In this embodiment, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), or a Deep Learning Processing Unit (DPU).

[0299] Figure 7A is a schematic diagram of the structure of the first device proposed in an embodiment of this disclosure. As shown in Figure 7A, the first device 7100 may include at least one of a transceiver module 7101, a processing module 7102, etc. The first device 7100 may include:

[0300] The transceiver module 7101 is used to receive first decoded data, wherein the first decoded data is determined by the second device based on semantic data; and to receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0301] The processing module 7102 is used to train the first decoding unit at least once based on the first decoded data and semantic data until the first decoding unit is trained. In this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device. The module also generates an encoder based on the first information.

[0302] In some embodiments of this disclosure, the first device further includes: a first encoding unit and a second encoding unit; wherein, the processing module 7102 is used for:

[0303] Configure the first encoding unit according to the first information;

[0304] An encoder is generated based on the configured first and second encoding units.

[0305] In some embodiments of this disclosure, wherein,

[0306] Processing module 7102 is used to extract semantic features of semantic data through the second encoding unit;

[0307] The transceiver module 7101 is used to send semantic features, wherein the first decoded data is obtained by the second device decoding the semantic features.

[0308] In some embodiments of this disclosure, the processing module 7102 is configured to:

[0309] Based on the first decoded data and semantic data, the first decoding unit is trained this time, and the loss value is obtained;

[0310] If the loss value meets the loss condition, the training of the first decoding unit is considered complete.

[0311] If the loss value does not meet the loss condition, it is determined that the first decoding unit has not been trained completely.

[0312] The first decoding unit is then retrained based on the loss value until the first decoding unit is fully trained.

[0313] In some embodiments of this disclosure, the processing module 7102 is configured to:

[0314] The first decoded data is processed by the first decoding unit to obtain the second decoded data;

[0315] The loss value is determined based on the semantic data and the second decoded data.

[0316] In some embodiments of this disclosure, wherein,

[0317] The transceiver module 7101 is used to send second information, wherein the second information is used to indicate the parameters of the first decoding unit; and to receive target information, wherein the target information is obtained by aggregating multiple pieces of second information, different pieces of second information correspond to different first decoding units, and different first decoding units belong to different first devices.

[0318] In some embodiments of this disclosure, the processing module 7102 is configured to:

[0319] Based on the loss value, determine the first adjustment information;

[0320] The parameters of the first decoding unit are adjusted based on the first adjustment information and the target information;

[0321] The first decoding unit is obtained through retraining and adjustment.

[0322] In some embodiments of this disclosure, the transceiver module 7101 is used for:

[0323] Send the loss value, whereby the loss value is used by the fine-tuning layer to determine the first adjustment information;

[0324] Receive the first adjustment information.

[0325] In some embodiments of this disclosure, the first device is a vehicle.

[0326] Figure 7B is a schematic diagram of the structure of the second device proposed in an embodiment of this disclosure. As shown in Figure 7B, the second device 7200 may include at least one of a transceiver module 7201, a processing module 7202, etc. The second device 7200 may include:

[0327] The processing module 7202 is used to process semantic data through the second decoding unit to obtain first decoded data, wherein the first decoded data is used to determine the loss value; and to train the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed, wherein in this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining.

[0328] The transceiver module 7201 is used to send first decoded data and first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

[0329] In some embodiments of this disclosure, the processing module 7202 is used for:

[0330] Semantic features of received semantic data;

[0331] The semantic features are decoded by the second decoding unit to obtain the first decoded data.

[0332] In some embodiments of this disclosure, wherein,

[0333] The transceiver module 7201 is used to receive multiple second pieces of information, wherein different second pieces of information correspond to different first decoding units, and different first decoding units belong to different first devices.

[0334] Processing module 7202 is used to aggregate multiple pieces of second information to obtain target information;

[0335] The transceiver module 7201 is also used to send target information.

[0336] In some embodiments of this disclosure, wherein,

[0337] The transceiver module 7201 is used to receive the loss value, wherein the loss value is determined based on semantic data and second decoded data, and the second decoded data is obtained by processing the first decoded data.

[0338] In some embodiments of this disclosure, the processing module 7202 is used for:

[0339] If the loss value meets the loss condition, the training of the second decoding unit is complete.

[0340] The loss value does not meet the loss condition, indicating that the second decoding unit has not been fully trained.

[0341] The second decoding unit is then retrained based on the loss value until the second decoding unit is fully trained.

[0342] In some embodiments of this disclosure, wherein,

[0343] The processing module 7202 is used to process the loss value through the fine-tuning layer to obtain the first adjustment information, wherein the first adjustment information is used to adjust the parameters of the first decoding unit in the first device;

[0344] The transceiver module 7201 is used to send the first adjustment information.

[0345] In some embodiments of this disclosure, the processing module 7202 is used for:

[0346] The loss value is processed by a fine-tuning layer to obtain the second adjustment information;

[0347] The parameters of the second decoding unit are adjusted according to the second adjustment information.

[0348] In some embodiments of this disclosure, the second device further includes: a third decoding unit; wherein the processing module 7202 is further configured to:

[0349] Configure the third decoding unit according to the target information;

[0350] The decoder is generated based on the configuration of the third and second decoding units.

[0351] In some embodiments of this disclosure, the second device is an edge cloud device.

[0352] In some embodiments, the transceiver module may include a transmitting module and / or a receiving module, which may be separate or integrated. Optionally, the transceiver module may be interchangeable with a transceiver.

[0353] In some embodiments, the processing module may be a single module or may include multiple sub-modules. Optionally, the multiple sub-modules may each perform all or part of the steps required by the processing module. Optionally, the processing module may be interchangeable with a processor.

[0354] Figure 8A is a schematic diagram of the structure of the communication device proposed in an embodiment of this disclosure. The communication device 8100 can be a terminal, a network device, a chip, chip system, or processor that supports the terminal in implementing any of the above methods, or a chip, chip system, or processor that supports the network device in implementing any of the above methods. The communication device 8100 can be used to implement the methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0355] As shown in Figure 8A, the communication device 8100 includes one or more processors 8101. The processor 8101 can be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit (CPU). The baseband processor can be used to process communication protocols and communication data, while the CPU can be used to control communication devices (e.g., base stations, baseband chips, terminal devices, terminal device chips, DUs or CUs, etc.), execute programs, and process program data. The communication device 8100 is used to execute any of the above methods.

[0356] In some embodiments, the communication device 8100 further includes one or more memories 8102 for storing instructions. Optionally, all or part of the memories 8102 may also be located outside the communication device 8100.

[0357] In some embodiments, the communication device 8100 further includes one or more transceivers 8103. When the communication device 8100 includes one or more transceivers 8103, the transceivers 8103 perform at least one of the communication steps such as sending and / or receiving in the above method, and the processor 8101 performs other steps.

[0358] In some embodiments, a transceiver may include a receiver and / or a transmitter, which may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, transceiver circuit, etc., may be used interchangeably; the terms transmitter, transmitting unit, transmitter, transmitting circuit, etc., may be used interchangeably; and the terms receiver, receiving unit, receiver, receiving circuit, etc., may be used interchangeably.

[0359] In some embodiments, the communication device 8100 may include one or more interface circuits 8104. Optionally, the interface circuit 8104 is connected to the memory 8102, and the interface circuit 8104 can be used to receive signals from the memory 8102 or other devices, and can be used to send signals to the memory 8102 or other devices. For example, the interface circuit 8104 can read instructions stored in the memory 8102 and send the instructions to the processor 8101.

[0360] The communication device 8100 described in the above embodiments may be a terminal, a network device, or a third entity, but the scope of the communication device 8100 described in this disclosure is not limited thereto, and the structure of the communication device 8100 may not be limited by FIG8A. The communication device may be a standalone device or may be part of a larger device. For example, the communication device may be: (1) a standalone integrated circuit IC, or chip, or chip system or subsystem; (2) a collection of one or more ICs, optionally, the IC collection may also include storage components for storing data and programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, smart terminal device, cellular phone, wireless device, handheld device, mobile unit, vehicle device, network device, cloud device, artificial intelligence device, etc.; (6) others, etc.

[0361] Figure 8B is a schematic diagram of the chip structure proposed in an embodiment of this disclosure. For cases where the communication device 8100 can be a chip or a chip system, please refer to the schematic diagram of the chip 8200 shown in Figure 8B, but it is not limited thereto.

[0362] Chip 8200 includes one or more processors 8201, which are used to perform any of the above methods.

[0363] In some embodiments, chip 8200 further includes one or more interface circuits 8202. Optionally, the interface circuit 8202 is connected to memory 8203, and the interface circuit 8202 can be used to receive signals from memory 8203 or other devices, and the interface circuit 8202 can be used to send signals to memory 8203 or other devices. For example, the interface circuit 8202 can read instructions stored in memory 8203 and send the instructions to processor 8201.

[0364] In some embodiments, the interface circuit 8202 performs at least one of the communication steps such as sending and / or receiving in the above method, while the processor 8201 performs other steps.

[0365] In some embodiments, the terms interface circuit, interface, transceiver pin, transceiver, etc., can be used interchangeably.

[0366] In some embodiments, chip 8200 further includes one or more memories 8203 for storing instructions. Optionally, all or part of the memories 8203 may be located outside of chip 8200.

[0367] This disclosure also proposes a storage medium storing instructions that, when executed on a communication device 8100, cause the communication device 8100 to perform any of the above methods. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but not limited thereto; it may also be a storage medium readable by other devices. Optionally, the storage medium may be a non-transitory storage medium, but not limited thereto; it may also be a temporary storage medium.

[0368] This disclosure also provides a program product that, when executed by the communication device 8100, causes the communication device 8100 to perform any of the above methods. Optionally, the program product is a computer program product.

[0369] This disclosure also proposes a computer program that, when run on a computer, causes the computer to perform any of the above methods.

[0370] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer programs. When the computer program is loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this disclosure are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer program can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program can be transferred from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0371] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0372] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0373] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A communication control method characterized by comprising: Performed by a first device, the first device comprising at least: a first decoding unit; wherein, the method includes: Receive first decoded data, wherein the first decoded data is determined by the second device based on semantic data; The first decoding unit is trained at least once based on the first decoded data and the semantic data until the first decoding unit is trained. During this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device. Receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer; and An encoder is generated based on the first information.

2. The method of claim 1, wherein, The first device further includes: a first encoding unit and a second encoding unit; wherein, generating the encoder based on the first information includes: Configure the first encoding unit according to the first information; The encoder is generated based on the configured first encoding unit and the second encoding unit.

3. The method of claim 2, wherein, The method further includes: The semantic features of the semantic data are extracted using the second encoding unit; The semantic features are sent, wherein the first decoded data is obtained by the second device decoding the semantic features.

4. The method according to any one of claims 1 to 3, characterized in that, The step of training the first decoding unit at least once based on the first decoded data and the semantic data until the first decoding unit is trained successfully includes: Based on the first decoded data and the semantic data, the first decoding unit is trained this time to obtain the loss value; If the loss value satisfies the loss condition, it is determined that the training of the first decoding unit is complete. If the loss value does not meet the loss condition, it is determined that the first decoding unit has not been trained. The first decoding unit is then retrained based on the loss value until the first decoding unit is fully trained.

5. The method of claim 4, wherein, The step of training the first decoding unit based on the first decoded data and the semantic data to obtain a loss value includes: The first decoded data is processed by the first decoding unit to obtain the second decoded data; The loss value is determined based on the semantic data and the second decoded data.

6. The method of claim 5, wherein, The method further includes: Send a second message, wherein the second message is used to indicate the parameters of the first decoding unit; Receive target information, wherein the target information is obtained by aggregating multiple second information, different second information corresponds to different first decoding units, and different first decoding units belong to different first devices.

7. The method according to any one of claims 4 to 6, wherein, The step of retraining the first decoding unit based on the loss value includes: Based on the loss value, determine the first adjustment information; The parameters of the first decoding unit are adjusted based on the first adjustment information and the target information; The first decoding unit is obtained through retraining and adjustment.

8. The method of claim 7, wherein, Determining the first adjustment information based on the loss value includes: The loss value is sent, wherein the loss value is used by the fine-tuning layer to determine the first adjustment information; Receive the first adjustment information.

9. The method according to any one of claims 1 to 8, wherein, The first device is a vehicle.

10. A communication control method characterized by comprising: Performed by a second device, the second device comprising: a fine-tuning layer and a second decoding unit; wherein, the method includes: The semantic data is processed by the second decoding unit to obtain the first decoded data, wherein the first decoded data is used to determine the loss value; Send the first decoded data; The second decoding unit is trained at least once based on the loss value until the training of the second decoding unit is completed. During this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining. Send a first message, wherein the first message is used to indicate the parameters of the fine-tuning layer.

11. The method of claim 10, wherein, The step of processing semantic data through the second decoding unit to obtain the first decoded data includes: The semantic features of the received semantic data; The semantic features are decoded by the second decoding unit to obtain the first decoded data.

12. The method according to any one of claims 10-11, characterized in that, The method further includes: Receive multiple pieces of second information, wherein different pieces of second information correspond to different first decoding units, and different first decoding units belong to different first devices; The target information is obtained by aggregating the multiple pieces of second information. Send the target information.

13. The method of claim 12, wherein, The method further includes: The loss value is received, wherein the loss value is determined based on the semantic data and the second decoded data, and the second decoded data is obtained by processing the first decoded data.

14. The method according to any one of claims 10 to 13, wherein, The step of training the second decoding unit at least once based on the loss value until the training of the second decoding unit is completed includes: If the loss value satisfies the loss condition, it is determined that the training of the second decoding unit is complete. If the loss value does not meet the loss condition, it is determined that the second decoding unit has not been trained. The second decoding unit is then retrained based on the loss value until the second decoding unit is fully trained.

15. The method according to any one of claims 10 to 14, wherein, The method further includes: The loss value is processed by the fine-tuning layer to obtain first adjustment information, wherein the first adjustment information is used to adjust the parameters of the first decoding unit in the first device; Send the first adjustment information.

16. The method according to any one of claims 10 to 15, wherein, The parameter adjustment of the second decoding unit based on the fine-tuning layer includes: The loss value is processed by the fine-tuning layer to obtain the second adjustment information; The parameters of the second decoding unit are adjusted according to the second adjustment information.

17. The method of any one of claims 12-16, wherein, The second device further includes: a third decoding unit; wherein, the method further includes: Configure the third decoding unit according to the target information; A decoder is generated based on the configured third decoding unit and the second decoding unit.

18. The method of any one of claims 10-17, wherein, The second device is an edge cloud device.

19. A communication control method characterized by comprising: The method includes: The second device processes semantic data through the second decoding unit to obtain first decoded data and sends the first decoded data, wherein the first decoded data is used to determine the loss value; The first device trains the first decoding unit at least once based on the first decoded data and the semantic data until the first decoding unit is trained. During this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer. The adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device. The second device trains the second decoding unit at least once based on the loss value until the second decoding unit is trained. During this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted parameters are... The second decoding unit is used for retraining; and for sending first information, wherein the first information is used to indicate the parameters of the fine-tuning layer; The first device generates an encoder based on the first information.

20. A first device, comprising: The first device includes: The transceiver module is used to receive first decoded data, wherein the first decoded data is determined by the second device based on semantic data; and to receive first information, wherein the first information is used to indicate the parameters of the fine-tuning layer. The processing module is configured to train a first decoding unit at least once based on the first decoded data and the semantic data until the first decoding unit is trained. In this training, the parameters of the first decoding unit are adjusted based on the fine-tuning layer, and the adjusted first decoding unit is used for retraining. The fine-tuning layer belongs to the second device. The module also generates an encoder based on the first information.

21. A second device, comprising: The second device includes: The processing module is used to process semantic data through the second decoding unit to obtain first decoded data, wherein the first decoded data is used to determine the loss value; and to train the second decoding unit at least once based on the loss value until the second decoding unit is trained, wherein in this training, the parameters of the second decoding unit are adjusted based on the fine-tuning layer, and the adjusted second decoding unit is used for retraining. The transceiver module is used to send the first decoded data and send first information, wherein the first information is used to indicate the parameters of the fine-tuning layer.

22. A communications device, characterized by include: One or more processors; The processor is used to execute the communication control method according to any one of claims 1-18.

23. A storage medium, the storage medium storing instructions, wherein, When the instruction is executed on the communication device, it causes the communication device to perform the communication control method as described in any one of claims 1-18.

24. A computer program product, characterised in that, It includes a computer program that, when executed by a processor, implements the communication control method according to any one of claims 1-18.

Images