Communication method and device

Abstract

The invention provides a communication method and device, and relates to the technical field of communication. In the method, after the model application equipment indicates the first state of one or more first models to the model management equipment, the first state can be maintained in the first time window, so that the condition that the model application equipment frequently reports the first state of one or more first models to the model management equipment is avoided; and information interaction between the model application equipment and the model management equipment is reduced.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a communication method and device. Background Art

[0002] The protocol proposes a solution for using models (such as artificial intelligence (AI) / machine learning (ML) models / functions) for air interface enhancement. The use of models for air interface enhancement includes the process of obtaining training data sets, the process of training and loading models, and the process of model management.

[0003] At present, one way to manage models is that the terminal device determines the information of the model status, feeds back the information of the model status to the network (NW), and the NW executes the model management strategy. In this way, when the model status of the terminal device changes, the information of the model status needs to be fed back to the NW, which makes the information exchange between the terminal device and the NW more frequent. When the number of models involved in the terminal device is large, the information exchange between the terminal device and the NW will be further increased. Summary of the invention

[0004] The embodiments of the present application provide a communication method and device to reduce information interaction.

[0005] In the first aspect, an embodiment of the present application provides a communication method, which can be executed by a model application device, or software or hardware (such as a chip) in the model application device. By way of example, the model application device can be one or more of a terminal device, an access network device (such as a base station), or a core network device, etc. For ease of description, the following description takes the execution of the method by a terminal device as an example. It can be understood that the terminal devices in the following description can be replaced with other model application devices, such as an access network device or a core network device. The method includes: sending a first message to a model management device, wherein the first message indicates that the state of one or more first models is a first state, and the first state indicates that the terminal device supports the use of the one or more first models, wherein the first state is valid within a first time window.

[0006] It should be understood that the model involved in the embodiment of the present application can be described as a function (such as an AI function or an ML function), a characteristic or an algorithm, etc., and the model includes, for example, an AI model or an ML model, etc. The terminal device supports the use of one or more first models, which can be described as the terminal device being able to execute or run one or more first models, or can be described as the terminal device supporting the execution of one or more first models, or can be described as the terminal device having the ability to execute one or more first models, or the terminal device supporting the execution (or running) of one or more first models, etc. The terminal device supports the use of one or more first models, for example, including the terminal device having locally loaded the configuration files of the one or more first models, or the terminal device can load the configuration files of the one or more first models from other devices (such as a configuration file management device, specifically, a cloud server). The first state is valid within the first time window, which can be understood as the terminal device not changing the state of one or more first models (i.e., the first state) within the first time window, or can be described as the terminal device maintaining the state of one or more first models within the first time window.

[0007] Optionally, the first time window may be preconfigured or predefined in the terminal device and the model management device, for example, preconfigured or predefined by a protocol, for example, at least one of the length (or window length), starting position (or starting time) and ending position (or ending time) of the first time window is preconfigured or predefined by a protocol, or the terminal device determines and notifies the model management device, for example, the terminal device sends second indication information to the model management device, and the second indication information indicates the first time window, for example, the first indication information indicates at least one of the length, starting position and ending position of the first time window, and the first indication information indicates the length, starting position and ending time of the first time window. In the case of one of the beam positions, at least one of the other two of the length, starting position and ending position of the first time window can be configured through the protocol, or the model management device determines and notifies the terminal device, for example, the model management device sends a third indication information to the terminal device, and the third indication information indicates the first time window. For example, the third indication information indicates at least one of the length, starting position and ending position of the first time window. In the case where the first indication information indicates one of the length, starting position and ending position of the first time window, at least one of the other two of the length, starting position and ending position of the first time window can be configured through the protocol, and the embodiments of the present application are not limited to this.

[0008] In an embodiment of the present application, the terminal device may not change the first state of one or more first models within the first time window, which means that the terminal device does not need to send the first state of the one or more first models to the model management device within the first time window, thereby reducing the information interaction between the terminal device and the model management device, and also reducing the information interaction in the network. In addition, since the information interaction between the terminal device and the model management device is reduced, it is also beneficial to save power of the terminal device and the model management device. In addition, since the model management device does not need to inquire the terminal device about the first state of the one or more first models within the first time window, it is equivalent to that the model management device can directly manage the one or more first models within the first time window based on the first information, which is beneficial to improve the efficiency of the model management device in managing the one or more first models of the terminal device.

[0009] In a possible implementation, the method further includes: after the first time window, determining that the status of M first models among the one or more first models is updated to a second state, the second state indicating that the terminal device does not support the use of the M first models, M is a positive integer; sending second information to the model management device, the second information indicating that the status of the M first models / functions is updated to the second state.

[0010] Optionally, the second state is valid within a fourth time window, and the fourth time window may be the same as or different from the first time window. The fourth time window is the same as the first time window, including the same starting time, length, and end time. The fourth time window is different from the first time window, including at least one of the following: the starting position of the fourth time window is different from the starting position of the first time window, the length of the fourth time window is different from the length of the first time window, and the end position of the fourth time window is different from the end position of the first time window. The way in which the model management device and the terminal device determine the fourth time window can refer to the content of determining the first time window in the previous text, and will not be listed here.

[0011] In the above embodiment, after the first time window, the terminal device can flexibly update the status of M first models among one or more first models. In this way, the information interaction between the terminal device and the model management device is relatively reduced, and the terminal device can be relatively guaranteed to have the flexibility to change the status of the M first models.

[0012] In a possible implementation, before sending the second information to the model management device, the method further includes: determining that a first moment has arrived, wherein the time interval between the first moment and the moment of sending the first information satisfies a first cycle, and the first cycle is a cycle for the terminal device to report the status of the M first models. Optionally, the length of the first time window is less than or equal to the length of the first cycle.

[0013] In the above implementation, the terminal device can report the status of the M first models according to the first cycle, which can also reduce the information interaction between the terminal device and the model management device to a certain extent. In addition, the length of the first time window is less than or equal to the length of the first cycle, which means that within the first cycle, the terminal device is more likely to change the status of the M models, which relatively reduces the possibility of the terminal device reporting the same status of the M first models multiple times, and reduces redundant interactions in the network.

[0014] In a possible implementation, the method further includes: after the first time window and when the first moment has been reached, determining that M first models among the states of the one or more first models have not been updated, wherein the time interval between the first moment and the moment of sending the first information satisfies a first period, the first period is a period for the terminal device to report the states of the M first models, M is a positive integer, and determining to skip sending the states of the M first models to the model management device at the first moment.

[0015] It should be understood that determining the state of skipping sending M first models to the model management device at the first moment can be described as skipping the state of sending M first models to the model management device at the first moment, or can be described as giving up the state of sending M first models to the model management device at the first moment, or can be understood as releasing the first resource, and the first resource is used to send (or report) the state of M first models to the model management device at the first moment.

[0016] In the above implementation, the terminal device reports the status of the M first models with reference to the first period, but at the reporting time (such as the first time), the terminal device determines that the status of the M first models has not changed, then the terminal device does not need to send the status of the M first models to the model management device. In this way, the possibility of the terminal device reporting the status of the same M first models multiple times is relatively reduced, and redundant interactions in the network are reduced.

[0017] In a possible implementation, the method further includes: after the first time window, receiving first indication information from the model management device, the first indication information indicating reporting the status of some or all of the one or more first models; and sending third information to the model management device, the third information indicating the status of some or all of the first models.

[0018] In the above implementation, the model management device can actively trigger the terminal device to report the status of part or all of the first models in one or more first models. In this way, the terminal device reports the third information only when the model management device has a demand, which is also conducive to reducing the invalid reporting of the status of part or all of the first models by the terminal device.

[0019] In a possible implementation, the method further includes: sending fourth information to the model management device, wherein the fourth information indicates that the state of one or more second models is a third state, and the third state indicates that the terminal device does not support the use of the one or more second models. Not supporting the use of one or more second models may be that the terminal device has not loaded the configuration files of the one or more second models locally, and is unable to obtain the configuration files of the one or more second models from other devices. Optionally, the first information and the fourth information may be carried in the same information, or may be carried in different information, which is not limited in this embodiment of the present application.

[0020] In the above implementation, the terminal device may also indicate the third state of one or more second models to the model management device, so that the model management device can obtain more information and manage the models in the terminal device more comprehensively and accurately.

[0021] In a possible implementation, the third state is valid within a second time window, and the second time window may be the same as or different from the first time window. The second time window and the first time window are the same including the starting time of the second time window and the first time window, the length of the second time window and the first time window, and the ending position of the second time window and the first time window are the same. The second time window and the first time window are different including at least one of the starting position of the second time window and the starting position of the first time window are different, the length of the second time window and the length of the first time window are different, and the ending position of the second time window is different from the ending position of the first time window. The way in which the model management device and the terminal device determine the second time window can refer to the content of determining the first time window in the previous text, which is not listed here.

[0022] In the above implementation, the terminal device can maintain the third state within the second time window, so that the information interaction between the terminal device and the model management device can be reduced. In addition, when the second time window is different from the first time window, the terminal device can relatively flexibly control the state of one or more first models and one or more second models.

[0023] In a possible implementation, the method further includes: sending fifth information to the model management device, the fifth information indicating that the status of some or all of the one or more second models is updated to a fourth status, the fourth status indicating that the terminal device supports the use of some or all of the second models, wherein the difference between the time when the terminal device sends the fifth information and the time when the terminal device sends the first information is less than the length of the first time window.

[0024] In the above implementation, the terminal device may send the updated status of part or all of the second model to the model management device when determining that the status of part or all of the second model is updated, thereby improving the flexibility of the terminal device in reporting the status of part or all of the second model.

[0025] In a possible implementation, the fourth state is valid within a third time window, wherein the third time window is the same as or different from the first time window. The difference between the third time window and the first time window may include at least one of a start time of the third time window being different from a start time of the first time window, a length of the third time window being different from a length of the first time window, and an end time of the third time window being different from an end time of the first time window. The manner in which the model management device and the terminal device determine the third time window may refer to the content of determining the first time window in the foregoing text, and will not be listed here.

[0026] In the above implementation, the terminal device can maintain the fourth state within the third time window, so that the information interaction between the terminal device and the model management device can be reduced. In addition, when the third time window is different from the first time window, the terminal device can relatively flexibly control the state of one or more first models and one or more second models.

[0027] In a possible implementation, the first information includes first delay information, and the first delay information indicates one or more delays associated with some or all of the one or more first models; wherein some or all of the first delay information indicates the first state. For example, some or all of the multiple bits included in the first delay information indicate the first state, or some or all of the multiple information included in the first delay information indicates the first state.

[0028] In the above implementation, the first delay information can be reused to indicate one or more delays related to part or all of the first model, as well as the first state. This can save bits occupied by the first information and reduce information interaction between the terminal device and the model management device.

[0029] In a possible implementation, the first information also includes: an identifier of the part or all of the first model; or, configuration information related to the part or all of the first model. The configuration information related to the part or all of the first model may include an identifier (or index) of the configuration related to the part or all of the first model. The configuration related to the part or all of the first model includes, for example, one or more of the following: the configuration of the input of the part or all of the first model (such as resources), the configuration of the output of the part or all of the first model (such as reporting the resources of the output of the part or all of the first model), the functions implemented by the part or all of the first model and the scenarios to which the part or all of the first model is applicable, and the information of the terminal device used to support the related functions of the part or all of the model, etc.

[0030] In a possible implementation manner, the first delay information includes at least one of the following: first activation delay information, the first activation delay information indicating K1 first activation delays, wherein one first activation delay indicates the delay of activating one of the K1 first models; first inference delay information, the first inference delay information indicating K2 first inference delays, wherein one first inference delay indicates the delay of inferring one of the K2 first models; first fallback delay information, the first fallback delay information indicating K3 first fallback delays, wherein one first fallback delay indicates the delay of falling back one of the K3 first models to a delay of implementing a first function without using a model, and the one first model can implement the first function; first monitoring delay information, wherein the first monitoring delay information indicates K3 first fallback delays, wherein one first fallback delay indicates the delay of implementing a first function without using a model; and The first monitoring delay information indicates K4 first monitoring delays, wherein one first monitoring delay indicates the delay of monitoring one of the K4 first models; the first switching delay information, wherein the first switching delay information indicates K5 first switching delays, wherein one first switching delay indicates the delay of switching one of the K5 first models to the third model; the second switching delay information, wherein the second switching delay information indicates K6 second switching delays, wherein one second switching delay indicates switching the fourth model to one of the K6 first models; or the first deactivation delay information, wherein the first deactivation delay information indicates K7 first deactivation delays, wherein one first deactivation delay indicates the delay of deactivating one of the K7 first models. Wherein, K1, K2, K3, K4, K5, K6 and K7 are all positive integers.

[0031] In the above implementation, multiple possible information included in the first delay information is provided, so that the model management device can obtain more comprehensive information, so as to more accurately manage the model in the terminal device. In addition, the terminal device can use at least one of the multiple possible information included in the first delay information to indicate the first state, which improves the flexibility of indicating the first delay information to indicate the first state.

[0032] In a possible implementation manner, part or all of the first delay information indicates the first state, including at least one of the following: the first activation delay information indicates that the states of P first models among the K1 first models are the first state, where P is a positive integer. For example, the values ​​of part or all of the first activation delays (such as P or less than P first activation delays) among the K1 first activation delays are all within a first value range, and the corresponding states of the P first models are the first state; the first switching delay information indicates that the states of Q first models among the K5 first models are the first state, where Q is a positive integer. For example, the values ​​of part or all of the first switching delays (such as less than or equal to Q first switching delays) among the K5 first switching delays are all within a second value range. Then the corresponding state of the Q first models is the first state; the second switching delay information indicates that the state of R first models among the K6 first models is the first state, R is a positive integer, for example, the values ​​of some or all of the second switching delays (such as R or less than R second switching delays) in the K6 second switching delays are in the third value range, then the corresponding state of the R first models is the first state; or, the first inference delay information indicates that the state of L first models among the K2 first models is the first state, L is a positive integer, for example, the values ​​of some or all of the first inference delays (such as less than or equal to L first inference delays) in the K2 first inference delays are in the fourth value range, then the corresponding state of the L first models is the first state.

[0033] In the above implementation, at least one of the first activation delay information, the first switching delay information, the second switching delay information, and the first inference delay information can be used to indicate the first state of some or all of the first models, which enriches the way of indicating the first state and reduces the signaling overhead. Specifically, the value of at least one of the activation delay, the first switching delay, the second switching delay, and the inference delay can be used to indicate the first state of some or all of the first models in one or more first models. In this way, the first state of some or all of the first models in some or all of the first models does not need to occupy the bits of the first information, which can further reduce the information interaction between the terminal device and the model management device.

[0034] In one possible implementation, the length of the first time window satisfies at least one of the following conditions: greater than the maximum value in the first value range; greater than the maximum value in the second value range; greater than the maximum value in the third value range; or greater than the maximum value in the fourth value range.

[0035] In the above embodiment, the length of the first time window may be greater than the maximum value of the value range of at least one of the activation delay, the first switching delay, the second switching delay and the inference delay, thereby avoiding the situation where the first time window is set too short and the terminal device does not have time to perform operations such as activation, switching or inference on the model.

[0036] In a possible implementation, the method further includes: sending sixth information to the terminal device within the first time window, the sixth information indicating a strategy for managing N first models among the one or more first models, where N is a positive integer.

[0037] It should be understood that the strategies for managing any two of the N first models may be the same or different. The strategies for managing the N first models include, for example, activating some or all of the N first models, deactivating some or all of the N first models, switching some or all of the N first models to other models (other models may be, for example, models in one or more second models, or models in one or more first models other than the N first models, etc.), and switching other models to some or all of the N first models, etc., and the embodiments of the present application do not limit this.

[0038] In one possible implementation, the fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with some or all of the one or more second models; wherein some or all of the second delay information also indicates the third state.

[0039] In the above implementation, the second delay information can be used to indicate the third state, which can save bits occupied by the fourth information and reduce information interaction between the terminal device and the model management device.

[0040] In a possible implementation, the second delay information includes at least one of the following: second activation delay information, the second activation delay information indicates H1 second activation delays, one of the second activation delays indicates the delay of activating one of the H1 second models; second inference delay information, the second inference delay information indicates H2 second inference delays, one of the second inference delays indicates the delay of inferring one of the H2 second models; second fallback delay information, the second fallback delay information indicates H3 second fallback delays, one of the second fallback delays indicates the delay of falling back one of the H3 second models to a delay of not using a model to implement a third function, and the one second model can implement the third function; second monitoring delay information, the second monitoring delay information indicates H4 second monitoring delay, wherein one second monitoring delay indicates the delay of monitoring one of the H4 second models; third switching delay information, wherein the third switching delay information indicates H5 third switching delays, wherein one third switching delay indicates the delay of switching one of the H5 second models to the third model; fourth switching delay information, wherein the fourth switching delay information indicates H6 fourth switching delays, wherein one fourth switching delay indicates switching the fourth model to one of the H6 second models; or, second deactivation delay information, wherein the second deactivation delay information indicates H7 second deactivation delays, wherein one second deactivation delay indicates the delay of deactivating one of the H7 second models, wherein H1, H2, H3, H4, H5, H6, and H7 are all positive integers.

[0041] In a possible implementation, part or all of the second delay information indicates the third state, including at least one of the following: the second activation delay information indicates that the state of U1 second model among the H1 second models is the third state, and U1 is a positive integer; the third switching delay information indicates that the state of U2 second model among the H5 second models is the third state, and U2 is a positive integer; the fourth switching delay information indicates that the state of U3 second model among the H6 second models is the third state, and U3 is a positive integer; or, the second inference delay information indicates that the state of U4 second model among the H2 second models is the third state, and U4 is a positive integer.

[0042] In the above implementation, at least one of the second activation delay information, the third switching delay information, the fourth switching delay information and the second inference delay information can be used to indicate the third state of part or all of the second models in part or all of the second model, thereby enriching the method of indicating the third state and reducing signaling overhead.

[0043] In a possible implementation, the values ​​of some or all of the H1 second activation delays (such as less than or equal to U1 second activation delays) are all within the fifth value range, and the corresponding state of the U1 second model is the third state; the values ​​of some or all of the H5 third switching delays (such as less than or equal to U5 third switching delays) are all within the sixth value range, and the corresponding state of the U5 second model is the third state; the values ​​of some or all of the H6 fourth switching delays (such as less than or equal to U3 fourth switching delays) are within the seventh value range, and the corresponding state of the U3 second models is the third state; the values ​​of some or all of the H2 second inference delays (such as less than or equal to U4 inference delays) are within the eighth value range, and the corresponding state of the U4 second models is the third state.

[0044] In the above embodiment, the value of at least one of the second activation delay, the third switching delay, the fourth switching delay and the second inference delay can be used to indicate the third state of part or all of the first model. In this way, the third state of part or all of the second model does not need to occupy the bits of the fourth information, which can further reduce the information interaction between the terminal device and the model management device.

[0045] In one possible implementation, the second time window satisfies at least one of the following conditions: greater than the maximum value in the fifth value range; greater than the maximum value in the sixth value range; greater than the maximum value in the seventh value range; or greater than the maximum value in the eighth value range.

[0046] In a possible implementation, the method further includes: after the first time window, sending seventh information to the model management device, wherein the seventh information indicates the status of some or all of the one or more first models. It should be understood that the some or all of the first models may be the first models whose status is updated in the one or more first models, or may be the one or more first models.

[0047] Optionally, the terminal device sends the seventh information at a second moment, and the time interval between the second moment and the moment of sending the first information satisfies a first period, and the first period is a period in which the terminal device reports the status of some or all of the one or more first models.

[0048] In a possible implementation, the method further includes: after the first time window, sending eighth information to the model management device, wherein the eighth information indicates the status of some or all of the one or more second models. It should be understood that the some or all of the second models may be the second models whose status is updated in the one or more second models, or may be the one or more second models.

[0049] Optionally, the terminal device sends the eighth information at a third moment, and the time interval between the third moment and the moment of sending the first information satisfies a first period, and the first period is a period in which the terminal device reports the status of some or all of the one or more second models.

[0050] In a second aspect, an embodiment of the present application provides a communication method, which can be executed by a model application device, or software or hardware (such as a chip) in the model application device. For example, the model application device can be one or more of a terminal device, an access network device (such as a base station), or a core network device, etc. For the sake of ease of description, the following description takes the execution of the method by a terminal device as an example. It can be understood that the terminal devices in the following description can be replaced with other model application devices, such as an access network device or a core network device. The method includes: sending a first message to a model management device, wherein the first message includes a first delay information, and the first delay information indicates one or more delays associated with some or all of the one or more first models; wherein some or all of the first delay information indicates a first state, and the first state indicates that the terminal device supports the use of the one or more first models.

[0051] In an embodiment of the present application, the first delay information in the first information can indicate part or all of the delay related to the first model, and can also indicate the first state. In this way, there is no need to use a special bit to indicate the first state, which is beneficial to saving bits occupied by the first information and reducing information interaction between the terminal device and the model management device.

[0052] In a possible implementation, the first information also includes an identifier of the part or all of the first model; or, configuration information related to the part or all of the first model. The configuration information related to the part or all of the first model may include an identifier (or index) of the configuration related to the part or all of the first model. The configuration related to the part or all of the first model includes, for example, one or more of the following: the configuration of the input of the part or all of the first model (such as resources), the configuration of the output of the part or all of the first model (such as reporting the resources of the output of the part or all of the first model), the functions implemented by the part or all of the first model and the scenarios to which the part or all of the first model is applicable, and the information of the terminal device used to support the related functions of the part or all of the model, etc.

[0053] In a possible implementation manner, the first delay information includes at least one of the following: first activation delay information, the first activation delay information indicating K1 first activation delays, wherein one first activation delay indicates the delay of activating one of the K1 first models; first inference delay information, the first inference delay information indicating K2 first inference delays, wherein one first inference delay indicates the delay of inferring one of the K2 first models; first fallback delay information, the first fallback delay information indicating K3 first fallback delays, wherein one first fallback delay indicates the delay of falling back one of the K3 first models to a delay of implementing a first function without using a model, and the one first model can implement the first function; first monitoring delay information, wherein the first monitoring delay information indicates K3 first fallback delays, wherein one first fallback delay indicates the delay of implementing a first function without using a model; and The first monitoring delay information indicates K4 first monitoring delays, wherein one first monitoring delay indicates the delay of monitoring one of the K4 first models; the first switching delay information, wherein the first switching delay information indicates K5 first switching delays, wherein one first switching delay indicates the delay of switching one of the K5 first models to the third model; the second switching delay information, wherein the second switching delay information indicates K6 second switching delays, wherein one second switching delay indicates switching the fourth model to one of the K6 first models; or the first deactivation delay information, wherein the first deactivation delay information indicates K7 first deactivation delays, wherein one first deactivation delay indicates the delay of deactivating one of the K7 first models. Wherein, K1, K2, K3, K4, K5, K6 and K7 are all positive integers.

[0054] In a possible implementation manner, part or all of the first delay information indicates the first state, including at least one of the following: the first activation delay information indicates that the states of P first models among the K1 first models are the first state, where P is a positive integer. For example, the values ​​of part or all of the first activation delays (such as P or less than P first activation delays) among the K1 first activation delays are all within a first value range, and the corresponding states of the P first models are the first state; the first switching delay information indicates that the states of Q first models among the K5 first models are the first state, where Q is a positive integer. For example, the values ​​of part or all of the first switching delays (such as less than or equal to Q first switching delays) among the K5 first switching delays are all within a second value range. Then the corresponding state of the Q first models is the first state; the second switching delay information indicates that the state of R first models among the K6 first models is the first state, R is a positive integer, for example, the values ​​of some or all of the second switching delays (such as R or less than R second switching delays) in the K6 second switching delays are in the third value range, then the corresponding state of the R first models is the first state; or, the first inference delay information indicates that the state of L first models among the K2 first models is the first state, L is a positive integer, for example, the values ​​of some or all of the first inference delays (such as less than or equal to L first inference delays) in the K2 first inference delays are in the fourth value range, then the corresponding state of the L first models is the first state.

[0055] In a possible implementation manner, the first state is valid within a first time window. The content of the first time window can refer to the content of the first time window in the first aspect above, and will not be listed here.

[0056] In one possible implementation, the length of the first time window satisfies at least one of the following conditions: greater than the maximum value in the first value range; greater than the maximum value in the second value range; greater than the maximum value in the third value range; or greater than the maximum value in the fourth value range.

[0057] In a possible implementation, the method further includes: after the first time window, determining that the status of M first models among the one or more first models is updated to a second state, the second state indicating that the terminal device does not support the use of the M first models, M is a positive integer; sending second information to the model management device, the second information indicating that the status of the M first models / functions is updated to the second state.

[0058] Optionally, the second state is valid within a fourth time window, and the fourth time window may be the same as or different from the first time window.

[0059] In a possible implementation, before sending the second information to the model management device, the method further includes: determining that a first moment has arrived, wherein the time interval between the first moment and the moment of sending the first information satisfies a first cycle, and the first cycle is a cycle for the terminal device to report the status of the M first models. Optionally, the length of the first time window is less than or equal to the length of the first cycle.

[0060] In a possible implementation, the method further includes: after the first time window and when a first moment has been reached, determining that the states of the M first models of the one or more first models have not been updated, wherein the time interval between the first moment and the moment of sending the first information satisfies a first period, the first period being a period for the terminal device to report the states of the M first models in the one or more first models, M being a positive integer, and determining to skip sending the states of the M first models to the model management device at the first moment.

[0061] In a possible implementation, the method further includes: after the first time window, receiving first indication information from the model management device, the first indication information indicating reporting the status of some or all of the one or more first models; and sending third information to the model management device, the third information indicating the status of some or all of the first models.

[0062] In a possible implementation, the method further includes: sending fourth information to the model management device, wherein the fourth information indicates that the status of one or more second models is a third status, and the third status indicates that the terminal device does not support the use of the one or more second models.

[0063] In a possible implementation, the third state is valid within a second time window, and the second time window may be the same as or different from the first time window. The second time window and the first time window being the same includes the start time of the second time window and the first time window, the length of the second time window and the first time window, and the end time of the second time window and the first time window being the same. The second time window and the first time window being different may include at least one of the start time of the second time window being different from the start time of the first time window, the length of the second time window being different from the length of the first time window, and the end time of the second time window being different from the end time of the first time window.

[0064] In a possible implementation, the method further includes: sending fifth information to the model management device, the fifth information indicating that the status of some or all of the one or more second models is updated to a fourth status, the fourth status indicating that the terminal device supports the use of some or all of the second models, wherein the difference between the time when the terminal device sends the fifth information and the time when the terminal device sends the first information is less than the length of the first time window.

[0065] In a possible implementation, the fourth state is valid within a third time window, wherein the third time window is the same as or different from the first time window. The difference between the third time window and the first time window may include at least one of a start time of the third time window being different from a start time of the first time window, a length of the third time window being different from a length of the first time window, and an end time of the third time window being different from an end time of the first time window. The manner in which the terminal device and the model management device determine the third time window may refer to the content of determining the first time window in the first aspect, and will not be listed here.

[0066] In one possible implementation, the fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with some or all of the one or more second models; wherein some or all of the second delay information also indicates the third state.

[0067] In a possible implementation, the second delay information includes at least one of the following: the second delay information includes at least one of the following: second activation delay information, the second activation delay information indicates H1 second activation delays, one of the second activation delays indicates the delay of activating one of the H1 second models; second inference delay information, the second inference delay information indicates H2 second inference delays, one of the second inference delays indicates the delay of inferring one of the H2 second models; second fallback delay information, the second fallback delay information indicates H3 second fallback delays, one of the second fallback delays indicates the delay of falling back one of the H3 second models to a delay of not using a model to implement a third function, and the one second model can implement the third function; second monitoring delay information, the second monitoring The delay information indicates H4 second monitoring delays, wherein one second monitoring delay indicates the delay of monitoring one of the H4 second models; the third switching delay information, wherein the third switching delay information indicates H5 third switching delays, wherein one third switching delay indicates the delay of switching one of the H5 second models to the third model; the fourth switching delay information, wherein the fourth switching delay information indicates H6 fourth switching delays, wherein one fourth switching delay indicates switching the fourth model to one of the H6 second models; or, the second deactivation delay information, wherein the second deactivation delay information indicates H7 second deactivation delays, wherein one second deactivation delay indicates the delay of deactivating one of the H7 second models, wherein H1, H2, H3, H4, H5, H6, and H7 are all positive integers.

[0068] In a possible implementation, part or all of the second delay information indicates the third state, including at least one of the following: the second activation delay information indicates that the state of U1 second model among the H1 second models is the third state, and U1 is a positive integer; the third switching delay information indicates that the state of U2 second model among the H5 second models is the third state, and U2 is a positive integer; the fourth switching delay information indicates that the state of U3 second model among the H6 second models is the third state, and U3 is a positive integer; or, the second inference delay information indicates that the state of U4 second model among the H2 second models is the third state, and U4 is a positive integer.

[0069] In a possible implementation, the values ​​of some or all of the H1 second activation delays (e.g., U1 or less than U1 second activation delays) are all within the fifth value range, and the corresponding state of the U1 second model is the third state; the values ​​of some or all of the H5 third switching delays (e.g., U5 or less than U5 third switching delays) are all within the sixth value range, and the corresponding state of the U5 second model is the third state; the values ​​of some or all of the fourth switching delays (e.g., U3 or less than U3 fourth switching delays) among the H6 fourth switching delays are within the seventh value range, and the corresponding state of the U3 second model is the third state; the values ​​of some or all of the reasoning delays (e.g., U4 or less than U4 reasoning delays) among the H2 second reasoning delays are within the eighth value range, and the corresponding state of the U4 second model is the third state.

[0070] In one possible implementation, the second time window satisfies at least one of the following conditions: greater than the maximum value in the fifth value range; greater than the maximum value in the sixth value range; greater than the maximum value in the seventh value range; or greater than the maximum value in the eighth value range.

[0071] It should be noted that the contents of the above-mentioned first aspect and any possible implementation manner can be cited in the possible implementation manners of the second aspect and are not listed here again.

[0072] In a third aspect, an embodiment of the present application provides a communication method, which can be executed by a model application device, or software or hardware (such as a chip) in the model application device. For example, the model application device can be one or more of a terminal device, an access network device (such as a base station), or a core network device, etc. For the sake of convenience of description, the following description takes the execution of the method by a terminal device as an example. It can be understood that the terminal devices in the following description can be replaced with other model application devices, such as an access network device or a core network device. The method includes: sending fourth information to the model management device, wherein the fourth information includes second delay information, and the second delay information indicates one or more delays associated with some or all of the first models in the one or more second models; wherein some or all of the first delay information indicates a third state, and the third state indicates that the terminal device does not support the use of the one or more second models.

[0073] In an embodiment of the present application, the second delay information in the fourth information can indicate part or all of the delay related to the second model, and can also indicate the third state. In this way, there is no need to use a special bit to indicate the third state, which is beneficial to saving bits occupied by the fourth information and reducing the information interaction between the terminal device and the model management device.

[0074] In a possible implementation manner, the third state is valid within the second time window.

[0075] In one possible implementation, the fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with some or all of the one or more second models; wherein some or all of the second delay information also indicates the third state.

[0076] In a possible implementation, the second delay information includes at least one of the following: second activation delay information, the second activation delay information indicates H1 second activation delays, one of the second activation delays indicates the delay of activating one of the H1 second models; second inference delay information, the second inference delay information indicates H2 second inference delays, one of the second inference delays indicates the delay of inferring one of the H2 second models; second fallback delay information, the second fallback delay information indicates H3 second fallback delays, one of the second fallback delays indicates the delay of falling back one of the H3 second models to a delay of not using a model to implement a third function, and the one second model can implement the third function; second monitoring delay information, the second monitoring delay information indicates H4 second monitoring delay, wherein one second monitoring delay indicates the delay of monitoring one of the H4 second models; third switching delay information, wherein the third switching delay information indicates H5 third switching delays, wherein one third switching delay indicates the delay of switching one of the H5 second models to the third model; fourth switching delay information, wherein the fourth switching delay information indicates H6 fourth switching delays, wherein one fourth switching delay indicates switching the fourth model to one of the H6 second models; or, second deactivation delay information, wherein the second deactivation delay information indicates H7 second deactivation delays, wherein one second deactivation delay indicates the delay of deactivating one of the H7 second models, wherein H1, H2, H3, H4, H5, H6, and H7 are all positive integers.

[0077] In a possible implementation, the values ​​of some or all of the H1 second activation delays (e.g., U1 or less than U1 second activation delays) are all within the fifth value range, and the corresponding state of the U1 second model is the third state; the values ​​of some or all of the H5 third switching delays (e.g., U5 or less than U5 third switching delays) are all within the sixth value range, and the corresponding state of the U5 second model is the third state; the values ​​of some or all of the fourth switching delays (e.g., U3 or less than U3 fourth switching delays) among the H6 fourth switching delays are within the seventh value range, and the corresponding state of the U3 second model is the third state; the values ​​of some or all of the reasoning delays (e.g., U4 or less than U4 reasoning delays) among the H2 second reasoning delays are within the eighth value range, and the corresponding state of the U4 second model is the third state.

[0078] It should be noted that the contents of the above-mentioned first aspect and any possible implementation manner can be cited in the possible implementation manners of the third aspect and are not listed here again.

[0079] In a fourth aspect, an embodiment of the present application provides a communication method, which can be executed by a model application device, or software or hardware (such as a chip) in the model application device. For example, the model application device can be one or more of a terminal device, an access network device (such as a base station), or a core network device, etc. For ease of description, the following description takes the execution of the method by a terminal device as an example. It can be understood that the terminal devices in the following description can be replaced with other model application devices, such as an access network device or a core network device. The method includes: sending ninth information to a model management device, wherein the ninth information indicates one or more delays associated with one or more first models, and the ninth information is used to determine tenth information, and the tenth information is related information to N1 first models among the one or more first models, and N1 is a positive integer.

[0080] In an embodiment of the present application, the terminal device may send ninth information to the model management device, and the ninth information may be used to determine more information related to N1 first models, so that the model management device can obtain more information based on the ninth information, which is conducive to more accurate management of the N1 first models.

[0081] In an embodiment of the present application, the terminal device can send one or more time delays related to one or more first models to the model management device, so that the model management device can more accurately determine the relevant information related to N1 first models, so that the model management device can subsequently more accurately determine the tenth information.

[0082] In a possible implementation, the ninth information includes at least one of the following: activation delay information, the activation delay information indicating one or more activation delays, one of which indicates the delay of activating one of the one or more first models; reasoning delay information, the reasoning delay information indicating one or more reasoning delays, one of which indicates the delay of reasoning one of the one or more first models; fallback delay information, the fallback delay information indicating one or more fallback delays, one of which indicates the delay of falling back one of the one or more first models to a delay of not using a model to implement a first function, and the one first model can implement the first function; monitoring delay information, the monitoring delay information indicating One or more monitoring delays, wherein one monitoring delay indicates the delay of monitoring one of the one or more first models; first switching delay information, wherein the first switching delay information indicates one or more first switching delays, wherein one of the first switching delays indicates the delay of switching one of the one or more first models to a third model; second switching delay information, wherein the second switching delay information indicates one or more second switching delays, wherein one of the second switching delays indicates switching a fourth model to one of the one or more first models; or, deactivation delay information, wherein the deactivation delay information indicates one or more deactivation delays, wherein one of the deactivation delays indicates the delay of deactivating one of the one or more first models.

[0083] In one possible implementation, the ninth information includes the activation delay information; when the activation delay corresponding to the one first model is within the value range #1, the tenth information indicates the terminal device to activate the one first model; and / or, when the activation delay corresponding to the one first model is within the value range #2, the tenth information indicates the terminal device not to activate the one first model; wherein the minimum value within the value range #2 is greater than the maximum value within the value range #1.

[0084] In the above implementation, the first model with too long activation delay may not be activated, so as to avoid the situation where the terminal device takes too long to activate the first model or even fails to activate the first model.

[0085] In a possible implementation, the tenth information also indicates a first duration corresponding to the first model, wherein the first duration is the time required from the terminal device confirming the activation of the first model to the use of the first model; wherein, the smaller the activation delay corresponding to the first model, the smaller the first duration.

[0086] In the above implementation, the first duration can be configured for a first model based on the size of the activation delay of the first model, which can give full play to the advantages of the high-specification first model (for example, the activation delay of a first model is relatively short), stimulate continuous innovation in the industry, further improve network performance and user experience, and also take into account the low-specification first model (for example, the activation delay of a first model is relatively long), so as to meet the various needs of models of different specifications or terminal devices where the models are located.

[0087] In a possible implementation, after a fourth time period, first configuration information is received from the model management device, the first configuration information indicating the configuration required for the terminal device to use the first model, and the fourth time period is less than or equal to the activation delay corresponding to the first model.

[0088] In the above implementation, the model management device may send the first configuration information in a timely manner before a first model is activated, so that the terminal device can run the first model faster.

[0089] In one possible implementation, the ninth information includes the inference delay information; when the inference delay corresponding to the one first model is within the value range #3, the tenth information instructs the terminal device to activate the one first model; and / or, when the inference delay corresponding to the one first model is within the value range #4, the tenth information instructs the terminal device not to activate the one first model; wherein the minimum value within the value range #4 is greater than the maximum value within the value range #3.

[0090] In the above implementation, the first model with too large inference delay may not be activated to avoid the situation where the terminal device takes too long to infer the first model or even cannot use the first model for inference.

[0091] In a possible implementation, the tenth information also indicates a second duration corresponding to the one first model, wherein the second duration is the duration between detecting a third signal from the terminal device and reporting a fourth signal, the third signal is related to the input of the one first model, and the fourth signal is related to the output of the one first model; wherein, the smaller the inference delay corresponding to the one first model, the smaller the second duration. For example, in an AI-CSI prediction scenario, for terminal devices that can complete CSI prediction in a shorter time, the model management device can configure a more timely CSI feedback time for them, that is, shorten the time interval between CSI-RS transmission and CSI feedback, which is beneficial to improve CSI prediction accuracy and better assist the model management device in optimizing downlink transmission.

[0092] In the above implementation, the second duration can be configured for a first model based on the size of the inference delay of the first model, thereby meeting various requirements of models of different specifications or terminal devices where the models are located.

[0093] In a possible implementation, the ninth information includes the fallback delay information; when the fallback delay corresponding to the one first model is within the value range #5, the tenth information instructs the terminal device to perform fallback processing on the one first model; when the fallback delay corresponding to the one first model is within the value range #6, the tenth information instructs the terminal device not to perform fallback processing on the one first model; wherein the minimum value of the value range #6 is greater than the maximum value of the value range #5.

[0094] In the above implementation, the first model with a large fallback delay may not be activated to avoid the situation where the terminal device takes too long to fall back to the first model or even cannot fall back to the first model.

[0095] In one possible implementation, after the fifth time period, second configuration information is received from the model management device, wherein the second configuration information indicates that the terminal device does not use the model to implement the configuration required for the first function, and the fifth time period is less than or equal to the fallback delay corresponding to the first model.

[0096] In the above implementation, the model management device may send the second configuration information in a timely manner before rolling back a first model, so that the terminal device can execute the rolled-back service more quickly.

[0097] In a possible implementation, the ninth information includes the fallback delay information; the method also includes: sending fourth indication information to the model management device, the fourth indication information requesting fallback of the one first model; determining that within the fallback delay corresponding to the one first model, no fallback indication information is received from the model management device; and after the fallback delay corresponding to the one first model, falling back on the one first model.

[0098] In the above implementation, the terminal device and the model management device can reach a consistent understanding on the fallback of a first model based on the fallback delay information, and in this way, the terminal device can decide to fall back to the first model on its own, which is conducive to improving the autonomy of the terminal device in managing the first model.

[0099] In a possible implementation, the ninth information includes the monitoring delay information; the tenth information indicates a third time duration corresponding to the first model, the third time duration being the time duration between the terminal device detecting the fifth signal and reporting the sixth signal, wherein the smaller the monitoring delay corresponding to the first model, the smaller the third time duration, the third time duration being the time duration between the terminal device detecting the fifth signal and reporting the sixth signal, the fifth signal being used to monitor the first model, and the sixth signal being the result of monitoring the first model; or, the fifth signal being a reference signal for monitoring the first model, and the sixth signal being used to determine the result of monitoring the first model.

[0100] In the above implementation, the third duration may be configured for a first model based on the size of the monitoring delay of the first model, thereby meeting various requirements of models of different specifications or terminal devices where the models are located.

[0101] In one possible implementation, the ninth information includes the first switching delay information; when the first switching delay corresponding to the one first model is within the value range #7, the tenth information instructs the terminal device to switch the one first model to the third model; and / or, when the first switching delay corresponding to the one first model is within the value range #8, the tenth information instructs the terminal device not to switch the one first model; wherein the minimum value within the value range #8 is greater than the maximum value within the value range #7.

[0102] In the above implementation, the first model with too large a first switching delay may not be activated, so as to avoid the situation where the terminal device takes too long to switch to the first model or even fails to switch to the first model.

[0103] In a possible implementation, the ninth information includes the first switching delay information; the method also includes: sending fifth indication information to the model management device, the fifth indication information requesting to switch the one first model to the third model; determining that within the first switching delay corresponding to the one first model, the first switching indication information is not received from the model management device; after the first switching delay corresponding to the one first model, switching the one first model to the third model.

[0104] In the above implementation, the terminal device and the model management device can reach a consistent understanding on the switching of a first model based on the first switching delay, and in this way, the terminal device can decide to switch the first model on its own, which is conducive to improving the autonomy of the terminal device in managing the first model.

[0105] In a possible implementation, after the sixth time period, third configuration information is received from the model management device, the third configuration information indicating the configuration required for the terminal device to use the third model, and the sixth time period is less than or equal to the first switching delay corresponding to the first model.

[0106] In the above implementation, the model management device may issue the third configuration information in a timely manner before switching to a certain first model, so that the terminal device can switch to the first model more quickly.

[0107] In one possible implementation, the ninth information includes the second switching delay information; when the second switching delay corresponding to the one first model is within the value range #9, the tenth information instructs the terminal device to switch the fourth model to the one first model; and / or, when the second switching delay corresponding to the one first model is within the value range #10, the tenth information instructs the terminal device not to switch the fourth model; wherein the minimum value within the value range #10 is greater than the maximum value within the value range #9.

[0108] In the above implementation, the first model with a too large second switching delay may not be activated, so as to avoid the situation where the terminal device takes too long to switch to the first model or even fails to switch to the first model.

[0109] In a possible implementation, after the seventh time period, fourth configuration information is received from the model management device, the fourth configuration information indicating the configuration required for the terminal device to use the one first model, and the seventh time period is less than or equal to the second switching delay corresponding to the one first model.

[0110] In the above implementation, the model management device may issue the fourth configuration information in a timely manner before switching to a certain first model, so that the terminal device can switch to the first model more quickly.

[0111] In a possible implementation, the ninth information includes the second switching delay information; the method also includes: sending sixth indication information to the model management device, the sixth indication information requesting to switch the fourth model to the one first model; determining that within the second switching delay corresponding to the one first model, the second switching indication information is not received from the model management device; after the second switching delay corresponding to the one first model, switching the fourth model to the one first model.

[0112] In the above implementation, the terminal device and the model management device can reach a consistent understanding on the switching of a first model based on the second switching delay, and in this way, the terminal device can decide to switch the first model on its own, which is conducive to improving the autonomy of the terminal device in managing the first model.

[0113] In one possible implementation, the ninth information includes the deactivation delay information; when the deactivation delay corresponding to the one first model is within the value range #11, the tenth information instructs the terminal device to deactivate the one first model; and / or, when the deactivation delay corresponding to the one first model is within the value range #12, the tenth information instructs the terminal device not to deactivate the one first model; wherein the minimum value within the value range #11 is greater than the maximum value within the value range #12.

[0114] In one possible implementation, after an eighth time period, fifth configuration information is received from the model management device, the eighth time period is less than or equal to a deactivation delay corresponding to the one first model, and the fifth configuration information indicates the configuration required for the terminal device to implement a second function after deactivating the one first model, wherein the one first model can implement the second function.

[0115] In the above implementation, the model management device may send the fifth configuration information in time before deactivating a first model, so that the terminal device can implement the second function more quickly after deactivating the first model.

[0116] In the above implementation, the first model with too long deactivation delay may not be deactivated, so as to avoid the situation where the terminal device takes too long to deactivate the first model or even fails to deactivate the first model.

[0117] In a possible implementation, the method further includes: sending a seventh indication message to the model management device, the seventh indication message requesting deactivation of the one first model; determining that within the deactivation delay corresponding to the one first model, no deactivation indication message is received from the model management device; and deactivating the one first model after the deactivation delay corresponding to the one first model.

[0118] In the above implementation, the terminal device and the model management device can reach a consensus on the deactivation of a first model based on the deactivation delay, and in this way, the terminal device can decide to deactivate the first model on its own, which is conducive to improving the autonomy of the terminal device in managing the first model.

[0119] In a fifth aspect, an embodiment of the present application provides a communication method, which can be executed by a model application device, or software or hardware (such as a chip) in the model application device. For example, the model application device can be one or more of a terminal device, an access network device (such as a base station), or a core network device. For ease of description, the following description is taken as an example of a terminal device executing the method. It can be understood that the terminal devices in the following description can be replaced by other model application devices, such as an access network device or a core network device. The method includes: receiving an eleventh message from a model management device, the eleventh message indicating model parameters of one or more first models, wherein the model parameters of the one or more first models are valid within a fifth time window. The model parameters of one of the first models can be understood as configuration variables inside the first model, and the first model needs to use the model parameters in the reasoning process, and the values ​​of the model parameters define the available models. The values ​​of the model parameters can be obtained based on training data, or based on empirical determination, or obtained from information sources.

[0120] In the embodiment of the present application, the model management device sends one or more model parameters of the first model to the terminal device, and the terminal device does not need to pre-store the model parameters of the one or more first models. In addition, the model management device can keep the model parameters of the one or more first models unchanged within the fifth time window, which reduces the number of times the model management device sends model parameters to the terminal device, thereby reducing the information interaction between the terminal device and the model management device.

[0121] In a possible implementation, the method further includes: after the fifth time window, receiving twelfth information from the model management device, the twelfth information indicating updated model parameters of N2 first models among the one or more first models, where N2 is a positive integer.

[0122] In a sixth aspect, an embodiment of the present application provides a communication method, which can be executed by a model application device, or software or hardware (such as a chip) in the model application device. For example, the model application device can be one or more of a terminal device, an access network device (such as a base station), or a core network device, etc. For the convenience of description, the following description takes the execution of the method by a terminal device as an example. It can be understood that the terminal devices in the following description can be replaced with other model application devices, such as an access network device or a core network device. The method includes: sending a thirteenth message to a model management device, wherein the thirteenth message indicates the status of one or more first models, and the status of the one or more first models includes at least a first state or a second state, the first state indicates that the terminal device supports the use of the one or more first models, and the second state indicates that the terminal device does not support the use of the one or more models, and the first state is valid within a first time window.

[0123] In a possible implementation manner, the second state is valid within a second time window. Optionally, the second time window is the same as or different from the first time window.

[0124] The specific indication methods of the first state and the second state in this aspect can refer to the description of the first state and the third state in the first aspect, respectively, and will not be repeated here. The possible implementation methods related to the first state or the second state in the first aspect can be applied to this aspect after replacing the third state with the second state.

[0125] In a seventh aspect, an embodiment of the present application provides a communication method, which can be executed by a model management device, or can be executed by a software or hardware module (such as a chip) of the model management device. For ease of description, the following description is given by taking the model management device as executing the method. The method includes: receiving first information from a terminal device, the first information indicating that the state of the one or more models is a first state, the first state indicating that the terminal device supports the use of the one or more models, wherein the first state is valid in the first time window.

[0126] In a possible implementation, the method further includes: receiving second information from the terminal device, wherein the second information indicates that the status of M first models among the one or more first models is updated to the second state, and the second state indicates that the terminal device does not support the use of the M first models, and M is a positive integer.

[0127] In a possible implementation, the time of sending the second information is the first time, wherein the time interval between the first time and the time of sending the first information satisfies a first period, and the first period is a period for the terminal device to report the status of the M first models.

[0128] In a possible implementation, the method further includes: sending first indication information to the terminal device after the first time window, the first indication information indicating reporting the status of some or all of the one or more first models; and receiving third information from the terminal device, the third information indicating the status of some or all of the one or more first models.

[0129] In a possible implementation, the method further includes: receiving fourth information from the terminal device, the fourth information indicating that the state of one or more second models is a third state, and the third state indicates that the terminal device does not support the use of the one or more second models.

[0130] In a possible implementation, the third state is valid within a second time window, wherein the second time window is the same as or different from the first time window.

[0131] In a possible implementation, the method further includes: receiving fifth information from the terminal device, the fifth information indicating that the status of some or all of the one or more second models is updated to a fourth state, the fourth state indicating that the terminal device is capable of executing some or all of the second models, wherein the difference between the time when the terminal device sends the fifth information and the time when the terminal device sends the first information is less than the length of the first time window.

[0132] In a possible implementation manner, the fourth state is valid within a third time window, wherein the third time window is the same as or different from the first time window.

[0133] In a possible implementation, second indication information is received from a terminal device, where the second indication information indicates the first time window; or third indication information is sent to the terminal device, where the third indication information indicates the first time window.

[0134] In a possible implementation manner, the first delay information includes at least one of the following: first activation delay information, the first activation delay information indicating K1 first activation delays, wherein one first activation delay indicates the delay of activating one of the K1 first models; first inference delay information, the first inference delay information indicating K2 first inference delays, wherein one first inference delay indicates the delay of inferring one of the K2 first models; first fallback delay information, the first fallback delay information indicating K3 first fallback delays, wherein one first fallback delay indicates the delay of falling back one of the K3 first models to a delay of implementing a first function without using a model, and the one first model can implement the first function; first monitoring delay information, wherein the first monitoring delay information indicates K3 first fallback delays, wherein one first fallback delay indicates the delay of implementing a first function without using a model; and The first monitoring delay information indicates K4 first monitoring delays, wherein one first monitoring delay indicates the delay of monitoring one of the K4 first models; the first switching delay information, wherein the first switching delay information indicates K5 first switching delays, wherein one first switching delay indicates the delay of switching one of the K5 first models to the third model; the second switching delay information, wherein the second switching delay information indicates K6 second switching delays, wherein one second switching delay indicates switching the fourth model to one of the K6 first models; or the first deactivation delay information, wherein the first deactivation delay information indicates K7 first deactivation delays, wherein one first deactivation delay indicates the delay of deactivating one of the K7 first models. Wherein, K1, K2, K3, K4, K5, K6 and K7 are all positive integers.

[0135] In a possible implementation manner, part or all of the first delay information indicates the first state, including at least one of the following: the first activation delay information indicates that the states of P first models among the K1 first models are the first state, where P is a positive integer. For example, part or all of the first activation delays among the K1 first activation delays, such as P or less than P first activation delays, all have values ​​within a first value range, and the corresponding states of the P first models are the first state; the first switching delay information indicates that the states of Q first models among the K5 first models are the first state, where Q is a positive integer. For example, part or all of the first switching delays among the K5 first switching delays, such as less than or equal to Q first switching delays, all have values ​​within a second value range. , then the corresponding state of the Q first models is the first state; the second switching delay information indicates that the state of R first models among the K6 first models is the first state, R is a positive integer, for example, some or all of the second switching delays in the K6 second switching delays, such as R or less than R second switching delays, have a value in the third value range, then the corresponding state of the R first models is the first state; or, the first inference delay information indicates that the state of L first models among the K2 first models is the first state, L is a positive integer, for example, some or all of the first inference delays in the K2 first inference delays, such as less than or equal to L first inference delays, have a value in the fourth value range, then the corresponding state of the L first models is the first state.

[0136] In one possible implementation, the length of the first time window satisfies at least one of the following conditions: greater than the maximum value in the first value range; greater than the maximum value in the second value range; greater than the maximum value in the third value range; or greater than the maximum value in the fourth value range.

[0137] In one possible implementation, the fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with some or all of the one or more second models; wherein some or all of the second delay information also indicates the third state.

[0138] In a possible implementation, the second delay information includes at least one of the following: second activation delay information, the second activation delay information indicates H1 second activation delays, one of the second activation delays indicates the delay of activating one of the H1 second models; second inference delay information, the second inference delay information indicates H2 second inference delays, one of the second inference delays indicates the delay of inferring one of the H2 second models; second fallback delay information, the second fallback delay information indicates H3 second fallback delays, one of the second fallback delays indicates the delay of falling back one of the H3 second models to a delay of not using a model to implement a third function, and the one second model can implement the third function; second monitoring delay information, the second monitoring delay information indicates H4 second monitoring delay, wherein one second monitoring delay indicates the delay of monitoring one of the H4 second models; third switching delay information, wherein the third switching delay information indicates H5 third switching delays, wherein one third switching delay indicates the delay of switching one of the H5 second models to the third model; fourth switching delay information, wherein the fourth switching delay information indicates H6 fourth switching delays, wherein one fourth switching delay indicates switching the fourth model to one of the H6 second models; or, second deactivation delay information, wherein the second deactivation delay information indicates H7 second deactivation delays, wherein one second deactivation delay indicates the delay of deactivating one of the H7 second models, wherein H1, H2, H3, H4, H5, H6, and H7 are all positive integers.

[0139] In a possible implementation, part or all of the second delay information indicates the third state, including at least one of the following: the second activation delay information indicates that the state of U1 second model among the H1 second models is the third state, and U1 is a positive integer; the third switching delay information indicates that the state of U2 second model among the H5 second models is the third state, and U2 is a positive integer; the fourth switching delay information indicates that the state of U3 second model among the H6 second models is the third state, and U3 is a positive integer; or, the second inference delay information indicates that the state of U4 second model among the H2 second models is the third state, and U4 is a positive integer.

[0140] In a possible implementation, some or all of the H1 second activation delays, such as U1 or less than U1 second activation delays, have values ​​that are within the fifth value range, and the corresponding state of the U1 second model is the third state; some or all of the H5 third switching delays, such as U5 or less than U5 third switching delays, have values ​​that are within the sixth value range, and the corresponding state of the U5 second model is the third state; some or all of the H6 fourth switching delays, such as U3 or less than U3 fourth switching delays, have values ​​that are within the seventh value range, and the corresponding state of the U3 second model is the third state; some or all of the H2 second inference delays, such as U4 or less than U4 inference delays, have values ​​that are within the eighth value range, and the corresponding state of the U4 second model is the third state.

[0141] In a possible implementation, the method further includes: sending fifth information to the terminal device within the first time window, the fifth information indicating a strategy for managing N first models among the one or more first models, where N is a positive integer.

[0142] In a possible implementation, the method further includes: after the first time window, receiving seventh information from a terminal device, wherein the seventh information indicates the state of some or all of the one or more first models. It should be understood that the part or all of the first models may be the first models whose states are updated in the one or more first models, or may be the one or more first models.

[0143] Optionally, the terminal device sends the seventh information at a second moment, and the time interval between the second moment and the moment of sending the first information satisfies a first period, and the first period is a period in which the terminal device reports the status of some or all of the one or more first models.

[0144] In a possible implementation, the method further includes: after the first time window, receiving eighth information from a terminal device, the eighth information indicating the state of some or all of the one or more second models. It should be understood that the part or all of the second models may be the second models whose states are updated in the one or more second models, or may be the one or more second models.

[0145] Optionally, the terminal device sends the eighth information at a third moment, and the time interval between the third moment and the moment of sending the first information satisfies a first period, and the first period is a period in which the terminal device reports the status of some or all of the one or more second models.

[0146] In an eighth aspect, an embodiment of the present application provides a communication method, which can be executed by a model management device, or can be executed by a software or hardware module (such as a chip) of the model management device, etc. For ease of description, the following description is based on the execution of the method by a model management device. The method comprises: receiving first information from a terminal device, wherein the first information comprises first delay information, and the first delay information indicates one or more delays associated with some or all of the one or more first models; wherein some or all of the first delay information indicates a first state, and the first state indicates that the terminal device supports the use of the one or more first models.

[0147] In a possible implementation manner, the first delay information includes at least one of the following: first activation delay information, the first activation delay information indicating K1 first activation delays, wherein one first activation delay indicates the delay of activating one of the K1 first models; first inference delay information, the first inference delay information indicating K2 first inference delays, wherein one first inference delay indicates the delay of inferring one of the K2 first models; first fallback delay information, the first fallback delay information indicating K3 first fallback delays, wherein one first fallback delay indicates the delay of falling back one of the K3 first models to a delay of implementing a first function without using a model, and the one first model can implement the first function; first monitoring delay information, wherein the first monitoring delay information indicates K3 first fallback delays, wherein one first fallback delay indicates the delay of implementing a first function without using a model; and The first monitoring delay information indicates K4 first monitoring delays, wherein one first monitoring delay indicates the delay of monitoring one of the K4 first models; the first switching delay information, wherein the first switching delay information indicates K5 first switching delays, wherein one first switching delay indicates the delay of switching one of the K5 first models to the third model; the second switching delay information, wherein the second switching delay information indicates K6 second switching delays, wherein one second switching delay indicates switching the fourth model to one of the K6 first models; or the first deactivation delay information, wherein the first deactivation delay information indicates K7 first deactivation delays, wherein one first deactivation delay indicates the delay of deactivating one of the K7 first models. Wherein, K1, K2, K3, K4, K5, K6 and K7 are all positive integers.

[0148] In a possible implementation manner, part or all of the first delay information indicates the first state, including at least one of the following: the first activation delay information indicates that the states of P first models among the K1 first models are the first state, where P is a positive integer. For example, the values ​​of part or all of the first activation delays (such as less than or equal to P first activation delays) among the K1 first activation delays are all within a first value range, and the corresponding states of the P first models are the first state; the first switching delay information indicates that the states of Q first models among the K5 first models are the first state, where Q is a positive integer. For example, the values ​​of part or all of the first switching delays (such as less than or equal to Q first switching delays) among the K5 first switching delays are all within a second value range. , then the corresponding state of the Q first models is the first state; the second switching delay information indicates that the state of R first models among the K6 first models is the first state, R is a positive integer, for example, the value of some or all of the second switching delays (such as less than or equal to R second switching delays) in the K6 second switching delays is in the third value range, then the corresponding state of the R first models is the first state; or, the first inference delay information indicates that the state of L first models among the K2 first models is the first state, L is a positive integer, for example, the value of some or all of the first inference delays (such as less than or equal to L first inference delays) in the K2 first inference delays is in the fourth value range, then the corresponding state of the L first models is the first state.

[0149] In a possible implementation manner, the first state is valid within a first time window.

[0150] In one possible implementation, the length of the first time window satisfies at least one of the following conditions: greater than the maximum value in the first value range; greater than the maximum value in the second value range; greater than the maximum value in the third value range; or greater than the maximum value in the fourth value range.

[0151] In a possible implementation, the method further includes: after the first time window, receiving second information from the terminal device, the second information indicating that the status of the M first models / functions is updated to a second state, the second state indicating that the terminal device does not support the use of the M first models, and M is a positive integer.

[0152] Optionally, the second state is valid within a fourth time window, and the fourth time window may be the same as or different from the first time window.

[0153] In a possible implementation, the time of sending the second information is the first time, wherein the time interval between the first time and the time of sending the first information satisfies a first period, and the first period is a period for the terminal device to report the status of the M first models.

[0154] In a possible implementation, the method further includes: sending first indication information to the terminal device after the first time window, the first indication information indicating reporting the status of some or all of the one or more first models; and receiving third information from the terminal device, the third information indicating the status of some or all of the one or more first models.

[0155] In a possible implementation, the method further includes: receiving fourth information from the terminal device, the fourth information indicating that the state of one or more second models is a third state, and the third state indicates that the terminal device does not support the use of the one or more second models.

[0156] In a possible implementation, the third state is valid within a second time window, wherein the second time window is the same as or different from the first time window.

[0157] In a possible implementation, the method further includes: receiving fifth information from the terminal device, the fifth information indicating that the status of some or all of the one or more second models is updated to a fourth state, the fourth state indicating that the terminal device is capable of executing some or all of the second models, wherein the difference between the time when the terminal device sends the fifth information and the time when the terminal device sends the first information is less than the length of the first time window.

[0158] In a possible implementation manner, the fourth state is valid within a third time window, wherein the third time window is the same as or different from the first time window.

[0159] In a possible implementation, second indication information is received from a terminal device, where the second indication information indicates the first time window; or third indication information is sent to the terminal device, where the third indication information indicates the first time window.

[0160] In a possible implementation manner, the fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with some or all of the one or more second models; wherein some or all of the second delay information also indicates the third state. The details of the second delay information may refer to the content discussed in the third aspect above, and will not be repeated here.

[0161] In one possible implementation, the second time window satisfies at least one of the following conditions: greater than the maximum value in the fifth value range; greater than the maximum value in the sixth value range; greater than the maximum value in the seventh value range; or greater than the maximum value in the eighth value range.

[0162] It should be noted that the contents of the above-mentioned first aspect and any possible implementation manner can be cited in the possible implementation manners of the eighth aspect and are not listed here again.

[0163] In a ninth aspect, an embodiment of the present application provides a communication method, which can be executed by a model management device, or can be executed by a software or hardware module (such as a chip) of the model management device, etc. For ease of description, the following description is based on the execution of the method by a model management device. The method includes: receiving fourth information from a terminal device, wherein the fourth information includes second delay information, and the second delay information indicates one or more delays associated with some or all of the first models in the one or more second models; wherein some or all of the first delay information indicates a third state, and the third state indicates that the terminal device does not support the use of the one or more second models.

[0164] In a possible implementation manner, the third state is valid within the second time window.

[0165] In one possible implementation, the fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with some or all of the one or more second models; wherein some or all of the second delay information also indicates the third state.

[0166] In a possible implementation, the second delay information includes at least one of the following: second activation delay information, the second activation delay information indicates H1 second activation delays, one of the second activation delays indicates the delay of activating one of the H1 second models; second inference delay information, the second inference delay information indicates H2 second inference delays, one of the second inference delays indicates the delay of inferring one of the H2 second models; second fallback delay information, the second fallback delay information indicates H3 second fallback delays, one of the second fallback delays indicates the delay of falling back one of the H3 second models to a delay of not using a model to implement a third function, and the one second model can implement the third function; second monitoring delay information, the second monitoring delay information indicates H4 second monitoring delay, wherein one second monitoring delay indicates the delay of monitoring one of the H4 second models; third switching delay information, wherein the third switching delay information indicates H5 third switching delays, wherein one third switching delay indicates the delay of switching one of the H5 second models to the third model; fourth switching delay information, wherein the fourth switching delay information indicates H6 fourth switching delays, wherein one fourth switching delay indicates switching the fourth model to one of the H6 second models; or, second deactivation delay information, wherein the second deactivation delay information indicates H7 second deactivation delays, wherein one second deactivation delay indicates the delay of deactivating one of the H7 second models, wherein H1, H2, H3, H4, H5, H6, and H7 are all positive integers.

[0167] In a possible implementation, part or all of the second delay information indicates the third state, including at least one of the following: the second activation delay information indicates that the state of U1 second model among the H1 second models is the third state, and U1 is a positive integer; the third switching delay information indicates that the state of U2 second model among the H5 second models is the third state, and U2 is a positive integer; the fourth switching delay information indicates that the state of U3 second model among the H6 second models is the third state, and U3 is a positive integer; or, the second inference delay information indicates that the state of U4 second model among the H2 second models is the third state, and U4 is a positive integer.

[0168] In a possible implementation, some or all of the H1 second activation delays, such as U1 or less than U1 second activation delays, have values ​​that are within the fifth value range, and the corresponding state of the U1 second model is the third state; some or all of the H5 third switching delays, such as U5 or less than U5 third switching delays, have values ​​that are within the sixth value range, and the corresponding state of the U5 second model is the third state; some or all of the H6 fourth switching delays, such as U3 or less than U3 fourth switching delays, have values ​​that are within the seventh value range, and the corresponding state of the U3 second model is the third state; some or all of the H2 second inference delays, such as U4 or less than U4 inference delays, have values ​​that are within the eighth value range, and the corresponding state of the U4 second model is the third state.

[0169] It should be noted that the contents of the above-mentioned first aspect and any possible implementation manner can be cited in the possible implementation manners of the ninth aspect and are not listed here again.

[0170] In a tenth aspect, an embodiment of the present application provides a communication method, which can be executed by a model management device, or can be executed by a software or hardware module (such as a chip) of the model management device. For ease of description, the following description is based on the execution of the method by a model management device. The method includes: receiving ninth information from a terminal device, wherein the ninth information indicates one or more delays associated with one or more first models, and the first information is used to determine tenth information, and the tenth information is related information to N first models of the one or more first models, where N is a positive integer.

[0171] In a possible implementation, the ninth information includes at least one of the following: activation delay information, the activation delay information indicating one or more activation delays, one of which indicates the delay of activating one of the one or more first models; reasoning delay information, the reasoning delay information indicating one or more reasoning delays, one of which indicates the delay of reasoning one of the one or more first models; fallback delay information, the fallback delay information indicating one or more fallback delays, one of which indicates the delay of falling back one of the one or more first models to a delay of not using a model to implement a first function, and the one first model can implement the first function; monitoring delay information, the monitoring delay information indicating One or more monitoring delays, wherein one monitoring delay indicates the delay of monitoring one of the one or more first models; first switching delay information, wherein the first switching delay information indicates one or more first switching delays, wherein one of the first switching delays indicates the delay of switching one of the one or more first models to a third model; second switching delay information, wherein the second switching delay information indicates one or more second switching delays, wherein one of the second switching delays indicates switching a fourth model to one of the one or more first models; or, deactivation delay information, wherein the deactivation delay information indicates one or more deactivation delays, wherein one of the deactivation delays indicates the delay of deactivating one of the one or more first models.

[0172] In one possible implementation, the ninth information includes the activation delay information; when the activation delay corresponding to the one first model is within the value range #1, the tenth information indicates the terminal device to activate the one first model; and / or, when the activation delay corresponding to the one first model is within the value range #2, the tenth information indicates the terminal device not to activate the one first model; wherein the minimum value within the value range #2 is greater than the maximum value within the value range #1.

[0173] In a possible implementation, the tenth information also indicates a first duration corresponding to the first model, wherein the first duration is the time required from the terminal device confirming the activation of the first model to the use of the first model, wherein the smaller the activation delay corresponding to the first model, the smaller the first duration.

[0174] In a possible implementation, after a fourth time duration, first configuration information is sent to the terminal device, the first configuration information indicating the configuration required for the terminal device to use the first model, and the fourth time duration is less than or equal to the activation delay corresponding to the first model.

[0175] In one possible implementation, the ninth information includes the inference delay information; when the inference delay corresponding to the one first model is within the value range #3, the tenth information instructs the terminal device to activate the one first model; and / or, when the inference delay corresponding to the one first model is within the value range #4, the tenth information instructs the terminal device not to activate the one first model; wherein the minimum value within the value range #4 is greater than the maximum value within the value range #3.

[0176] In a possible implementation, the tenth information also indicates a second duration corresponding to the one first model, wherein the second duration is the duration between detecting a third signal from the terminal device and reporting a fourth signal, the third signal is related to the input of the one first model, and the fourth signal is related to the output of the one first model; wherein, the smaller the inference delay corresponding to the one first model, the smaller the second duration.

[0177] In a possible implementation, the ninth information includes the fallback delay information; when the fallback delay corresponding to the one first model is within the value range #5, the tenth information instructs the terminal device to perform fallback processing on the one first model; when the fallback delay corresponding to the one first model is within the value range #6, the tenth information instructs the terminal device not to perform fallback processing on the one first model; wherein the minimum value of the value range #6 is greater than the maximum value of the value range #5.

[0178] In one possible implementation, after the fifth time duration, second configuration information is sent to the terminal device, the second configuration information indicating that the terminal device does not use the model to implement the configuration required for the first function, and the fifth time duration is less than or equal to the fallback delay corresponding to the first model.

[0179] In a possible implementation, the ninth information includes the fallback delay information; the method also includes: receiving fourth indication information from the terminal device, the fourth indication information requesting fallback to the one first model; after the fallback delay corresponding to the one first model, determining that the terminal device has fallen back to the one first model.

[0180] In a possible implementation, the ninth information includes the monitoring delay information; the tenth information indicates a third time duration corresponding to the first model, the third time duration being the time duration between the terminal device detecting the fifth signal and reporting the sixth signal, wherein the smaller the monitoring delay corresponding to the first model, the smaller the third time duration, the third time duration being the time duration between the terminal device detecting the fifth signal and reporting the sixth signal, the fifth signal being used to monitor the first model, and the sixth signal being the result of monitoring the first model; or, the fifth signal being a reference signal for monitoring the first model, and the sixth signal being used to determine the result of monitoring the first model.

[0181] In one possible implementation, the ninth information includes the first switching delay information; when the first switching delay corresponding to the one first model is within the value range #7, the tenth information instructs the terminal device to switch the one first model to the third model; and / or, when the first switching delay corresponding to the one first model is within the value range #8, the tenth information instructs the terminal device not to switch the one first model; wherein the minimum value within the value range #8 is greater than the maximum value within the value range #7.

[0182] In a possible implementation, the ninth information includes the first switching delay information; the method also includes: receiving fifth indication information from the terminal device, the fifth indication information requesting to switch the one first model to the third model; after the first switching delay corresponding to the one first model, determining that the terminal device has switched the one first model to the third model.

[0183] In a possible implementation, after the sixth time duration, third configuration information is sent to the terminal device, the third configuration information indicating the configuration required for the terminal device to use the third model, and the sixth time duration is less than or equal to the first switching delay corresponding to the first model.

[0184] In a possible implementation, the ninth information includes the first switching delay information; the method also includes: receiving fifth indication information from the terminal device, the fifth indication information requesting to switch the one first model to the third model; after the first switching delay corresponding to the one first model, determining that the terminal device switches the one first model to the third model.

[0185] In one possible implementation, the ninth information includes the second switching delay information; when the second switching delay corresponding to the one first model is within the value range #9, the tenth information instructs the terminal device to switch the fourth model to the one first model; and / or, when the second switching delay corresponding to the one first model is within the value range #10, the tenth information instructs the terminal device not to switch the fourth model; wherein the minimum value within the value range #10 is greater than the maximum value within the value range #9.

[0186] In a possible implementation, after the seventh time duration, fourth configuration information is sent to the terminal device, the fourth configuration information indicating the configuration required for the terminal device to use the one first model, and the seventh time duration is less than or equal to the second switching delay corresponding to the one first model.

[0187] In a possible implementation, the ninth information includes the second switching delay information; the method also includes: receiving sixth indication information from the terminal device, the sixth indication information requesting to switch the fourth model to the one first model; after the second switching delay corresponding to the one first model, determining that the terminal device has switched the fourth model to the one first model.

[0188] In one possible implementation, after an eighth time period, fifth configuration information is sent to the terminal device, the eighth time period is less than or equal to the deactivation delay corresponding to the one first model, the fifth configuration information indicates the configuration required for the terminal device to implement a second function after deactivating the one first model, wherein the one first model can implement the second function.

[0189] In one possible implementation, the ninth information includes the deactivation delay information; when the deactivation delay corresponding to the one first model is within the value range #11, the tenth information instructs the terminal device to deactivate the one first model; and / or, when the deactivation delay corresponding to the one first model is within the value range #12, the tenth information instructs the terminal device not to deactivate the one first model; wherein the minimum value within the value range #11 is greater than the maximum value within the value range #12.

[0190] In a possible implementation, the method further includes: receiving a seventh indication message from a terminal device, the seventh indication message requesting deactivation of the one first model; and determining that the terminal device has deactivated the one first model after a deactivation delay corresponding to the one first model.

[0191] In an eleventh aspect, an embodiment of the present application provides a communication method, which can be executed by a model management device, or can be executed by a software or hardware module (such as a chip) of the model management device. For ease of description, the following description is based on the execution of the method by a model management device. The method includes: sending eleventh information to a terminal device, the eleventh information indicating model parameters of one or more first models, wherein the model parameters of the one or more first models are valid within a fifth time window.

[0192] In a possible implementation, the method further includes: after the fifth time window, sending twelfth information to the terminal device, the twelfth information indicating updated model parameters of N first models among the one or more first models, where N is a positive integer.

[0193] In a twelfth aspect, an embodiment of the present application provides a communication method, which can be executed by a model management device, or can be executed by a software or hardware module (such as a chip) of the model management device, etc. For ease of description, the following description is given by taking the model management device as executing the method. The method includes: receiving thirteenth information from a terminal device, wherein the thirteenth information indicates the status of one or more first models, the status of the one or more first models includes at least a first state or a second state, the first state indicates that the terminal device supports the use of the one or more first models, the second state indicates that the terminal device does not support the use of the one or more models, and the first state is valid within a first time window.

[0194] In a possible implementation manner, the second state is valid within a second time window. Optionally, the second time window is the same as or different from the first time window.

[0195] The specific indication methods of the first state and the second state in this aspect can refer to the description of the first state and the third state in the first aspect, respectively, and will not be repeated here. The possible implementation methods related to the first state or the second state in the first aspect can be applied to this aspect after replacing the third state with the second state.

[0196] In the thirteenth aspect, an embodiment of the present application provides a communication device, which may be a model application device provided in any one of the first to sixth aspects, or software or hardware (such as a chip) in the model application device. For example, the model application device may be one or more of a terminal device, an access network device (such as a base station), or a core network device, etc. The communication device includes corresponding means (means) or modules for executing any one of the first to sixth aspects or any possible implementation methods. For example, the communication device includes a processing module (sometimes also referred to as a processing unit), and a transceiver module (sometimes also referred to as a transceiver unit).

[0197] The transceiver module communicates with the model management device under the control of the processing module. For example, the transceiver module may send first information to the model management device under the control of the processing module, wherein the first information indicates that the state of one or more first models is a first state, the first state indicates that the terminal device supports the use of the one or more first models, and the first state is valid within a first time window.

[0198] Optionally, the communication device can also be used to implement the first to sixth aspects or any possible implementation methods mentioned above, which are not listed one by one here.

[0199] In a fourteenth aspect, an embodiment of the present application provides a communication device, which may be a model management device described in any one of the seventh to twelfth aspects above, or a software module or hardware model (such as a chip) configured in the model management device. The communication device includes corresponding means (means) or modules for executing any one of the seventh to twelfth aspects above or any possible implementation method. For example, the communication device includes a processing module (sometimes also referred to as a processing unit), and a transceiver module (sometimes also referred to as a transceiver unit).

[0200] The transceiver module communicates with the model application device under the control of the processing module. For example, the transceiver module may receive first information from the terminal device under the control of the processing module, wherein the first information indicates that the state of the one or more models is a first state, wherein the first state indicates that the terminal device supports the use of the one or more models, and wherein the first state is valid in the first time window.

[0201] Optionally, the communication device can also be used to implement the seventh to twelfth aspects above or any possible implementation manner, which are not listed one by one here.

[0202] In the fifteenth aspect, an embodiment of the present application provides a communication device, comprising a processor and an interface circuit, wherein the interface circuit is used to receive signals from other communication devices outside the communication device and transmit them to the processor or send signals from the processor to other communication devices outside the communication device, and the processor is used to implement any method described in the first to twelfth aspects and any possible implementation manner through a logic circuit or executing code instructions.

[0203] In the specific implementation process, the communication device may be a chip, and the processor may be a transistor, a gate circuit, a trigger, and various logic circuits, etc. The embodiment of the present application does not limit the specific implementation method of the processor.

[0204] In one implementation, the communication device may be a wireless communication device, that is, a computer device supporting wireless communication functions. Specifically, the wireless communication device may be a terminal such as a smart phone, or a wireless access network device such as a base station.

[0205] In another implementation, the communication device may be a part of a wireless communication device, such as an integrated circuit product such as a system chip or a communication chip. The system chip may also be referred to as a system on chip (SoC), or simply as an SoC chip. The communication chip may include a baseband processing chip and a radio frequency processing chip. The baseband processing chip is sometimes also referred to as a modem or a baseband chip. The radio frequency processing chip is sometimes also referred to as a radio frequency transceiver or a radio frequency chip. In a physical implementation, some or all of the chips in the communication chip may be integrated inside the SoC chip. For example, the baseband processing chip is integrated in the SoC chip, and the radio frequency processing chip is not integrated with the SoC chip. The interface circuit may be a radio frequency processing chip in a wireless communication device, and the processor may be a baseband processing chip in a wireless communication device. The interface circuit may be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin or a related circuit on the chip or the chip system. The processor may also be embodied as a processing circuit or a logic circuit.

[0206] In another implementation, the communication device may be a part of a wireless communication device, such as an integrated circuit product such as a system chip or a communication chip. The interface circuit may be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin or a related circuit on the chip or the chip system. The processor may also be embodied as a processing circuit or a logic circuit.

[0207] In the sixteenth aspect, an embodiment of the present application provides a communication device, comprising: a processing circuit and an interface circuit; wherein: the interface circuit is used to couple with a memory outside the communication device and provide a communication interface for the processing circuit to access the memory; the processing circuit is used to execute program instructions in the memory to implement a method as described in any one of the first to twelfth aspects and any possible implementation manner.

[0208] In the specific implementation process, the communication device can be a chip, the input circuit can be an input pin, the output circuit can be an output pin, and the processing circuit can be a transistor, a gate circuit, a trigger, and various logic circuits. The input signal received by the input circuit can be, for example, but not limited to, received and input by a receiver, and the signal output by the output circuit can be, for example, but not limited to, output to a transmitter and transmitted by the transmitter, and the input circuit and the output circuit can be the same circuit, which is used as an input circuit and an output circuit at different times. The embodiments of the present application do not limit the specific implementation methods of the processor and various circuits.

[0209] In one implementation, the communication device may be a wireless communication device, that is, a computer device that supports wireless communication functions. Specifically, the wireless communication device may be a terminal such as a smart phone, or a wireless access network device such as a base station. The system chip may also be referred to as a system on chip (SoC), or simply as an SoC chip. The communication chip may include a baseband processing chip and a radio frequency processing chip. The baseband processing chip is sometimes also referred to as a modem or a baseband chip. The radio frequency processing chip is sometimes also referred to as a radio frequency transceiver or a radio frequency chip. In a physical implementation, some or all of the chips in the communication chip may be integrated inside the SoC chip. For example, the baseband processing chip is integrated in the SoC chip, and the radio frequency processing chip is not integrated with the SoC chip. The interface circuit may be a radio frequency processing chip in a wireless communication device, and the processing circuit may be a baseband processing chip in a wireless communication device.

[0210] In another implementation, the communication device may be a part of a wireless communication device, such as an integrated circuit product such as a system chip or a communication chip. The interface circuit may be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin or a related circuit on the chip or the chip system. The processor may also be embodied as a processing circuit or a logic circuit.

[0211] In the seventeenth aspect, an embodiment of the present application provides a communication device, comprising: a processor and a memory; the memory is used to store one or more computer programs, and the one or more computer programs include computer execution instructions. When the communication device is running, the processor executes the one or more computer programs stored in the memory, so that the communication device performs any method described in the first aspect to the twelfth aspect and any possible implementation manner.

[0212] Optionally, the communication device further includes other components, such as an antenna, an input / output module, an interface, etc. These components may be hardware, software, or a combination of software and hardware.

[0213] In an eighteenth aspect, an embodiment of the present application provides a chip system, the chip system comprising: a processor and an interface. The processor is used to call and run instructions from the interface, and when the processor executes the instructions, the method described in any one of the first to twelfth aspects and any possible implementation manner is implemented.

[0214] In the nineteenth aspect, an embodiment of the present application provides a computer-readable storage medium, which is used to store computer programs or instructions. When the computer-readable storage medium is executed, it implements the method described in any of the first to twelfth aspects and any possible implementation methods.

[0215] In the twentieth aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a computer, implements any of the methods described in the first to twelfth aspects and any possible implementation manner.

[0216] Regarding the beneficial effects of any technical solution in the above-mentioned second to third aspects, and fifth to twentieth aspects and any possible implementation methods, reference can be made to the beneficial effects discussion of the corresponding technical solution in the first aspect or the fourth aspect, and the repeated parts will not be listed here. BRIEF DESCRIPTION OF THE DRAWINGS

[0217] Figure 1 A schematic diagram of the structure of a communication system applicable to an embodiment of the present application;

[0218] Figure 2 A schematic diagram of the structure of another communication system applicable to the embodiments of the present application;

[0219] Figure 3 A schematic diagram of an application framework of a communication system applicable to an embodiment of the present application;

[0220] Figure 4 A schematic diagram of the structure of a communication system provided in an embodiment of the present application;

[0221] Figure 5 is a schematic diagram of a method for managing a first model;

[0222] Figure 6 A schematic diagram of a communication method provided in an embodiment of the present application;

[0223] Figures 7 to 9 Schematic diagram of three communication methods provided in embodiments of the present application;

[0224] Fig.10 A schematic diagram of another communication method provided in an embodiment of the present application;

[0225] Fig.11A schematic diagram of another communication method provided in an embodiment of the present application;

[0226] Figures 12 to 17 A schematic diagram of a process for managing N first models provided in an embodiment of the present application;

[0227] Fig.18 A schematic diagram of another communication method provided in an embodiment of the present application;

[0228] Fig.19 A schematic diagram of another communication method provided in an embodiment of the present application;

[0229] Fig. 20 A schematic diagram of the structure of a communication device provided in an embodiment of the present application;

[0230] Fig.21 A schematic diagram of the structure of another communication device provided in an embodiment of the present application;

[0231] Fig. 22 A schematic diagram of the structure of another communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0232] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0233] In the embodiments of the present application, the number of nouns, unless otherwise specified, means "singular noun or plural noun", that is, "one or more". "At least one" means one or more, and "plural" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. For example, A / B means: A or B. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c means: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0234] It should be understood that in various embodiments of the present application, indications include direct indications (also referred to as explicit indications) and implicit indications. Wherein, direct indication of information A means including the information A; implicit indication of information A means indicating information A through the correspondence between information A and information B and direct indication of information B. Wherein, the correspondence between information A and information B may be predefined, pre-stored, pre-burned, or pre-configured. In various embodiments of the present application, information C is used to determine information D, including information D being determined based only on information C, and information D being determined based on information C and other information. In addition, information C is used to determine information D, and it may also be indirectly determined, for example, information D is determined based on information E, and information E is determined based on information C. In addition, "network element A sends information A to network element B" in various embodiments of the present application can be understood as the destination end of the information A or the intermediate network element in the transmission path between the destination end and the information A is network element B, which may include sending information to network element B directly or indirectly. "Network element B receives information A from network element A" can be understood as the source of the information A or the intermediate network element in the transmission path between the source and the source is network element A, which may include directly or indirectly receiving information from network element A. The information may be processed as necessary between the source and the destination of the information, such as format changes, but the destination can understand the valid information from the source. Similar expressions in the various embodiments of the present application can be understood similarly and will not be repeated here.

[0235] The technical solution provided in the embodiments of the present application can be applied to various communication systems (or networks), such as: the fifth generation (5th generation, 5G) or new radio (new radio, NR) system, long term evolution (long term evolution, LTE) system, LTE frequency division duplex (frequency division duplex, FDD) system, LTE time division duplex (time division duplex, TDD) system, wireless local area network (wireless local area network, WLAN) system, satellite communication system, future communication system, such as the sixth generation (6th generation, 6G) mobile communication system, or a fusion system of multiple systems, etc. The technical solution provided in the present application can also be applied to device to device (D2D) communication, vehicle to everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and Internet of things (IoT) communication system or other communication systems.

[0236] For example, see Figure 1 , is a schematic diagram of a communication system applicable to the embodiments of the present application. Figure 1 As shown, the communication system includes at least one terminal device and at least one network device. Figure 1 The example is that the number of terminal devices and network devices are both 1, and the number of terminal devices and network devices is not actually limited. Any terminal device in at least one terminal device interacts with part or all of at least one network device. The signal may include information, signaling, or data, etc. The device may also be replaced by a network element, an entity, a network entity, a communication device, a communication module, a node, a communication node, or a communication device, etc. The embodiment of the present application is described by taking the device as an example. It can be understood that the terminal device in the embodiment of the present application can be replaced by a first network element or a first communication device, and the network device can be replaced by a second network element or a second communication device, and the two perform the corresponding communication method in the embodiment of the present application.

[0237] The following is an introduction to possible implementations of terminal devices and network devices.

[0238] 1. A terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device.

[0239] The terminal device may be a device that provides voice / data, such as a handheld device or a vehicle-mounted device with a wireless connection function. At present, some examples of terminals are: mobile phones, tablet computers, laptop computers, PDAs, mobile internet devices (MID), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication functions, computing devices or other processing devices connected to wireless modems, wearable devices, terminal devices in 5G networks or future evolved public land mobile communication networks (public land mobile The embodiments of the present application do not limit this.

[0240] As an example but not limitation, in the embodiments of the present application, the terminal device may also be a wearable device. Wearable devices may also be referred to as wearable smart devices, which are a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also powerful functions achieved through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include full-featured, large-sized, and fully or partially independent of smartphones, such as smart watches or smart glasses, as well as devices that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various types of smart bracelets and smart jewelry for vital sign monitoring.

[0241] In the embodiment of the present application, the device for realizing the function of the terminal device may be a terminal device, or a device capable of supporting the terminal device to realize the function, such as a chip system, which may be installed in the terminal device or used in combination with the terminal device. In the embodiment of the present application, the chip system may be composed of a chip, or may include a chip and other discrete devices. In the embodiment of the present application, only the device for realizing the function of the terminal device is used as an example for explanation, and the scheme of the embodiment of the present application is not limited.

[0242] 2. The network device may be a device for communicating with the terminal device, and the network device may be, for example, at least one of an access network device, a wireless access network device, or a third-party device (such as a third-party server). For example, the network device may be a base station. The network device in the embodiment of the present application may be, for example, a wireless access network (RAN) device (or node) that connects the terminal device to a wireless network. A base station may broadly cover various names as follows, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, access point, transmission point (TRP), transmission point (TP), master station, auxiliary station, multi-standard wireless (motor slide retainer, MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. A base station may be a macro base station, a micro base station, a relay node, a donor node, or the like, or a combination thereof. The base station may also refer to a communication module, modem or chip used to be set in the aforementioned equipment or device. The base station may also be a mobile switching center and a device that performs the base station function in D2D, V2X, and M2M communications, a network side device in a 6G network, and a device that performs the base station function in future communication systems. The base station can support networks with the same or different access technologies. Optionally, the RAN node may also be a server, a wearable device, a vehicle or an on-board device, etc. For example, the access network device in the vehicle to everything (V2X) technology may be a road side unit (RSU). The embodiments of the present application do not limit the specific technology and specific device form adopted by the network equipment.

[0243] In some deployments, the access network device mentioned in the embodiments of the present application may be a device including a CU, or a DU, or a device including a CU and a DU, or a control plane CU node (central unit control plane (central unit-control plane, CU-CP)) and a user plane CU node (central unit user plane (central unit user plane, CU-UP)) and a DU node. For example, the access network device may include a gNB-CU-CP, a gNB-CU-UP, and a gNB-DU.

[0244] In some deployments, multiple RAN nodes collaborate to assist terminal devices in achieving wireless access, and different RAN nodes implement part of the functions of the base station. For example, the RAN node can be a CU, DU, CU-CP, CU-UP, or RU. The CU and DU can be set separately, or they can also be included in the same network element, such as a BBU. The RU can be included in a radio frequency device or radio frequency unit, such as an RRU, AAU, or RRH.

[0245] The RAN node may support one or more types of fronthaul interfaces, and different fronthaul interfaces correspond to DUs and RUs with different functions. If the fronthaul interface between the DU and the RU is a common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and the RU is another interface, relative to the CPRI, part of the downlink and / or uplink baseband functions, such as, for downlink, one or more of precoding, digital beamforming (BF), or inverse fast Fourier transform (IFFT) / cyclic prefix (CP) are moved from the DU to the RU for implementation, and for uplink, one or more of digital beamforming (BF), or fast Fourier transform (FFT) / cyclic prefix (CP) are moved from the DU to the RU for implementation. In a possible implementation, the interface may be an enhanced common public radio interface (eCPRI). In the eCPRI architecture, the division between DU and RU is different, corresponding to different types (category, Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, F.

[0246] Taking eCPRI Cat A as an example, for downlink transmission, based on layer mapping, the DU is configured to implement layer mapping and one or more functions before it (i.e., one or more functions of coding, rate matching, scrambling, modulation, and layer mapping), while other functions after layer mapping (e.g., RE mapping, digital beamforming (BF), or one or more functions of inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP)) are moved to the RU for implementation. For uplink transmission, based on de-RE mapping, the DU is configured to implement de-mapping and one or more functions before it (i.e., one or more functions of decoding, de-rate matching, de-scrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and de-RE mapping), while other functions after de-mapping (e.g., one or more functions of digital BF or fast Fourier transform (FFT) / removing CP) are moved to the RU for implementation. It can be understood that for the functional description of DU and RU corresponding to various types of eCPRI, reference can be made to the eCPRI protocol and will not be repeated here.

[0247] In one possible design, the processing unit for implementing the baseband function in the BBU is called a baseband high layer (BBH) unit, and the processing unit for implementing the baseband function in the RRU / AAU / RRH is called a baseband low layer (BBL) unit.

[0248] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, CU may also be called O-CU (open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. Any unit in the CU (or CU-CP, CU-UP), DU and RU in this application may be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0249] In the embodiments of the present application, the device for realizing the function of the network device may be a network device; or it may be a device capable of supporting the network device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. The device may be installed in the network device or used in combination with the network device. In the embodiments of the present application, only the device for realizing the function of the network device is a network device as an example for explanation, and the scheme of the embodiments of the present application is not limited.

[0250] The network device and / or terminal device can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; it can also be deployed on the water surface; it can also be deployed on aircraft, balloons and satellites in the air. The scenarios in which the network device and the terminal device are located are not limited in the embodiments of the present application. In addition, the terminal device and the network device can be hardware devices, or they can be software functions running on dedicated hardware, software functions running on general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities including dedicated or general-purpose hardware devices and software functions. The present application does not limit the specific forms of the terminal device and the network device.

[0251] In order to support AI technology in the communication network, an AI node is also introduced into the communication network. Optionally, the AI ​​node may be, for example, an AI network element or an AI module.

[0252] The AI ​​node can be deployed in one or more of the following locations in the communication system: access network equipment, terminal equipment, or core network equipment, etc., or the AI ​​node can also be deployed separately, for example, deployed in a location other than any of the above devices, such as a host or cloud server in an over the top (OTT) system. The AI ​​node can communicate with other devices in the communication system, and the other devices can be, for example, one or more of the following: network equipment, terminal equipment, or network elements of the core network, etc.

[0253] It is understood that the present application does not limit the number of AI nodes. For example, when there are multiple AI nodes, the multiple AI nodes can be divided based on functions, such as different AI nodes are responsible for different functions.

[0254] It can also be understood that AI nodes can be independent devices, or they can be integrated into the same device to implement different functions, or they can be network elements in hardware devices, or they can be software functions running on dedicated hardware, or they can be virtualized functions instantiated on a platform (for example, a cloud platform). This application does not limit the specific form of the above-mentioned AI nodes.

[0255] AI nodes can be used to implement corresponding AI functions. The AI ​​module model can implement different functions according to different parameter configurations. The AI ​​node model can be configured based on one or more of the following parameters: structural parameters (such as the number of neural network layers, neural network width, connection relationship between layers, neuron weights, neuron activation functions, or at least one of the biases in the activation function), input parameters (such as the type of input parameters and / or the dimension of input parameters), or output parameters (such as the type of output parameters and / or the dimension of output parameters). Among them, the bias in the activation function can also be called the bias of the neural network.

[0256] An AI node may have one or more models / functions, or an AI node may include one or more models. The model in the embodiment of the present application can be inferred to obtain an output, which includes one parameter or multiple parameters. The learning process, training process, or reasoning process of different models can be deployed in different devices, or can be deployed in the same device. Alternatively, the model can be understood to be used to implement specific functions or features, such as a model or algorithm related to a specific function or a specific one. Models include, for example, AI models, ML models, AI functions, or ML functions, which can be simply described as AI / ML models / functions. Optionally, the AI ​​model may also be referred to as an AI function, the ML model may also be referred to as an ML function, and the AI ​​model may include an ML model.

[0257] The model can be understood as a specific implementation of a function (such as an AI function), or can be understood as a characteristic or function of a terminal device. The model represents the mapping relationship between the input and output of the model. The model can be a neural network, a linear regression model, a decision tree model, a support vector machine (SVM), a Bayesian network, a Q learning model or other machine learning models. AI functions may include at least one of the following: data collection (collecting training data and / or reasoning data), data preprocessing, model training (or model learning), model information release (configuring model information), model verification (verifying the trained model), model reasoning (using the trained model for reasoning), or reasoning result release. Among them, reasoning can also be called prediction, use or determination.

[0258] Figure 2 A possible structural diagram of a communication system is shown in FIG. Figure 2As shown, the devices in the communication system are connected through interfaces (such as NG, Xn) or air interfaces. These devices, such as core network devices, access network devices (such as RAN devices), terminal devices, or one or more devices in operation, administration and maintenance (OAM), are equipped with one or more AI modules. Figure 1 In the example, the number of AI modules set in one device is 1, and the number of AI modules set is not actually limited. AI modules are used to implement corresponding functions. The AI ​​modules deployed in any two devices in one or more devices may be completely identical, partially identical, or completely different, and the embodiments of the present application do not specifically limit this.

[0259] The access network device may be a separate RAN node or may include multiple RAN nodes, for example, a CU and a DU. The CU and / or DU may also be provided with one or more AI modules. Optionally, the CU may also be split into a CU-CP and a CU-UP. One or more AI modules are provided in the CU-CP and / or CU-UP. Figure 2 At least one of the access network equipment, core network equipment, CU in the access network equipment, DU in the access network equipment, CU-CP in the access network equipment or CU-UP in the access network equipment involved can be regarded as Figure 1 An example of the network equipment involved.

[0260] Figure 3 Figure 1 is a possible application framework diagram of a communication system. Figure 3 As shown, the communication system includes a RAN intelligent controller (RAN intelligent controller, RIC). For example, the RIC can be Figure 2 The AI ​​module shown in is used to implement AI-related functions. The RIC includes near-real time RIC (near-real time RIC, near-RT RIC) and non-real time RIC (non-real time RIC, Non-RT RIC). Non-real-time RIC mainly processes non-real-time information, such as data that is not sensitive to delay, and the delay of this data can be in the order of seconds. Real-time RIC mainly processes near real-time information, such as data that is relatively sensitive to delay, and the delay of this data is in the order of tens of milliseconds. Optionally, near-real-time RIC and non-real-time RIC can also be set separately as a network element.

[0261] The near real-time RIC is used for model training and reasoning. For example, it is used to train a model and use the model for reasoning. The near real-time RIC can obtain information on the network side and / or the terminal device from the RAN node (such as CU, CU-CP, CU-UP, DU and / or RU) and / or the terminal device. This information can be used as training data or reasoning data. Optionally, the near real-time RIC can submit the reasoning results to the RAN node and / or the terminal device. Optionally, the reasoning results can be exchanged between the CU and the DU, and / or between the DU and the RU. For example, the near real-time RIC submits the reasoning results to the DU, and the DU sends it to the RU.

[0262] The non-real-time RIC is also used for model training and reasoning. For example, it is used to train a model and use the model for reasoning. The non-real-time RIC can obtain information on the network side and / or the terminal side from a wireless access device (such as a CU, CU-CP, CU-UP, DU and / or RU) and / or a terminal device. The information can be used as training data or reasoning data, and the reasoning results can be submitted to the RAN node and / or the terminal. Optionally, the reasoning results can be exchanged between the CU and the DU, and / or between the DU and the RU. For example, the non-real-time RIC submits the reasoning results to the DU, and the DU sends it to the RU.

[0263] The near real-time RIC and the non-real-time RIC may also be separately set as a network element. Optionally, the near real-time RIC and the non-real-time RIC may also be part of other devices, for example, the near real-time RIC is set in the access network device (for example, in the CU, DU), and the non-real-time RIC is set in the OAM, the cloud server, the core network device or other network devices.

[0264] Figure 3 At least one of the access network device, near real-time RIC, non-real-time RIC, CU in the access network device, DU in the access network device, CU-CP in the access network device, or CU-UP in the access network device involved may be regarded as Figure 1 An example of the network equipment involved.

[0265] The present application embodiment can be applied to the scenario of management model, please refer to Figure 4 , is a schematic diagram of a communication system applicable to the embodiments of the present application. Or, Figure 4 It can also be regarded as a schematic diagram of a scenario provided by an embodiment of the present application. Figure 4 As shown, the scenario includes a model management device and a model application device. The model application device may be, for example, the terminal device mentioned above (such as Figures 1 to 3Any terminal device involved), access network device, third-party device, server or core network device, etc., which is not specifically limited in the present application embodiment. Optionally, the scenario also includes a configuration file management device, such as a server, access network device or core network device.

[0266] The model management device may be, for example, Figure 1 The network device shown, or Figure 2 At least one of the access network equipment, core network equipment, CU in the access network equipment, DU in the access network equipment, CU-CP in the access network equipment or CU-UP in the access network equipment, or Figure 3 The access network device involved, near real-time RIC, non-real-time RIC, CU in the access network device, DU in the access network device, CU-CP in the access network device or at least one of the CU-UP in the access network device, the embodiment of the present application does not limit the specific implementation form of the model management device.

[0267] The model application device may be provided with one or more AI modules, and one or more AI modules may be deployed with models. The AI ​​module may be Figure 3 The RIC shown is, for example, a near real-time RIC or a non-real-time RIC, etc. For example, the near real-time RIC is set in a RAN device (eg, in a CU, DU), while the non-real-time RIC is set in an OAM, a cloud server, a core network device, or other network devices.

[0268] For example, taking the model management device as RIC, RIC can obtain a training data set (such as a training data set from a model application device) from an access network device (such as a CU, CU-CP, CU-UP, DU and / or RU), train the model involved in the model application device, and obtain the output of the trained model through RIC. The input and output of the trained model generator can be sent to the model application device through the access network device or directly to the model application device as training data in the training data set.

[0269] The model application device may be pre-configured with a configuration file of the model. The configuration file may be understood as at least one of the code, instructions, and data required for the model application device to run or execute the model. Optionally, the model application device may also obtain the configuration file of the model from a configuration file management device. In other words, the configuration file management device may be used to manage the configuration file of the model. The model management device may manage the model in the model application device.

[0270] exist Figure 4The example in the figure is that the number of model management devices, model application devices, and configuration file management devices is 1. In practice, there is no limit to the number of model management devices, model application devices, and configuration file management devices.

[0271] The following takes the model management device as an access network device as an example. Figure 5 The schematic diagram of the management model shown introduces the process of the management model mentioned in the background technology. Figure 5 The three steps S501 and S503 are illustrated and will be introduced below respectively.

[0272] S501: The access network device instructs the terminal device to report the status of the model.

[0273] S502: The terminal device sends the status of the model to the access network device.

[0274] S503: The access network device sends the policy of the management model to the terminal device.

[0275] It can be seen that before the access network device indicates the policy of the management model, it needs to inquire the terminal device about the state of the model, which makes the interaction between the access network device and the terminal device frequent.

[0276] In view of this, an embodiment of the present application provides a communication method, in which a model application device, such as a terminal device, can keep the state unchanged within a first time window after indicating to a model management device that it supports a model (such as one or more first models). In other words, the model application device, such as a terminal device, does not need to send the state of the one or more first models to the model management device within the first time window, which is conducive to reducing the information interaction between the model application device, such as a terminal device, and the model management device. In addition, before the model management device manages the one or more first models within the first time window, it is not necessary to instruct the model application device, such as a terminal device, to report the state of the one or more first models, which is also conducive to improving the efficiency of the model management device in managing the one or more first models.

[0277] The following describes the method provided by the embodiments of the present application in conjunction with the accompanying drawings. In the accompanying drawings corresponding to the various embodiments of the present application, all steps indicated by dotted lines are optional steps. Also, the model management device involved in the various embodiments of the present application, for example, can be the aforementioned Figure 1 The network device shown, or Figure 2 At least one of the access network equipment, core network equipment, CU in the access network equipment, DU in the access network equipment, CU-CP in the access network equipment or CU-UP in the access network equipment, or Figure 3At least one of the access network device, the near real-time RIC, the non-real-time RIC, the CU in the access network device, the DU in the access network device, the CU-CP in the access network device, or the CU-UP in the access network device, or Figure 4 The model management device involved. The various embodiments of the present application are exemplified by taking the model application device as the terminal device. The terminal device involved in the various embodiments of the present application, for example, can be the aforementioned Figures 1 to 4 Any terminal device involved, etc. The network device involved in each embodiment of the present application may be, for example, the Figure 1 The network device shown, or Figure 2 At least one of the access network equipment, core network equipment, CU in the access network equipment, DU in the access network equipment, CU-CP in the access network equipment or CU-UP in the access network equipment, or Figure 3 The access network device involved, the near real-time RIC, the non-real-time RIC, the CU in the access network device, the DU in the access network device, the CU-CP in the access network device or at least one of the CU-UP in the access network device, etc. The configuration file management device involved in each embodiment of the present application is, for example, Figure 4 It should be understood that if the technical solutions provided in the various embodiments of the present application are applied to other communication systems, the name and / or function of the device may be changed, and no limitation is imposed on this.

[0278] It can be understood that the various embodiments of the present application can be applied between the model management device and the model application device, wherein the model application device can be one or more of a terminal device, an access network device, or a core network device, or a third-party device, such as a server. For details, please refer to the description of the aforementioned AI module. The model management device, which may also be referred to as a model management entity, may be an access network device, a core network device, or a third-party device, such as a server, or may be the part of an access network device, a core network device, or a third-party device used to manage the model, which is not limited here.

[0279] The following embodiments are described by taking the communication between a terminal device and a model management device as an example. In the following description, the terminal device is replaced with other model application devices, such as an access network device, and the following embodiments can be applied to other model application devices.

[0280] Please refer to Figure 6 , which is a schematic diagram of a communication method provided in an embodiment of the present application. Figure 6 The steps S601 to S607 are illustrated and are introduced below respectively.

[0281] S601: The terminal device sends information 1 to the model management device. Correspondingly, the model management device receives information 1 from the terminal device. Information 1 may also be referred to as first information.

[0282] Information 1 may be carried in proprietary information or in a radio resource control (RRC) message, for example. The RRC message may be, for example, capability information of a terminal device or user equipment assistance information (UAI), which is not limited in the embodiments of the present application.

[0283] Information 1 indicates that the state of one or more first models is the first state, which can be simply described as information 1 indicating the first state. The one or more first models may be part or all of the first models that the terminal device can perceive. The first state means that the terminal device supports the use of (or can execute, or supports running, or can run) the one or more first models, or can be described as the one or more first models can be executed, or can be described as the one or more first models are in an executable state. It can be further understood that the terminal device can execute or run the one or more first models, or the terminal device has the conditions to execute or run one or more first models.

[0284] The following introduces the manner in which the terminal device determines that it supports one or more first models.

[0285] B1. The terminal device determines that a configuration file of at least one first model among the one or more first models has been loaded, and then determines that at least one first model is supported. The at least one first model belongs to the one or more first models.

[0286] Under B1, the time delay of the terminal device activating a first model in the at least one first model (which can be referred to as the activation delay of a first model) can be expressed as t0. The time delay of the terminal device activating a first model can be understood as the time length from the current moment to the time when the terminal device runs or executes a first model.

[0287] It should be understood that the values ​​of the first delays corresponding to any two first models in the at least one first model may be the same or different, and this embodiment of the present application does not specifically limit this.

[0288] B2. The terminal device does not locally load a configuration file of at least one of the one or more first models, but the terminal device can obtain a configuration file of at least one model from a configuration file management device.

[0289] Under B2, the time delay for the terminal device to activate one of the at least one first model can be expressed as t0+t′, wherein t′ can be the time required for the terminal device to obtain a configuration file of the first model from the configuration file management device.

[0290] It should be understood that the values ​​of the second delays corresponding to any two first models in the at least one first model may be the same or different, and this embodiment of the present application does not specifically limit this.

[0291] Among them, one or more first models can be used to implement at least one function (or characteristic) in C1 to C5 below, which are introduced below respectively.

[0292] C1. Predict channel state information (CSI) (or may be described as inferring CSI or determining CSI).

[0293] For example, the one or more first models include model 1 and model 2, model 1 can be used to predict CSI in case 1, and model 2 can be used to predict the AI-CSI prediction function in case 2. Case 1 and case 2 are, for example, different periods for reporting CSI by the terminal device.

[0294] Optionally, the duration of the period for reporting CSI by the terminal device in case 1 is shorter than that of reporting CSI by the terminal device in case 2. The period for reporting CSI by the terminal device in case 1 is, for example, 20 milliseconds (ms). The period for reporting CSI by the terminal device in case 2 is, for example, 40 ms. Alternatively, the moving (or motion) speed of the terminal device in case 1 is different from the range of values ​​of the moving speed of the terminal device in case 2. For example, case 1 corresponds to a medium-speed scenario, such as a range of values ​​of the moving speed of the terminal device of 60-120 km / h, and case 2 corresponds to a high-speed scenario, such as a range of values ​​of the moving speed of the terminal device of 120-300 km / h.

[0295] C2. Predicted beam (or may be described as inferred beam or determined beam). The beam includes the transmit beam and / or receive beam of the terminal device.

[0296] C3. Predicting the channel type (or can be described as inferring the channel type or determining the channel type).

[0297] Channel types may include line of sight (LOS) or non line of sight (NLOS), for example. Line of sight means that there is no obstruction between two devices (such as a terminal device and a network device, or a terminal device and a model management device) that affects signal transmission. Non-line of sight means that there is an obstruction between two devices (such as a terminal device and a network device, or a terminal device and a model management device) that affects signal transmission.

[0298] C4, compressed CSI.

[0299] For example, the one or more first models include AI model 3 and AI model 4, where AI model 3 is used to compress CSI with rank (RANK) = 1. AI model 4 is used to compress CSI with RANK = 4.

[0300] Of course, the above C1-C4 are just examples of functions that may be implemented by one or more first models. In fact, with the evolution of the standard, one or more first models may have more functions, and the embodiments of the present application do not limit this.

[0301] C5. Positioning, which can be used to determine the location of the terminal device.

[0302] The models that the terminal device can perceive may not be limited to one or more first models, and may also include one or more second models. The function implemented by any second model in the one or more second models may be the same as the function implemented by one first model in the one or more first models, or the function implemented by at least one second model in the one or more second models may be different from the function implemented by any first model in the one or more first models, and the embodiments of the present application do not limit this.

[0303] Optionally, the terminal device may send the state of the model to the model management device with the function (or characteristic) as the granularity. In this case, the one or more first models and the one or more second models may be used to implement a function. Alternatively, the terminal device may send the state of the model to the model management device with the terminal device as the granularity, in which case the one or more first models and the one or more second models may be all models that the terminal device can perceive.

[0304] After introducing the first state and the content of one or more first models, the following describes the manner in which information 1 indicates the first state. The manner in which information 1 indicates the first state may include the following first possible implementation manner and second possible implementation manner, which are described below respectively.

[0305] In a first possible implementation manner, information 1 includes first delay information, where the first delay information indicates a first state.

[0306] The first delay information indicates one or more delays associated with K first models in one or more first models. K is a positive integer. The K first models may be part or all of the first models in one or more first models. The first delay information may include at least one of the following D1-D7.

[0307] D1, first activation delay information.

[0308] The first activation delay information indicates K1 first activation delays. The K1 first activation delays correspond to K1 first models, wherein one first activation delay corresponds to one first model among the K1 first models. One first activation delay among the K1 first activation delays indicates the delay of activating one first model among the K1 first models. The first activation delay corresponding to one first model can be understood as the time required for the terminal device to execute (or run) the one first model from the current moment. K1 is a positive integer.

[0309] The first activation delay information may indicate K1 first activation delays in a quantitative manner. For example, the first activation delay information may indicate the K1 first activation delays through the value of the activation delay information. Alternatively, the first activation delay information may also indicate the index (or identifier) ​​corresponding to the K1 first activation delays to indicate the K1 activation delays. Alternatively, the first activation delay information may also indicate the index (or identifier) ​​corresponding to the value range to which the K1 first activation delays belong.

[0310] Exemplarily, the terminal device and the model management device both pre-store multiple value ranges of the first activation delay and identifiers corresponding to each value range, and the terminal device can report the identifier of the value range corresponding to each first activation delay in the K1 first activation delays to the model management device. In this way, reporting overhead can be saved and concise indication can be achieved.

[0311] D2, first inference delay information.

[0312] The first inference delay information indicates K2 first inference delays, wherein the method of indicating the K2 first inference delays can refer to the method of indicating the K2 first activation delays in the previous text, and will not be listed here. The K2 first inference delays correspond to K2 first models, wherein one first inference delay corresponds to one first model among the K2 first models. One first inference delay among the K2 first inference delays indicates the delay of inferring one first model among the K2 first models. The inference delay corresponding to a first model can be understood as the time required for the terminal device to implement the inference process of the first model, such as the time required from receiving input from the first model to obtaining output from the first model. K2 is a positive integer.

[0313] Optionally, a first inference delay corresponding to a first model may be related to the complexity of the first model and / or the hardware capabilities of the terminal device.

[0314] For example, when the hardware capabilities of the terminal devices are the same, the more complex a first model is, the longer the first inference delay of the first model may be. For another example, for the same first model, different chips or even different chip versions corresponding to the terminal devices may result in different inference delays for the first model due to different compilation methods of the first model by the terminal devices.

[0315] D3, first fallback delay information.

[0316] The first fallback delay information indicates K3 first fallback delays, wherein the method of indicating K3 first fallback delays can refer to the method of indicating K3 first activation delays in the previous text, and is not listed here. K3 first fallback delays correspond to K3 first models, wherein one first fallback delay corresponds to one first model among the K3 first models. One fallback delay among the K3 first fallback delays indicates the delay of falling back one first model among the K3 first models to the delay of not using the model to implement the first function, and one first model can implement the first function. K3 is a positive integer.

[0317] For example, one or more first models include first models of multiple ranks (RANK) for compressing CSI (for example, including AI model 1 and AI model 2). That is, the first function includes compressing CSI. Not using a model to implement the first function can be understood as not using a model to compress CSI (also referred to as a traditional method for compressing CSI).

[0318] AI model 1 may occupy some of the same hardware resources as CSI without model compression, while AI model 2 and CSI without model compression occupy completely different hardware resources. Therefore, the time required for AI model 1 to fall back to CSI without model compression (i.e., the fallback delay corresponding to AI model 1) may be longer than the first fallback delay corresponding to AI model 2.

[0319] D4, first monitoring delay information.

[0320] The first monitoring delay information indicates K4 first monitoring delays, wherein the method of indicating the K4 first monitoring delays can refer to the method of indicating the K4 first activation delays in the previous text, and will not be listed here. The K4 first monitoring delays correspond to K4 first models, wherein one first monitoring delay corresponds to one first model among the K4 first models. One monitoring delay among the K4 first monitoring delays indicates the delay of monitoring one first model among the K4 first models. The monitoring delay of a first model can be understood as the delay required for the monitoring simulation to implement the reasoning process with the said first model. Optionally, the terminal device may use a proxy model to obtain the first monitoring delays of the K4 first models. K4 is a positive integer.

[0321] Optionally, the first monitoring delay of a first model may be related to the complexity of the first model. For example, the higher the complexity of the first model, the longer the first monitoring delay of the first model.

[0322] For example, one or more first models include AI model 1 and AI model 2, AI model 1 is used to compress CSI, and AI model 2 is used to decompress CSI. The terminal device can use the proxy model to simulate the reasoning process of AI model 1 and the reasoning process of AI model 2, so as to obtain the first monitoring delay corresponding to AI model 1 and the first monitoring delay of AI model 2 respectively. The complexity of AI model 2 is higher than that of AI model 1. Therefore, the first monitoring delay corresponding to AI model 1 is less than the first monitoring delay of AI model 2.

[0323] D5. First switching delay information.

[0324] The first switching delay information indicates K5 first switching delays, wherein the method of indicating the K5 first switching delays can refer to the method of indicating the K5 activation delays in the previous text, and is not listed here. The K5 first switching delays correspond to the K5 first models, wherein one of the first switching delays corresponds to a first model in the K5 first models. A first switching delay in the K5 first switching delays indicates the delay of switching a first model in the K5 first models to a third model. The third model may belong to the K5 first models, or may belong to the K5 second models. K5 is a positive integer.

[0325] For example, one or more models include AI model 1 and AI model 2. AI model 1 and AI model 2 share some hardware resources, while the hardware resources of AI model 1 and AI model 3 are completely different. Then, the first switching delay required for AI model 1 to switch to AI model 2 may be longer than the first switching delay required for AI model 1 to switch to AI model 3.

[0326] D6, second switching delay information.

[0327] The second switching delay information indicates K6 second switching delays, wherein the manner of indicating the K6 second switching delays may refer to the manner of indicating the K6 activation delays in the previous text, and will not be listed here. The K6 second switching delays correspond to the K6 first models, wherein one second switching delay corresponds to a first model in the K6 first models. A second switching delay in the K6 second switching delays indicates that the fourth model is switched to a first model in the K6 first models. The fourth model may belong to the K6 first models, or may belong to the K6 second models. The fourth model may be the same as or different from the third model in the previous text. K6 is a positive integer.

[0328] D7: first deactivation delay information.

[0329] The first deactivation delay information indicates K7 first deactivation delays, wherein the method of indicating K7 first deactivation delays can refer to the method of indicating K7 first activation delays in the previous text, and will not be listed here. The K7 first deactivation delays correspond to K7 first models, wherein one first deactivation delay corresponds to a first model among the K7 first models. A first deactivation delay among the K7 first deactivation delays indicates the delay of deactivating a first model among the K7 first models. The first activation delay of a first model can be the time required for the terminal device to stop using a first model after reporting the deactivation of a first model. K7 is a positive integer.

[0330] For example, when a first model is activated, the network device is instructed to deactivate the first model. After a time interval T1, even if the terminal device does not receive the deactivation instruction from the network device, the terminal device can also deactivate the first model. The time interval T1 here can be regarded as the first deactivation delay corresponding to the first model.

[0331] Optionally, the one or more delays indicated by the first delay information include at least one of a first activation delay, a first inference delay, a first fallback delay, a first monitoring delay, a first switching delay, a second switching delay and a first deactivation delay.

[0332] Optionally, the first delay information may indicate the value corresponding to one or more first models under each delay of one or more delays. Alternatively, the first delay information may indicate one or more delays corresponding to each first model in some or all of the first models and each first model. The following takes one or more first models including model A, model B and model C as an example to illustrate the specific content of the first delay information.

[0333] For example, the first delay information includes first inference delay information, first activation delay information, first fallback delay information, first switching delay information, and first deactivation delay information. The first inference delay information, the first activation delay information, the first fallback delay information, the first switching delay information, and the first deactivation delay information can be respectively expressed as follows.

[0334] 1. The inference delay information can be expressed as: {t inference,A ,t inference,B ,t inference,C}. Among them, t inference,A ,t inference,B and t inference,C They represent the inference delays corresponding to model A, model B, and model C respectively.

[0335] 2. Monitoring delay information can be expressed as: {t proxy,A ,t proxy,B ,t proxy,C}. Among them, t proxy,A ,t proxy,B and t proxy,C They can represent the monitoring delays corresponding to model A, model B, and model C respectively.

[0336] 3. The activation delay information can be expressed as: {t activation,A ,t activation,B ,t activation,C}. Among them, t activation,A ,t activation,B and t activation,C They represent the activation delays corresponding to model A, model B, and model C respectively.

[0337] 4. The backoff delay can be expressed as: {t fallback,A ,t fallback,B ,t fallback,C}. Among them, t fallback,A ,t fallback,B and t fallback,C They represent the backoff delays corresponding to model A, model B, and model C respectively.

[0338] 5. The deactivation delay information can be expressed as: {t deact,A ,t deact,B ,t deact,C}. Among them, t deact,A ,t deact,B and t deact,C They represent the deactivation delays corresponding to model A, model B, and model C respectively.

[0339] 6. The first switching delay information can be expressed as: {t switch,A→B ,t switch,A→C ,t switch,B→A ,t switch,B→C ,t switch,C→A}.

[0340] t switch ,A→B,t switch,A→C ,t switch,B→A t switch,B→C ,t switch,C→A They respectively represent the switching delay from model A to model B, the switching delay from model A to model C, the switching delay from model B to model A, the switching delay from model B to model C, and the switching delay from model C to model A.

[0341] Alternatively, the first delay information includes: inference,A ,t proxy,A ,t activation,A ,t fallback,A ,tdeact,A ,t switch,A→B ,t switch,A→C}, {t inference,B ,t proxy,B ,t activation,B ,t fallback,B ,t deact,B ,t switch,B→A ,t switch,B→C} and {t inference,C ,t proxy,C ,t activation,C ,t fallback,C ,t deact,C ,t switch,C→B ,t switch,C→A}.

[0342] After introducing the content of the first delay information, the manner in which the first delay information indicates the first state is introduced below.

[0343] Exemplarily, part or all of the first delay information may indicate the first state, for example, part or all of the multiple bits included in the first delay information indicate the first state, or at least one of the information shown in D1-D7 included in the first delay information indicates the first state. E1-E4 is used as an example for introduction.

[0344] E1. The first activation delay information in the first delay information indicates the first states of P first models among K1 first models, where P is a positive integer. In other words, the first activation delay information indicates K1 first activation delays and also indicates that the states of the P first models are the first state.

[0345] The P first models may be part or all of the first models in one or more first models. For example, if the values ​​of part or all of the first activation delays (for example, less than or equal to the P first activation delays) in the K1 first activation delays are all within the first value range, then the state of the P first models is the first state. The first value range may be preconfigured or predefined in the terminal device and the model management device, or may be determined by negotiation between the terminal device and the model management device.

[0346] The following takes the example that P first models include a first model A, a first model B, and a first model C to illustrate the manner in which the activation delay indicates the state of the first model.

[0347] 1. The first model A.

[0348] For example, the first activation delay corresponding to the first model A is t A , indicating that the terminal device has loaded the configuration file of the first model A locally. This also means that the terminal device supports the use of the first model A.

[0349] Alternatively, the first activation delay corresponding to the first model A is t A +Δ A , indicating that the terminal device has not loaded the configuration file of the first model A locally, but can load the configuration file of the first model A from the configuration file management device. This also means that the terminal device supports the use of the first model A.

[0350] Alternatively, the first activation delay corresponding to the first model A is +∞, indicating that the terminal device has not loaded the configuration file of the first model A locally, and cannot load the configuration file of the first model A from the configuration file management device at the current moment. This means that the terminal device does not support the use of the first model A. Not supporting use can also be referred to as being in an unexecutable state.

[0351] 2. The first model B.

[0352] For example, the first activation delay corresponding to the first model B is t B , indicating that the terminal device has loaded the configuration file of the first model B locally. This also means that the terminal device supports the use of the first model B.

[0353] Alternatively, the first activation delay corresponding to the first model B is t B +Δ B , indicating that the terminal device has not loaded the configuration file of the first model B locally, but can load the configuration file of the first model B from the configuration file management device. This also means that the terminal device supports the use of the first model B.

[0354] Alternatively, the first activation delay corresponding to the first model B is +∞, indicating that the terminal device has not loaded the configuration file of the first model B locally, and cannot load the configuration file of the first model B from the configuration file management device at the current moment. This means that the terminal device does not support the use of the first model B.

[0355] 3. The first model C.

[0356] For example, the first activation delay corresponding to the first model C is t C , indicating that the terminal device has loaded the configuration file of the first model C locally. This also means that the terminal device supports the first model C.

[0357] Alternatively, the first activation delay corresponding to the first model C is t C +Δ C , indicating that the terminal device has not loaded the configuration file of the first model C locally, but can load the configuration file of the first model C from the configuration file management device. This also means that the terminal device supports the use of the first model C.

[0358] Alternatively, the first activation delay corresponding to the first model C is +∞, indicating that the terminal device has not loaded the configuration file of the first model C locally, and cannot load the configuration file of the first model C from the configuration file management device at the current moment. This means that the terminal device does not support the use of the first model C.

[0359] E2. The first switching delay information in the first delay information indicates the first states of Q first models among the K first models, where Q is a positive integer. In other words, the first switching delay information indicates K5 first switching delays and also indicates that the states of the Q first models are the first state.

[0360] The Q first models may be part or all of the first models in one or more first models. For example, if the values ​​of part or all of the first switching delays (such as less than or equal to the Q first switching delays) in the K5 first switching delays are all in the second value range, then the state of the Q first models is the first state. The second value range may be preconfigured or predefined in the terminal device and the model management device, or may be determined by negotiation between the terminal device and the model management device.

[0361] The following takes the Q first models including the first model A and the third model including the first model B as an example to illustrate the manner in which the first switching delay indicates the state of the first model.

[0362] For example, the first switching delay from the first model A to the first model B is t A→B , indicating that the UE has locally loaded the first model A and the first model B, which means that the terminal device supports the use of the first model A and the first model B.

[0363] Alternatively, the first switching delay from the first model A to the first model B is t A→B +Δ B , indicating that the terminal device has loaded the first model A locally, and the terminal device has not loaded the first model B, but can load the first model B from the configuration file management device, which means that the terminal device supports the use of the first model A and the first model B.

[0364] Alternatively, the first switching delay from the first model A to the first model B is +∞, indicating that the terminal device may have loaded the first model A, but has not loaded it locally, and cannot load the first model B from the configuration file management device at the current moment, which means that the terminal device supports the first model A but does not support the use of the first model B.

[0365] E3. The second switching delay information in the first delay information indicates that the states of R first models among K6 first models are the first state, where R is a positive integer. In other words, the second switching delay information indicates K6 second switching delays and also indicates that the states of R first models are the first state.

[0366] The R first models may be some or all of the first models in the K6 first models. For example, if the values ​​of some or all of the second switching delays (such as less than or equal to the R first switching delays) in the K6 first switching delays are all within the third value range, then the state of the R first models is the first state. The third value range may be preconfigured or predefined in the terminal device and the model management device, or may be determined by negotiation between the terminal device and the model management device.

[0367] The following takes the example that R first models include the first model A and the fourth model includes the first model B to illustrate the manner in which the first switching delay indicates the state of the first model.

[0368] For example, the second switching delay of switching from the first model B to the first model A is t B→A , indicating that the UE has locally loaded the first model A and the first model B, which means that the terminal device supports the use of the first model A and the first model B.

[0369] Alternatively, the second switching delay of switching from the first model B to the first model A is t B→A +Δ B , indicating that the terminal device has loaded the first model B locally, and the terminal device has not loaded the first model A, but can load the first model A from the configuration file management device, which means that the terminal device supports the use of the first model A and the first model B.

[0370] Alternatively, the second switching delay from the first model B to the first model A is +∞, indicating that the terminal device may have loaded the first model B, but has not loaded it locally and cannot load the first model A from the configuration file management device at the current moment, which means that the terminal device supports the first model B but does not support the use of the first model A.

[0371] E4. The first inference delay information indicates that the states of L first models among the K2 first models are the first state, where L is a positive integer.

[0372] For example, if the values ​​of some or all of the inference delays (such as less than or equal to L inference delays) in the K2 first inference delays are all within the fourth value range, then the states of the L first models are the first state. The fourth value range may be preconfigured or predefined in the terminal device and the model management device, or may be determined by negotiation between the terminal device and the model management device.

[0373] The following example uses L first models including model A.

[0374] For example, the first inference delay of model A is t A , indicating that the configuration file of model A has been loaded locally on the terminal device, which means that the terminal device supports the use of model A, that is, model A is in an executable state.

[0375] Alternatively, the activation delay of model A is +∞, which means that the terminal device has not loaded the configuration file of model A locally and cannot load the configuration file of model A from the configuration file management device at the current moment, which means that the terminal device does not support the use of model A, that is, model A is in an unexecutable state.

[0376] It should be understood that when the first delay information includes at least two of the first activation delay information, the first switching delay information, the second switching delay information and the first inference delay, each of the at least two pieces of information may indicate the first state of some or all of the K first models.

[0377] For example, the first delay information includes first activation delay information, first switching delay information, second switching delay information, and first inference delay information, the first activation delay information may indicate that the states of P first models are the first state, the first switching delay information indicates that the states of Q first models are the first state, and the second switching delay information indicates that the states of R first models are the first state, and the first inference delay information indicates L first models. In this case, the one or more first models may include P first models, Q first models, R first models, and L first models.

[0378] Optionally, the information 1 also includes identifications of some or all of the first models. The identifications of some or all of the first models are used to indicate K first models. The identifications of some or all of the first models may be preconfigured or predefined in the terminal device and the model management device, or may be determined by negotiation between the terminal device and the model management device. In this case, the terminal device may flexibly indicate one or more delays associated with each of the K first models.

[0379] Alternatively, the information 1 further includes configuration information related to part or all of the first model. The configuration information related to part or all of the first model is also used to indicate part or all of the first model.

[0380] Configuration information related to part or all of the first model may include an identifier (or index) of a configuration related to part or all of the first model. The configuration related to part or all of the first model may include, for example, one or more of the following: configuration of the input of part or all of the first model (such as resources), configuration of the output of part or all of the first model (such as reporting resources for the output of part or all of the first model), functions implemented by part or all of the first model and scenarios applicable to part or all of the first model, and information used by the terminal device to support related functions of part or all of the model. Among them, the configuration related to part or all of the first model may be bound to this part or all of the first model.

[0381] For example, part or all of the first model includes AI model 1 for compressing CSI with RANK=1, then the terminal device may indicate the index of the CSI with RANK=1 (i.e., regarded as an example of configuration information corresponding to the input of AI model 1) to the network device in information 1, which is equivalent to indicating AI model 1. Alternatively, the terminal device may indicate in information 1 that the terminal device supports compression of CSI with RANK=1, which is equivalent to indicating AI model 1.

[0382] In a second possible implementation manner, the information 1 includes multiple bits, and each s bits of the multiple bits indicate that the state of one first model among the one or more first models is the first state, and s is a positive integer.

[0383] Exemplarily, the terminal device indicates the first state in the form of a bitmap. The position of the bitmap indicates one or more first models, and the value of the bit in the bitmap indicates the state of each of the one or more first models. This method is applicable to the case where the terminal device and the model management device can predetermine the first model indicated by each position in the bitmap.

[0384] For example, one or more first models include first model 1, first model 2, first model 3 and first model 4, the first bit to the fourth bit position in the bit map correspond to the states of first model 1, first model 2, first model 3 and first model 4 respectively, and the bit value is 1, indicating that the terminal device can support the model, and the bit value is 0, indicating that the terminal device does not support the model. The value of information 1 is "1111", which means that the terminal device supports first model 1, first model 2, first model 3 and first model 4.

[0385] Optionally, the information 1 also includes an identifier of each of the one or more first models, or includes configuration information related to each of the one or more first models. The configuration information related to each first model may indicate one or more of the configuration required to use each of the first models, the configuration required to report the output of each of the first models, the configuration required for the input of each of the first models, the scenarios applicable to each of the first models, or the parameters bound or associated with each of the first models, etc., which is not specifically limited in the embodiments of the present application.

[0386] In the embodiment of the present application, the first state is valid within the first time window, or can be described as the terminal device maintaining the state of one or more first models in the first state within the first time window. The start time (or starting time) of the first time window is, for example, the time when the terminal device sends information 1 to the model management device, or the time when the model management device receives information 1.

[0387] The length of the first time window may be preconfigured or predefined in the terminal device and the model management device, or may be determined by negotiation between the terminal device and the model management device, or may be determined by the terminal device itself. In the case where the first time window is determined by the terminal device itself, the terminal device may send a second indication information to the model management device, the second indication information indicating the first time window, for example, the second indication information indicates at least one of the length, starting position and ending position of the first time window, when the second indication information only indicates one of the length, starting position and ending position, the other two of the length, starting position and ending position may be defined by the protocol. Alternatively, the terminal device receives a third indication information from the model management device, the third indication information indicating the first time window, for example, the second indication information indicates at least one of the length, starting position and ending position of the first time window, when the third indication information only indicates one of the length, starting position and ending position, the other two of the length, starting position and ending position may be defined by the protocol.

[0388] Exemplarily, the terminal device may report the status of some or all of the first models in one or more first models according to the first period, and the terminal device may set the length of the first time window to be less than or equal to the length of the first period. Alternatively, the terminal device may predict the average duration required for the status of one or more first models to change, and set the length of the first time window to the average duration. Of course, there may be many ways for the terminal device to determine the first time window, and the embodiments of the present application do not specifically limit this. The first period may be determined by negotiation between the terminal device and the model management device, the model management device may be configured for the terminal device, or may be predefined or preconfigured in the terminal device and the model management device, and the embodiments of the present application do not specifically limit this.

[0389] In a possible implementation, the length of the first time window may satisfy at least one of the following conditions: greater than the maximum value in the first value range, greater than the maximum value in the second value range, greater than the maximum value in the third value range, and greater than the maximum value in the fourth value range. In this way, it can be ensured that the terminal device has enough time to activate or switch the corresponding first model. The contents of the first value range, the second value range, the third value range, and the fourth value range can refer to the contents discussed above and are not listed here.

[0390] Optionally, the first time windows corresponding to the first models for implementing multiple functions may be the same, in which case the one or more first models may be used to implement multiple functions. For example, the function may be to predict CSI or positioning. Alternatively, the first time windows corresponding to the first models for implementing one function may be the same, in which case the one or more first models are used to implement the same function.

[0391] S602: The terminal device sends information 2 to the model management device. Correspondingly, the model management device receives information 2 from the terminal device. The information 2 may also be referred to as fourth information. The information 2 may indicate that the state of one or more second models is the third state.

[0392] In the case where the terminal device can still perceive one or more second models, the terminal device can send the status of the one or more second models to the model management device. In the embodiment of the present application, the status of the one or more second models is used as an example as the third state.

[0393] The third state means that the terminal device does not support the use (or operation) of the one or more second models, or can be described as the one or more second models cannot be executed (or cannot be executed), or can be described as the one or more second models are in an unexecutable state. It can be further understood that the terminal device does not support the use of the one or more second models, or the terminal device has the conditions to execute the one or more second models.

[0394] Exemplarily, the terminal device does not support the one or more second models, including that the terminal device has not loaded the configuration files of the one or more second models locally, and the terminal device cannot obtain the configuration files of the one or more second models from the configuration file management device. In this case, the delay of the terminal device activating the one or more second models (i.e., the activation delay of the one or more second models) can be expressed as +∞.

[0395] The terminal device cannot obtain the one or more second model configuration files from the configuration file management device, which may be because the configuration file management device does not store the one or more second model configuration files. Alternatively, the configuration file management device may not store the one or more second model configuration files, but the terminal device has insufficient storage space or power, so it cannot obtain the one or more second model configuration files from the configuration file management device.

[0396] Information 2 may include second delay information, and the second delay information may indicate one or more delays of the one or more second models. Exemplarily, the second delay information further indicates at least one of a second activation delay, a second reasoning delay, a second fallback delay, a second monitoring delay, a third switching delay, a fourth switching delay, and a second deactivation delay of the one or more second models.

[0397] The manner in which information 2 indicates the third state can refer to the content of information 1 indicating the first state, as illustrated below.

[0398] In a possible implementation, part or all of the second delay information indicates the third state. Exemplarily, part or all of the bits in the second delay information indicate the third state, or part or all of at least one information included in the second delay information indicates the third state.

[0399] The second delay information includes at least one of the following: at least one of second activation delay information, second inference delay information, second fallback delay information, second monitoring delay information, third switching delay information, fourth switching delay information, and second deactivation delay information.

[0400] Among them, the second activation delay information indicates H1 activation delays, and one second activation delay indicates the delay of activating one of the H1 second models. The second reasoning delay information indicates H2 reasoning delays, and one second reasoning delay indicates the delay of reasoning one of the H2 second models. The second fallback delay information indicates H3 fallback delays, and one second fallback delay indicates the delay of falling back one of the H3 second models to the delay of not using the model to implement the third function, and one second model can implement the third function. The second monitoring delay information indicates H4 second monitoring delays, and one second monitoring delay indicates the delay of monitoring one of the H4 second models. The third switching delay information indicates H5 third switching delays, and one third switching delay indicates the delay of switching one of the H5 second models to the third model. The fourth switching delay information indicates H6 fourth switching delays, and one fourth switching delay indicates switching the fourth model to one of the H6 second models. The second deactivation delay information indicates H7 second deactivation delays, wherein one second deactivation delay indicates a delay for deactivating one of the H7 second models, wherein H7 is a positive integer.

[0401] In a possible implementation manner, part or all of the second delay information indicating the third state includes at least one of the following.

[0402] (1) The second activation delay information indicates that the states of U1 second models among H1 second models are the third state, and U1 is a positive integer.

[0403] (2) The third switching delay information indicates that the states of U2 second models among H5 second models are the third state, and U2 is a positive integer.

[0404] (3) The fourth switching delay information indicates that the states of U3 second models among the H6 second models are the third state, and U3 is a positive integer.

[0405] (4) The second inference delay information indicates that the states of U4 second models among the H2 second models are the third state, and U4 is a positive integer.

[0406] In a possible implementation, if the values ​​of some or all of the second activation delays in H1 second activation delays are in the fifth value range, the corresponding state of the U1 second model is the third state. For example, if the values ​​of the H1 second activation delays that are less than or equal to the U1 second activation delays are in the fifth value range, the corresponding state of the U1 second model is the third state. If the values ​​of some or all of the third switching delays in H5 third switching delays are in the sixth value range, the state of the U5 second model is the third state. For example, if the values ​​of some or all of the third switching delays in H5 third switching delays that are less than or equal to the U5 third switching delays are in the sixth value range, the corresponding state of the U5 second model is the third state. If the values ​​of some or all of the fourth switching delays in H6 fourth switching delays are in the seventh value range, the third state of the U3 second model. For example, if the values ​​of the H6 fourth switching delays that are less than or equal to the U3 fourth switching delays are in the seventh value range, the corresponding state of the U3 second model is the third state. If the value of some or all of the H2 second inference delays is within the eighth value range, then the third state of the U4 second models. For example, if the value of the H2 second inference delays is less than or equal to the U4 second inference delays, both of which are within the eighth value range, then the corresponding state of the U4 second models is the third state.

[0407] The specific content of the third state indicated by the second delay information can also refer to the content of the first state indicated by the first delay information in the foregoing text, and the repeated parts will not be repeated here.

[0408] Optionally, the fourth information may also include configuration information related to the one or more second models or identifiers of the one or more second models. The content of the configuration information related to the one or more second models may refer to the content of the configuration information related to the one or more first models, and the content of the identifiers of the one or more second models may also refer to the content of the identifiers of the one or more first models.

[0409] Optionally, the third state is valid within the second time window, or can be described as the terminal device maintaining the state of one or more second models in the third state within the second time window. The start time (or starting time) of the second time window is, for example, the time when the terminal device sends information 2 to the model management device, or the time when the model management device receives information 2. The way in which the model management device and the terminal device determine the second time window can refer to the content of determining the first time window in the previous text, and the embodiment of the present application does not limit this.

[0410] The second time window may be the same as the first time window, that is, the first time window and the second time window are the same time window, or the second time window and the first time window are different, which is not specifically limited in the embodiment of the present application.

[0411] Optionally, information 1 and information 2 may be carried in one message. In this case, when the terminal device sends the one message to the model management device, it is equivalent to sending information 1 and information 2 to the model management device.

[0412] It should be understood that, in the case where the terminal device cannot perceive one or more second models, it is not necessary to execute the content of S602, that is, S602 is an optional step. Figure 6 Indicated by dotted line.

[0413] It should be noted that in another communication method (or may be considered as another embodiment) provided in the embodiment of the present application, the terminal device may send information 2 to the model management device. Correspondingly, the model management device receives information 2 from the terminal device. That is, in this communication method, S602 is a step that must be executed, and Figure 6 The remaining steps in are all optional steps. In the communication method, information 2 may include second delay information, and part or all of the second delay information indicates the third state. Figure 6 The repeated parts will not be repeated here.

[0414] S603: The terminal device sends information 3 to the model management device. Correspondingly, the model management device receives information 3 from the terminal device. Information 3 may also be referred to as fifth information.

[0415] The information 3 may be carried in proprietary information, or may be carried in an RRC message, for example, the RRC message may be carried in UAI, which is not limited in the embodiment of the present application.

[0416] The terminal device needs to maintain the first state of one or more first models unchanged within the first time window. Optionally, the terminal device may change the state of one or more second models within the first time window. If within the first time window, the terminal device determines that the state of one or more second models is updated to the fourth state, for example, the terminal device loads the configuration files of the one or more second models from the configuration file management device, then the terminal device may send information 3 to the model management device. Information 3 indicates that the state of one or more second models is updated to the fourth state.

[0417] Optionally, the fourth state is valid within a third time window. The starting time of the third time window is, for example, the time when the terminal device sends information 3, or the time when the model management device receives information 3. The third time window is the same as or different from the first time window, and this embodiment of the present application does not limit this. Under this implementation, the terminal device may also keep the fourth state of one or more second models unchanged, which can further reduce the interaction between the terminal device and the model management device.

[0418] If the third state is valid within the second time window, the terminal device may send information 3 to the model management device after the second time window.

[0419] S604: The model management device sends information 4 to the terminal device. Correspondingly, the terminal device receives information 4 from the model management device. Information 4 may also be referred to as sixth information.

[0420] Information 4, for example, may be carried in dedicated signaling, or may be carried in RRC signaling, which is not limited in the embodiments of the present application.

[0421] After the model management device determines the status of one or more first models, it can determine the policies of N first models among the one or more first models, and send the policies of the N first models to the terminal device. The N first models can be some or all of the first models among the one or more first models. It should be noted that when the value of N is greater than 1, the policies for managing any two first models among the N first models can be the same or different, and the embodiments of the present application do not limit this.

[0422] The following is an introduction using a strategy for managing one of N first models as an example.

[0423] F1. The strategy indicates activation of a first model.

[0424] Exemplarily, the model management device determines that the state of a first model is updated from the second state to the first state, and the model management device may determine to activate a first model. In other words, the policy may instruct the terminal device to activate a first model. Optionally, in this case, the policy may also indicate a model parameter of the first model.

[0425] F2. The policy indicates to activate a first model.

[0426] Exemplarily, when the model management device determines that the terminal device does not need to use a first model to implement a certain function (such as the first function), the model management device may instruct to deactivate the first model.

[0427] F3. The policy indicates switching a first model to a third model.

[0428] Exemplarily, the model management device determines that a first model is not effective, and determines to switch the first model to a third model, where the third model belongs to one or more second models, or to one or more first models other than N first models.

[0429] F4. The strategy indicates switching the fourth model to a first model.

[0430] Exemplarily, the model management device determines that the effect of using the fourth model is not good, or the fourth model is not suitable for current use of the terminal device, and then determines to switch the fourth model to a first model.

[0431] F5. The policy indicates that a rollback process is performed on a first model.

[0432] Exemplarily, the model management device determines that the terminal device does not use the first model effectively, or that the first model is not suitable for the current use of the terminal device, and thus can determine to roll back the first model to a delay of not using the model to implement the first function. A first model can implement the first function.

[0433] It should be noted that the above is an example introduction to the content of the policy. In fact, the content of the policy can be multiple, and the embodiments of the present application do not specifically limit this.

[0434] After receiving information 4, the terminal device may optionally send a first response message to the model management device, wherein the first response message indicates that the terminal device confirms that one or more of the N first models are processed according to the policy. The first response message may specifically indicate that the terminal device has received information 4, or indicates that the terminal device will execute according to the instructions of information 4, or indicates that the terminal device has executed according to the instructions of information 4, and the embodiments of the present application do not specifically limit this.

[0435] For example, if the policy indicates activation of N first models, the first response message may indicate confirmation of activation of the N first models; or, if the policy indicates deactivation of N first models, the first response message may indicate deactivation of the N first models.

[0436] Optionally, the information 4 may include multiple information sent sequentially by the model management device. At least two of the multiple information are used to indicate strategies for managing different first models in the one or more first models, and / or at least two of the multiple information can be used to indicate different strategies for managing at least one model in the one or more first models.

[0437] Exemplarily, within the first time window, the model management device can manage different first models in one or more first models separately. In these cases, the model management device may indicate to the terminal device multiple times the strategies used in different first models, which is equivalent to sending multiple messages, and at least two of the multiple messages are used to indicate the strategies for managing different first models in one or more first models.

[0438] Alternatively, within the first time window, the model management device may manage N first models multiple times. In this case, the model management device may indicate to the terminal device multiple times the strategies for managing the N first models, which is equivalent to sending multiple sub-information, and at least two of the multiple sub-information can be used to indicate different strategies for managing at least one model in one or more first models.

[0439] It should be noted that the execution order of S603 and S604 can be arbitrary, for example, execute S603 first and then execute S604; or, execute S603 first and then execute S604; or, execute S603 and S604 in parallel, and the embodiments of the present application do not make specific limitations on this.

[0440] In the case where S603 is executed first and then S604 is executed, the model management device can manage not only the N first models, but also one or more second models. For example, information 4 can also indicate a strategy for managing some or all of the one or more second models. The content of the strategy for managing some or all of the second models can refer to the content of the strategy for managing N first models in the previous text, which is not listed here.

[0441] In a possible implementation, if the model management device needs some auxiliary information to manage some or all of the N first models, the model management device may send an eighth indication message to the terminal device. Accordingly, the terminal device receives the eighth indication message from the model management device. The eighth indication message instructs the terminal device to report the auxiliary information of some or all of the N first models. The auxiliary information of some or all of the N first models is used to assist in the management of some or all of the N first models. After receiving the eighth indication message, the terminal device may send the auxiliary information of some or all of the N first models to the model management device. The auxiliary information of some or all of the first models in the N first models may be, for example, one or more of the location information of the terminal device, the moving (or motion) speed of the terminal device, or the Doppler frequency deviation.

[0442] If some or all of the N first models are different, then the contents of the auxiliary information of some or all of the N first models are also different. The embodiment of the present application does not specifically limit the contents of the auxiliary information of some or all of the N first models.

[0443] In this possible implementation, when executing S604, the model management device may also determine a strategy for managing the N first models based on the auxiliary information of some or all of the N first models.

[0444] For example, one or more first models include model A and model B, both of which are used to compress CSI, model A is suitable for medium-speed scenarios, and model B is suitable for high-speed scenarios. The terminal device sends to the model management device that the moving speed of the terminal device is 150km / h. The model management device determines that the terminal device is in a high-speed scenario, so it can determine to indicate that model B is activated. Correspondingly, the policy instructs the terminal device to activate model B.

[0445] In another possible implementation, the model management device may determine the auxiliary information of part or all of the N first models by itself, or obtain the auxiliary information of part or all of the N first models from other devices (such as core network devices or access network devices). The embodiments of the present application do not specifically limit this.

[0446] S602-S604 are all optional steps. Figure 6 Indicated by dotted line.

[0447] S605. The terminal device sends information 5 to the model management device. Accordingly, the model management device receives information 5 from the terminal device. Information 5 may also be referred to as second information. Information 5 indicates that the states of M first models in the one or more first models are updated to the second state. M is a positive integer, and the M first models may be part or all of the one or more first models.

[0448] The information 5 may be carried in proprietary information, or may be carried in an RRC message, for example. The RRC message may be in the capability information or UAI of the UE, which is not limited in the embodiment of the present application.

[0449] Exemplarily, after the first time window, the terminal device determines that the states of M first models in one or more first models are updated (or changed), for example, from the first state to the second state, then the terminal device may send information 5 to the model management device.

[0450] When the terminal device is in the state of sending the first model according to the first cycle, the terminal device can determine that the first moment has arrived, and then send information 5 to the model management device. The time between the first moment and the time of sending the first information is the first cycle.

[0451] In another possible implementation, if at the first moment, the states of the M first models in the one or more first models are not updated, then the terminal device may abandon (or skip) sending the states of the M first models in the one or more first models to the model management device at the first moment. In this case, step S605 may not be performed, that is, S605 is an optional step. Figure 6 Indicated by dotted line.

[0452] In a possible implementation, the information 5 may also indicate the status of part or all of the second models in the one or more second models. The part or all of the second models may be second models whose status is updated in the one or more second models, or may be one or more second models.

[0453] In another possible implementation, after the first time window, eighth information is sent to the model management device, and the eighth information indicates the status of some or all of the one or more second models. These some or all of the second models may be second models whose status is updated in one or more second models, or may be one or more second models. Optionally, the terminal device sends the eighth information at a third moment, and the time interval between the third moment and the moment of sending the first information satisfies a first period, and the first period is a period for the terminal device to report the status of some or all of the one or more second models.

[0454] S605 can be regarded as a way for the terminal device to report the status of part or all of the models, hereinafter referred to as reporting method one.

[0455] In another possible implementation, the terminal device sends seventh information to the model management device after the first time window. The seventh information indicates the status of some or all of the one or more first models. The some or all of the first models may be the first models whose status is updated in the one or more first models, or may be the one or more first models.

[0456] Optionally, the terminal device sends the seventh information at a second moment. The time interval between the second moment and the moment of sending the first information satisfies the first period.

[0457] S606: The model management device sends information 6 to the terminal device. Accordingly, the terminal device receives information 6 from the model management device. Information 6 may also be referred to as first indication information. The first indication information indicates reporting the status of some or all of the one or more first models.

[0458] Information 6, for example, may be carried in dedicated signaling, or may be carried in RRC signaling, which is not limited in the embodiments of the present application.

[0459] S607: The terminal device sends information 7 to the model management device. Correspondingly, the model management device receives information 7 from the terminal device. Information 7 may also be referred to as third information. Information 7 indicates the status of some or all of the first models in one or more first models.

[0460] The information 7 may be carried in proprietary information, or may be carried in an RRC message. The RRC message may be in the capability information or UAI of the UE, which is not limited in the embodiment of the present application.

[0461] In S607, the states of one or more first models are updated from the first state to the second state, or the states of some of the first models in the states of one or more first models are updated from the first state to the second state, or the states of one or more first models are not updated (or changed), and this embodiment of the present application does not limit this.

[0462] S606-S607 can be regarded as a way for the terminal device to report the status of part or all of the models, which is referred to as reporting method two below.

[0463] The terminal device may adopt the above-mentioned reporting method 1 or reporting method 2 to report the status of part or all of the first models in one or more first models. That is, S605 and S606-S607 are optional steps. Figure 6 Indicated by dotted line.

[0464] As discussed above, since the first state is valid within the first time window, the terminal device will not update the state of one or more first models within the first time window, so S605 to S607 are all executed after the first time window.

[0465] In an embodiment of the present application, the terminal device can maintain the status of one or more first models within the first time window. In this way, there is no need for frequent interaction between the terminal device and the model management device regarding the status of one or more first models, which is conducive to reducing data interaction between the terminal device and the model management device.

[0466] The following is an example in which a terminal device reports the status of one or more first models according to a first period, and one or more first models include model A and model B, one or more second models include model C, and the length of the first time window is less than the length of the first period, to introduce the communication method provided in an embodiment of the present application.

[0467] Please refer to Figure 7 , which is a schematic diagram of a communication method provided in an embodiment of the present application. Figure 7 The steps S701-S712 are illustrated and described below respectively.

[0468] S701, the model management device configures the first cycle for the terminal device. The content of the first cycle can be referred to in the previous text. Figure 6 The contents of the first cycle are not listed here.

[0469] For example, the model management device configures the first cycle for the terminal device at time t1. Optionally, the model management device may also indicate the time when the terminal device starts reporting model A and model B (such as time t2).

[0470] S702: The terminal device sends information to the model management device that models A and B are in an executable state, and model C is in an unexecutable state. The information that models A and B are in an executable state, and model C is in an unexecutable state can be regarded as an example of information 1. The content of information 1, the executable state, and the unexecutable state can be referred to in the previous text. Figure 6 The content is not listed here.

[0471] S703: The model management device sends information for activating model A to the terminal device. The information for activating model A can be regarded as an example of information 3.

[0472] For the content of information 3, please refer to the previous article Figure 6 The content of the discussion will not be listed here.

[0473] S704: The terminal device sends information confirming activation of model A to the model management device. The information confirming activation of model A can be regarded as an example of a first response message. Model A can be regarded as an example of N first models. The content of the first response message can be referred to in the previous text. Figure 6 The content of the discussion.

[0474] S705 : The model management device sends information of switching from model A to model B to the terminal device. The information of switching from model A to model B can be regarded as an example of information 3 .

[0475] S706: The terminal device sends information that C is in an executable state to the model management device. The information that C is in an executable state can be regarded as an example of information 2.

[0476] S707: The terminal device sends information to the model management device confirming the switch from model A to model B. The information confirming the switch from model A to model B may be regarded as an example of the first response message.

[0477] S708: The model management device sends information to the terminal device to deactivate model B. The information indicating deactivation of model B can be regarded as an example of the content of information 3.

[0478] S709: The terminal device sends to the model management device information confirming the deactivation of model B. The information confirming the deactivation of model B may be regarded as an example of the content of the first response message.

[0479] S710: The model management device sends information for activating model C to the terminal device. The information for activating model C may be regarded as an example of information 3.

[0480] S711: The terminal device sends information that the activation of model C has been confirmed to the model management device. The information that the activation of model C has been confirmed can be regarded as an example of the first response message.

[0481] From S702 to t3, the first time window is satisfied. After the first time window, the terminal device can update the status of model A and model B. In the embodiment of the present application, the terminal device does not update the status of model A and model B after the first time window. In this case, the terminal device determines that the first time (such as Figure 7 As shown in t4), the status of models A and B has not changed, so the terminal device can skip reporting at the first moment, that is, the terminal device does not report (or send) the status of models A and B to the model management device.

[0482] The following uses an example in which one or more first models include model A and model B, and one or more second models include model C to introduce the communication method provided in an embodiment of the present application.

[0483] Please refer to Figure 8 , which is a schematic diagram of a communication method provided in an embodiment of the present application. Figure 8 The steps S801-S808 are illustrated and described below respectively.

[0484] S801. The terminal device sends to the model management device that models A and B are in an executable state, and model C is in an unexecutable state. Models A and B are in an executable state, and model C is in an unexecutable state, which can be regarded as an example of information 1.

[0485] S802: The model management device sends information for activating model A to the terminal device. The information for activating model A can be regarded as an example of information 4.

[0486] S803: The terminal device sends information to the model management device to confirm activation of model A. The information to confirm activation of model A may be regarded as an example of the first response message.

[0487] S804: The model management device sends information of switching from model A to model B to the terminal device. The information of switching from model A to model B can be regarded as an example of information 4.

[0488] S805: The terminal device sends information to the model management device confirming the switch from model A to model B. The information confirming the switch from model A to model B may be regarded as an example of the first response message.

[0489] S806: The model management device sends information for deactivating model B to the terminal device. The information for deactivating model B can be regarded as an example of information 4.

[0490] S807: The terminal device sends to the model management device information confirming the deactivation of model B. The information confirming the deactivation of model B may be regarded as an example of the first response message.

[0491] After S807, the terminal device may delete the configuration file of model B due to insufficient memory or computing resources, and model B is updated from an executable state to an unexecutable state. Therefore, after the first time window (such as the time window shown from t1 to t2), the terminal device may execute S808, that is, send information that model B is in an unexecutable state to the model management device. For example, the terminal device sends information that model B is in an unexecutable state to the model management device at time t3. The information that model B is in an unexecutable state can be regarded as an example of information 7.

[0492] The following introduces the communication method provided in an embodiment of the present application by taking one or more first models including model A and model B, one or more second models including model C, the first time window is different from the second time window, and the length of the first time window is less than the length of the second time window as an example.

[0493] Please refer to Fig. 9 , which is a schematic diagram of a communication method provided in an embodiment of the present application. Fig. 9 The steps S901-S910 are illustrated and will be introduced below respectively.

[0494] S901: The model management device sends a first time window and a second time window to the terminal device. The contents of the first time window and the second time window can be found in the previous text. Figure 6 The contents of the first cycle are not listed here.

[0495] S902: The terminal device sends information to the model management device that models A and B are in an executable state, and model C is in an unexecutable state. The information that models A and B are in an executable state, and model C is in an unexecutable state can be regarded as an example of information 1. The content of information 1, the executable state, and the unexecutable state can refer to the above. Figure 6 The content is not listed here.

[0496] S903: The model management device sends information to the terminal device to activate model A. The information to activate model A can be regarded as an example of information 4. The content of information 4 can be referred to in the previous text. Figure 6 The content of the discussion will not be listed here.

[0497] S904: The terminal device sends information confirming activation of model A to the model management device. The information confirming activation of model A can be regarded as an example of a first response message. Model A can be regarded as an example of N first models. The content of the first response message can be referred to in the previous text. Figure 6 The content of the discussion.

[0498] S905 : The model management device sends information of switching from model A to model B to the terminal device. The information of switching from model A to model B can be regarded as an example of information 4 .

[0499] S906: The terminal device sends information to the model management device confirming the switch from model A to model B. The information confirming the switch from model A to model B may be regarded as an example of the first response message.

[0500] S907: The model management device sends information to the terminal device to deactivate model B. The information indicating deactivation of model B can be regarded as an example of the content of information 4.

[0501] S908: The terminal device sends to the model management device information confirming the deactivation of model B. The information confirming the deactivation of model B may be regarded as an example of the content of the first response message.

[0502] S909: The model management device sends information that model B is in an unexecutable state to the terminal device. The information that model B is in an unexecutable state can be regarded as an example of information 7.

[0503] After the terminal device determines t2, it reaches the first time window. After the first time window, the terminal device releases (or deletes) the configuration file of model B due to insufficient memory or computing power. Then, the terminal device can send information 7 to the model management device after the first time window, and the information 7 indicates that model B is in an unexecutable state.

[0504] S910: The terminal device sends information to the model management device to confirm that the model C is in an executable state. The information that the model C is in an executable state can be regarded as an example of information 3.

[0505] After the terminal device determines t4, it reaches the second time window. After the second time window, the terminal device has installed the configuration file of model C. Then, the terminal device can send information 3 to the model management device after the second time window, and the information 3 indicates that model C is in an executable state.

[0506] In order to reduce the amount of data transmission between the terminal device and the model management device, an embodiment of the present application provides a communication method, in which the terminal device can report first information to the model management device, and the first information includes first delay information. The first delay information can indicate one or more delays related to K first models in one or more first models, K is a positive integer, and on the other hand, part or all of the first delay information can be used to indicate that the state of one or more first models is the first state. That is, the first delay information indicates a variety of contents, which is conducive to reducing the number of bits occupied by the first information, thereby reducing the amount of data transmission between the terminal device and the management device.

[0507] Please refer to Fig.10 , which is a schematic diagram of a communication method provided in an embodiment of the present application. Fig.10 The two steps S1001 and S1002 are illustrated and are introduced below respectively.

[0508] S1001: The terminal device sends first information to the model management device. Correspondingly, the model management device receives the first information from the terminal device.

[0509] The first information includes first delay information. The first delay information indicates one or more delays associated with some or all of the one or more first models. The content of some or all of the one or more first models can refer to the content discussed in S601 above, and the content of the first delay information can refer to the content of the first delay information discussed in S601 above.

[0510] In the embodiment of the present application, the first delay information further indicates that the state of one or more first models is the first state. The specific content of the first delay information indicating the first state can refer to the content discussed in the above S601.

[0511] Optionally, in the embodiment of the present application, the first state may also be valid within the first time window. The content of the first time window may refer to the above Figure 6 The content of the discussion will not be listed here.

[0512] S1002: The model management device sends second information to the terminal device. Correspondingly, the terminal device receives the second information from the model management device. The second information indicates a strategy for managing N first models.

[0513] After receiving the first information, the model management device can determine whether the terminal device supports the use of one or more first models, thereby determining the policies of the one or more first models. The content of determining the policy can refer to the content of S604 above, which is not listed here.

[0514] S1002 is an optional step. Fig.10 Indicated by dotted line.

[0515] In an embodiment of the present application, the first delay information can indicate the first state by indicating one or more delays, which can further reduce the interaction between the terminal device and the model management device.

[0516] In order to enable the model management device to manage the model of the terminal device more specifically, an embodiment of the present application provides a communication method, in which the terminal device may send first information to the model management device, and the first information indicates a delay associated with one or more first models, so that the model management device can determine the relevant information related to N first models based on the first information, which is equivalent to the model management device being able to obtain more information based on less information, thereby being able to determine the management or control of the N first models in a targeted manner.

[0517] Please refer to Fig.11 , which is a communication method provided in an embodiment of the present application. Fig.11 The two steps S1101 and S1102 are illustrated and are introduced below respectively.

[0518] S1101: The terminal device sends ninth information to the model management device. Correspondingly, the model management device receives the ninth information from the terminal device.

[0519] The ninth information indicates one or more delays associated with the one or more first models. Exemplarily, the ninth information may include at least one of the following G1-G7.

[0520] G1. Activation delay information.

[0521] The activation delay information indicates one or more activation delays. One or more activation delays correspond to one or more first models, and one activation delay corresponds to one first model in the one or more first models. One of the one or more activation delays indicates the delay of activating one of the one or more first models. The activation delay corresponding to one first model can be understood as the time required for the terminal device to execute (or run) the one first model from the current moment.

[0522] The manner in which the activation delay information indicates the one or more activation delays may refer to the content of the activation delay information indicating K activation delays discussed in S601 above, which will not be listed here.

[0523] G2, inference delay information.

[0524] The inference delay information indicates one or more inference delays, wherein the method of indicating one or more inference delays can refer to the method of indicating one or more activation delays in the previous text, and will not be listed here. One or more inference delays correspond to one or more first models, wherein one inference delay corresponds to one first model in the one or more first models. An inference delay among one or more inference delays indicates the delay of inferring one first model in one or more first models. The inference delay corresponding to a first model can be understood as the time required for the first model of the terminal device to implement the inference process, such as the time required from receiving input from the first model to obtaining output from the first model.

[0525] G3, fallback delay information.

[0526] The fallback delay information indicates one or more fallback delays, wherein the manner of indicating one or more fallback delays may refer to the manner of indicating one or more activation delays in the previous text, and will not be listed here. One or more fallback delays correspond to one or more first models, wherein one fallback delay corresponds to one first model in the one or more first models. One fallback delay in one or more fallback delays indicates a delay for falling back one first model in one or more first models to a delay for implementing a first function without using a model, and a first model can implement a first function.

[0527] G4. Monitor delay information.

[0528] The monitoring delay information indicates one or more monitoring delays, wherein the manner of indicating one or more monitoring delays may refer to the manner of indicating one or more activation delays in the previous text, and will not be listed here. One or more monitoring delays correspond to one or more first models, wherein one monitoring delay corresponds to one first model in the one or more first models. One of the one or more monitoring delays indicates the delay of monitoring one of the one or more first models. The monitoring delay of a first model may be understood as the delay required for the monitoring simulation to implement the reasoning process with the said first model. Optionally, the terminal device may use a proxy model to obtain the monitoring delays of one or more first models.

[0529] G5. First switching delay information.

[0530] The first switching delay information indicates one or more first switching delays, wherein the manner of indicating one or more first switching delays may refer to the manner of indicating one or more activation delays in the previous text, and will not be listed here. One or more first switching delays correspond to one or more first models, wherein one first switching delay corresponds to a first model in one or more first models. A first switching delay in one or more first switching delays indicates the delay of switching a first model in one or more first models to a third model. The third model may belong to one or more first models, or may belong to one or more second models.

[0531] The second switching delay information indicates one or more second switching delays, wherein the manner of indicating one or more second switching delays may refer to the manner of indicating one or more activation delays in the previous text, and will not be listed here. One or more second switching delays correspond to one or more first models, wherein one second switching delay corresponds to one first model in one or more first models. A second switching delay in one or more second switching delays indicates that the fourth model is switched to a first model in one or more first models. The fourth model may belong to one or more first models, or may belong to one or more second models. The fourth model may be the same as or different from the third model in the previous text.

[0532] G6, second switching delay information.

[0533] The second switching delay information indicates one or more second switching delays, wherein the manner of indicating one or more second switching delays may refer to the manner of indicating one or more activation delays in the previous text, and will not be listed here. One or more second switching delays correspond to one or more first models, wherein one second switching delay corresponds to one first model in one or more first models. A second switching delay in one or more second switching delays indicates that the fourth model is switched to a first model in one or more first models. The fourth model may belong to one or more first models, or may belong to one or more second models. The fourth model may be the same as or different from the third model in the previous text.

[0534] G7. Deactivate delay information.

[0535] The deactivation delay information indicates one or more deactivation delays, wherein the manner of indicating one or more deactivation delays may refer to the manner of indicating one or more activation delays in the previous text and will not be listed here. One or more deactivation delays correspond to one or more first models, wherein one deactivation delay corresponds to one first model in the one or more first models. A deactivation delay among the one or more deactivation delays indicates the delay of deactivating one first model in the one or more first models. The activation delay of a first model may be the time required for a terminal device to stop using a first model after reporting the deactivation of a first model.

[0536] Optionally, the one or more delays indicated by the ninth information include at least one of activation delay, inference delay, fallback delay, monitoring delay, first switching delay, second switching delay and deactivation delay of one or more first models.

[0537] S1102. The model management device sends the tenth information to the terminal device. Accordingly, the terminal device receives the tenth information from the model management device. The tenth information indicates the management of information related to N1 first models. The tenth information is related to the ninth information, or the tenth information is determined based on the ninth information. The tenth information may, for example, be a policy indicating the management of N1 first models and / or a configuration related to N1 first models. The configuration related to N1 first models includes one or more of the configuration required to use N1 first models, the configuration required to report the output of N1 first models, the configuration required for the input of N1 first models, the scenarios applicable to N1 first models, or the parameters bound or associated with N1 first models, etc., which are not limited in this embodiment of the present application. Among them, the content of the policy can refer to the above. Figure 6 Discuss the content.

[0538] Among them, if the content of the ninth information is different, the content of the tenth information will also be different, which will be explained below.

[0539] M1. The ninth information includes activation delay information.

[0540] Under M1, if the activation delay corresponding to a first model is within the value range #1, it means that the activation delay of a first model is relatively small, so the model management device can determine to activate a first model. Accordingly, the tenth information can indicate the activation of a first model. If the activation delay corresponding to a first model is within the value range #2, it means that the activation delay of a first model is relatively large, and even indicates that the terminal device may not be able to activate a first model, so the model management device can determine not to activate a first model, and accordingly, the tenth information indicates that the terminal device does not activate a first model.

[0541] The minimum value in the value range #2 is greater than the maximum value in the value range #1. The value range #1 and the value range #2 may be pre-configured or pre-defined in the model management device, or may be determined by negotiation between the model management device and the terminal device, which is not limited in the embodiments of the present application.

[0542] Optionally, the activation delay corresponding to a first model is within the value range #1, and the smaller the activation delay corresponding to a first model, the smaller the value of the first duration corresponding to the first model indicated by the tenth information. The first duration is the duration required from the terminal device confirming the activation of the first model to the use of the first model. For example, the second signal is the output of a first model, or the output of a first model is obtained by processing the second signal.

[0543] In the case where the tenth information indicates activation of a first model, the terminal device may activate a first model. Optionally, after a fourth duration, the terminal device receives first configuration information from the model management device, and the fourth duration is less than or equal to an activation delay corresponding to a first model. The first configuration information indicates the configuration required for the terminal device to infer a first model, for example, the first configuration information indicates an input resource of a first model.

[0544] It should be understood that, in another possible situation, the terminal device and the model management device have pre-negotiated the configuration required for inferring a first model. In this case, the model management device does not need to send the first configuration information to the terminal device.

[0545] For example, see Fig.12 , is a schematic diagram of managing N1 first models provided in an embodiment of the present application. Fig.12 In the example, N1 first models are model A and model B, both model A and model B are used to perform CSI compression with RANK=4, and the value range #1 is less than or equal to 5ms.

[0546] like Fig.12As shown, the terminal device indicates to the model management device that the activation delay of model A is 2ms, and the activation delay of model B is 3ms. The model management device determines that the activation delays of models A and B are both within the value range #1, and therefore indicates the activation of models A and B. The terminal device may indicate to the model management device to confirm the activation of models A and B, respectively. After activating models A and B, the terminal device may send the CSI compressed using model A to the model management device within the first duration corresponding to model A. In addition, the terminal device may send the CSI compressed using model B to the model management device within the first duration corresponding to model B. Since the activation delay of model B is greater than the activation delay of model A, the first duration corresponding to model B (for example, 4ms) may be greater than the first duration corresponding to model A (for example, 2ms).

[0547] M2. The ninth information includes inference delay information.

[0548] If the inference delay corresponding to a first model is within the value range #3, it means that the inference delay of a first model is relatively small, so the model management device can determine to activate a first model. Correspondingly, the tenth information instructs the terminal device to activate a first model. If the activation delay corresponding to a first model is within the value range #4, it means that the inference delay of a first model is relatively large, so the model management device can determine to activate a first model. Correspondingly, the tenth information instructs the terminal device not to activate a first model. Among them, the minimum value in the value range #4 is greater than the maximum value in the value range #3.

[0549] Among them, value range #3 and value range #4 can be pre-configured or pre-defined in the model management device, or can be determined by negotiation between the model management device and the terminal device, and the embodiments of the present application do not limit this.

[0550] In the case where the tenth information indicates the activation of a first model, the terminal device can activate a first model. Optionally, the smaller the inference delay corresponding to a first model, the smaller the value of the second time length corresponding to the first model indicated by the tenth information. The second time length is the time length between the terminal device detecting the third signal and reporting the fourth signal. The third signal is related to the input of a first model, and the fourth signal is related to the output of a first model. For example, the third signal is the input of a first model, or the input of a first model is obtained by processing the third signal. For example, the fourth signal is the output of a first model, or the output of a first model is obtained by processing the fourth signal.

[0551] For example, see Fig.13 , is a schematic diagram of a process for managing N1 first models provided in an embodiment of the present application. Fig.13In the figure, the N1 first models are taken as model A, model A is used to predict CSI, and the inference delay of model A is 0.2 ms.

[0552] like Fig.13 As shown, the terminal device receives channel state information reference signal 1 (CSI-RS1), CSI-RS2, CSI-RS3, CSI-RS4 and CSI-RS5 from the model management device in sequence, and accordingly, the terminal device uses model A to determine CSI1, CSI2, CSI3, CSI4 and CSI5 corresponding to CSI-RS1, CSI-RS2, CSI-RS3, CSI-RS4 and CSI-RS5 respectively. The second duration can be the duration between the terminal device detecting CSI-RS5 and the terminal device starting to send CSI1 to the model management device.

[0553] For example, see Fig.14 , is a schematic diagram of a process for managing N1 first models provided in an embodiment of the present application. Fig.14 In the figure, the N1 first models are taken as model A, model A is used to manage beams, and the inference delay of model A is 0.2ms.

[0554] like Fig.14 As shown, the terminal device receives synchronization signal 1 (synchronization signal 1, SSB1) (or channel state information reference signal CSI-RS1), SSB2 (or CSI-RS2), SSB3 (or CSI-RS3), SSB4 (or CSI-RS4) and SSB5 (or CSI-RS5) from the model management device in sequence. Accordingly, the terminal device uses model A to determine the beams corresponding to SSB1 (or CSI-RS1), SSB2 (or CSI-RS2), SSB3 (or CSI-RS3), SSB4 (or CSI-RS4) and SSB5 (or CSI-RS5), respectively. The second duration can be the duration between the terminal device detecting SSB5 (or CSI-RS5) and the terminal device starting to send the determined beam to the model management device. Among them, SSB is, for example, a synchronization signal / physical broadcast channel block SS / PBCH block.

[0555] M3. The ninth information includes fallback delay information.

[0556] If the fallback delay corresponding to a first model is within the value range #5, it means that the fallback delay of a first model is relatively small. Therefore, the model management device can determine to fallback a first model, that is, to fallback a first model to a situation where the model is not applicable to implement the first function. A first model can be used to implement the first function. In this case, the tenth information indicates that the terminal device does not fallback to a first model. If the fallback delay corresponding to a first model is within the value range #6, it means that the fallback delay of a first model is relatively large. Therefore, the model management device can determine not to fallback to a first model. Accordingly, the tenth information indicates that the terminal device falls back to a first model. Among them, the minimum value of the value range #6 is greater than the maximum value of the value range #5.

[0557] Among them, value range #5 and value range #6 can be pre-configured or pre-defined in the model management device, or can be determined by negotiation between the model management device and the terminal device, and the embodiments of the present application do not limit this.

[0558] In the case where the policy indicates that the first model is to be rolled back, the terminal device may roll back the first model to implement the first function without using the model. Optionally, after the fifth duration, the terminal device receives second configuration information from the model management device, and the second configuration information indicates that the terminal device does not use the configuration required for implementing the first function. The fifth duration is less than or equal to the rollback delay corresponding to the first model.

[0559] In another possible implementation, if the terminal device determines that it needs to roll back a first model due to its own reasons (for example, insufficient computing resources or memory resources of the terminal device), the terminal device may send a fourth indication message to the model management device. The fourth indication message requests to roll back a first model. If the terminal device determines that it has not received a rollback indication message from the model management device within a rollback delay corresponding to a first model, the terminal device may roll back a first model on its own after a rollback delay corresponding to a first model. Accordingly, the model management device may determine that the terminal device rolls back a first model after a rollback delay corresponding to a first model.

[0560] M4. The ninth information includes monitoring delay information.

[0561] If the monitoring delay corresponding to a first model is smaller, indicating that the time required to monitor a first model is smaller, then the tenth information indicates that the value of the third time duration corresponding to a first model is smaller. The third time duration is the time duration between the terminal device detecting the fifth signal and reporting the sixth signal.

[0562] The fifth signal is used to monitor a first model. For example, the fifth signal is auxiliary information for monitoring model A, and the sixth signal is the result of monitoring a first model (or can be called a monitoring result of a first model). This situation is applicable to the scenario where the terminal device performs monitoring on a first model. Alternatively, the fifth signal is a reference signal for monitoring a first model, and the sixth signal is used to determine the result of monitoring a first model. This situation is applicable to the scenario where the model management device performs monitoring on a first model.

[0563] For example, see Fig.15 , is a schematic diagram of a process for managing N1 first models provided in an embodiment of the present application. Fig.15 In the example, N1 first models include model A, and the monitoring delay corresponding to model A is 2ms.

[0564] like Fig.15 As shown, the terminal device reports the monitoring delay corresponding to model A to the model management device as 2ms, and the model management device sends a fifth signal to the terminal device (at Fig.15 In the example of monitoring the auxiliary information of model A), the terminal device feeds back the sixth signal to the model management device after 2 ms (in Fig.15 The monitoring results corresponding to model A are used as an example).

[0565] M5. The ninth information includes the first switching delay information.

[0566] If the first switching delay corresponding to a first model is within the value range #7, the tenth information indicates that the terminal device switches the first model to the third model. If the first switching delay corresponding to a first model is within the value range #8, the tenth information indicates that the terminal device does not switch the first model. The minimum value in the value range #8 is greater than the maximum value in the value range #7.

[0567] Among them, the value range #8 and the value range #7 can be pre-configured or pre-defined in the model management device, or can be determined by negotiation between the model management device and the terminal device, and the embodiments of the present application do not limit this.

[0568] In the case where the tenth information indicates that a first model is switched to a third model, the terminal device may revert the first model to not using a model to implement the first function. Optionally, after the sixth duration, third configuration information is received from the model management device. The third configuration information indicates the configuration required for the terminal device to use the third model. The sixth duration is less than or equal to a first switching delay corresponding to a first model.

[0569] In another possible implementation, the terminal device determines that it is necessary to switch a first model to a third model due to its own reasons (for example, insufficient computing resources or memory resources of the terminal device), and sends a fifth indication message to the model management device. The fifth indication message requests to switch a first model to a third model. If the terminal device determines that it has not received the first switching indication message from the model management device within the first switching delay corresponding to a first model, then the terminal device may switch a first model to a third model after the first switching delay corresponding to a first model.

[0570] For example, see Fig.16 , is a schematic diagram of a process for managing N1 first models provided in an embodiment of the present application. Fig.16 In the example, N1 first models include model A, the third model is model A, and the first switching delay corresponding to model A is 2ms.

[0571] like Fig.16 As shown, the terminal device reports to the model management device that the first switching delay corresponding to model A is 2ms, the model management device instructs the terminal device to switch from model A to model B, and the terminal device sends a confirmation to the model management device to switch from model A to model B. After 2ms, the terminal device feeds back the output of model B to the model management device.

[0572] M6. The ninth information includes the second switching delay information.

[0573] If the second switching delay corresponding to a first model is within the value range #9, the tenth information instructs the terminal device to switch the fourth model to a first model. If the second switching delay corresponding to a first model is within the value range #10, the tenth information instructs the terminal device not to switch the fourth model. The minimum value in the value range #10 is greater than the maximum value in the value range #9.

[0574] Among them, the value range #9 and the value range #10 can be pre-configured or pre-defined in the model management device, or can be determined by negotiation between the model management device and the terminal device, and the embodiments of the present application do not limit this.

[0575] In the case where the policy indicates that the terminal device switches the fourth model to a first model, the terminal device may switch the fourth model to a first model. After the seventh duration, fourth configuration information is received from the model management device. The fourth configuration information indicates the configuration required for the terminal device to use a first model. The seventh duration is less than or equal to a second switching delay corresponding to a first model. Accordingly, the model management device may determine that the terminal device has switched a first model to a third model after a first switching delay corresponding to a first model.

[0576] In another possible implementation, the terminal device determines that the fourth model needs to be switched to a first model due to its own reasons (for example, insufficient computing resources or memory resources of the terminal device), and sends a sixth indication message to the model management device. The sixth indication message requests that the fourth model be switched to a first model. If the terminal device determines that the second switching indication message from the model management device is not received within the second switching delay corresponding to a first model, the fourth model may be switched to a first model after the second switching delay corresponding to a first model. Accordingly, the model management device may determine that the terminal device has switched the fourth model to a first model after the second switching delay corresponding to a first model.

[0577] M7. The ninth information includes deactivation delay information.

[0578] If the deactivation delay corresponding to a first model is within the value range #11, it means that the deactivation delay corresponding to a first model is relatively small, so the model management device can determine to deactivate a first model. Correspondingly, the tenth information instructs the terminal device to deactivate a first model. If the deactivation delay corresponding to a first model is within the value range #12, it means that the deactivation delay corresponding to a first model is relatively large, so the model management device can determine to deactivate a first model. Correspondingly, the tenth information instructs the terminal device not to deactivate a first model. Among them, the minimum value in the value range #11 is greater than the maximum value in the value range #11.

[0579] Among them, the value range #11 and the value range #12 can be pre-configured or pre-defined in the model management device, or can be determined by negotiation between the model management device and the terminal device, and the embodiments of the present application do not limit this.

[0580] In one possible implementation, after an eighth time period, fifth configuration information is received from the model management device, the eighth time period is less than or equal to a deactivation delay corresponding to a first model, and the fifth configuration information indicates the configuration required for the terminal device to implement a second function after deactivating a first model, wherein a first model can implement the second function.

[0581] In another possible implementation, if the terminal device determines that a first model needs to be deactivated due to its own reasons (for example, insufficient computing resources or memory resources of the terminal device), the terminal device may send a seventh indication message to the model management device. The seventh indication message requests to deactivate a first model. If the terminal device determines that it has not received a deactivation indication message from the model management device within a deactivation delay corresponding to a first model, the first model may be deactivated after a deactivation delay corresponding to a first model. Accordingly, the model management device may determine that the terminal device has deactivated a first model after a deactivation delay corresponding to a first model.

[0582] For example, see Fig.17 , is a schematic diagram of a process for managing N1 first models provided in an embodiment of the present application. Fig.17 In the example, N1 first models include model A, the third model is model A, and the deactivation delay corresponding to model A is 2ms.

[0583] like Fig.17 As shown, the terminal device reports to the model management device that the deactivation delay corresponding to model A is 2ms, the model management device instructs the terminal device to deactivate model A, and the terminal device sends a confirmation to the model management device to deactivate model A. After 2ms, the terminal device feeds back the result corresponding to the second function to the model management device.

[0584] It should be understood that when the ninth information includes at least two of the above information M1-M7, the tenth information may also include the content of the tenth information corresponding to at least two of the information M1-M7, which is not listed here.

[0585] In order to reduce the data interaction between the model management device and the terminal device, an embodiment of the present application provides a communication method, in which the model management device may send ninth information to the terminal device, and the ninth information indicates the model parameters of one or more first models. The model management device may maintain the model parameters of the one or more first models unchanged within a fourth time window. In this way, the situation where the model management device frequently updates the model parameters of one or more first models is reduced, thereby reducing the data interaction between the model management device and the terminal device.

[0586] Please refer to Fig.18 , which is a communication method provided in an embodiment of the present application. Fig.18 The three steps S1801-S1803 are illustrated and are introduced below.

[0587] S1801: The terminal device sends status information to the model management device. Correspondingly, the model management device receives the status information from the terminal device. The eleventh information indicates that the status of one or more first models is the first status or the second status. The first status, the second status, and the content of one or more first models can all refer to the above Figure 6 content.

[0588] S1801 is an optional step. Fig.18 Indicated by dotted line.

[0589] S1802: The model management device sends eleventh information to the terminal device. Correspondingly, the terminal device receives the eleventh information from the model management device. The eleventh information indicates one or more model parameters of the first model.

[0590] In the embodiment of the present application, the model parameters of one or more first models are valid in the fifth time window. In other words, the model management device can maintain the model parameters of one or more first models in the fifth time window. Optionally, the start time of the fifth time window is, for example, the time when the model management device starts to send the eleventh information. The way in which the model management device and the terminal device determine the fifth time window can refer to the above Figure 6 The content of the first time window is determined in , and will not be listed here.

[0591] S1803: The model management device sends the twelfth information to the terminal device. Accordingly, the terminal device receives the twelfth information from the model management device. The second information indicates the updated model parameters of N2 first models, and the N2 first models may be part or all of the one or more first models.

[0592] S1801 and S1803 are optional steps. Fig.18 Indicated by dotted line.

[0593] In an embodiment of the present application, the model management device can maintain the model parameters of one or more first models unchanged, avoiding frequent changes in the model parameters of one or more first models, which is conducive to reducing the interaction between the terminal device and the model management device.

[0594] In order to reduce information interaction, an embodiment of the present application provides a communication method, in which, after the terminal device indicates to the model management device that it supports the use of one or more first models (including the first state and the second state), the first state can be kept unchanged within the first time window. In other words, the terminal device does not need to send the first state of the one or more first models to the model management device within the first time window, which is conducive to reducing the information interaction between the terminal device and the model management device. In addition, before the model management device manages the one or more first models in the first time window, it is not necessary to instruct the terminal device to report the first state of the one or more first models, which is also conducive to improving the efficiency of the model management device in managing the one or more first models.

[0595] S1901: The terminal device sends thirteenth information to the model management device. Correspondingly, the model management device receives the thirteenth information from the terminal device.

[0596] The thirteenth information indicates the state of one or more first models. The state includes at least the first state or the second state. The first state means that the terminal device supports the use of one or more first models. The content of supporting the use of one or more first models can refer to the above text. Figure 6 The second state means that the terminal device does not support the use of one or more first models, wherein the content of not supporting the use of one or more first models can refer to the above Figure 6 The content of the discussion.

[0597] In the embodiment of the present application, the first state may be valid within the first time window. The content of the first time window and the content valid in the first time window may refer to the above text. Figure 6 Optionally, the second state may be valid in a second time window. The second time window is the same as or different from the first time window. The starting time of the first time window and the second time window is, for example, the time when the terminal device sends the thirteenth information.

[0598] S1902: The model management device sends the sixteenth information to the terminal device. Correspondingly, the terminal device receives the sixteenth information from the model management device. The sixteenth information indicates a strategy for managing N4 first models. The N4 first models are part or all of the first models in one or more first models. The strategy can refer to the above Figure 6 The content of the strategies discussed will not be listed here.

[0599] After receiving the thirteenth information, the model management device may not determine the strategy for managing the N4 first models. In this case, there is no need to execute the content of S1902, that is, S1902 is an optional step.

[0600] In one possible implementation, the thirteenth information includes first delay information, and the first delay information indicates one or more delays associated with some or all of the one or more first models; wherein some or all of the first delay information indicates the state of one or more first models. For example, some or all of the multiple bits included in the first delay information indicate the state of one or more first models, or some or all of the multiple information included in the first delay information indicates the state of one or more first models. Optionally, the thirteenth information also includes identifications of some or all of the first models; or, configuration information related to some or all of the first models. The content of the configuration information related to K first models can be referred to in the previous text. Figure 6 The content of the discussion.

[0601] For example, the third delay information includes at least one of the following: third activation delay information, the third activation delay information indicates R1 third activation delays, one of the third activation delays indicates the delay of activating one of the R1 first models; third reasoning delay information, the third reasoning delay information indicates R2 third reasoning delays, one of the third reasoning delays indicates the delay of reasoning one of the R2 first models; third fallback delay information, the third fallback delay information indicates R3 third fallback delays, one of the third fallback delays indicates the delay of falling back one of the R3 first models to not using the model to implement the first function, and one first model can implement the first function; third monitoring delay information, the third monitoring delay information indicates R2 third reasoning delays, one of the third reasoning delays indicates the delay of reasoning one of the R2 first models. Information indicating R4 third monitoring delays, one of which indicates the delay of monitoring one of the first models in R4; fifth switching delay information, the fifth switching delay information indicating R5 fifth switching delays, one of which indicates the delay of switching one of the first models in R5 to the third model; sixth switching delay information, the sixth switching delay information indicating R6 sixth switching delays, one of which indicates switching the fourth model to one of the first models in R6; or, third deactivation delay information, the third deactivation delay information indicating R7 third deactivation delays, one of which indicates the delay of deactivating one of the first models in R7. Wherein, R1, R2, R3, R4, R5, R6 and R7 are all positive integers.

[0602] For example, the third activation delay information indicates that the state of P1 first model among R1 first models is the first state, and P1 is a positive integer; the fifth switching delay information indicates that the state of Q1 first model among R5 first models is the first state, and Q1 is a positive integer; the sixth switching delay information indicates that the state of R1 first model among R6 first models is the first state, and R1 is a positive integer; or, the third inference delay information indicates that the state of L1 first model among R2 first models is the first state, and L1 is a positive integer.

[0603] For example, if the values ​​of some or all of the third activation delays in R1 are in the value range #13, then the state of the P1 first model is the first state; for example, if the values ​​of some or all of the fifth switching delays in R5 are in the value range #14, then the state of the Q1 first model is the first state; for example, if the values ​​of some or all of the sixth switching delays in R6 are in the value range #15, then the state of the R1 first model is the first state; for example, if the values ​​of some or all of the third inference delays in R2 are in the value range #16, then the state of the L1 first model is the first state.

[0604] Or, for example, if the values ​​of some or all of the third activation delays in R1 third activation delays (such as less than or equal to P1 third activation delays) are all in the value range #17, then the state of the P first models is the first state; for example, if the values ​​of some or all of the fifth switching delays in R5 fifth switching delays (such as less than or equal to Q1 fifth switching delays) are all in the value range #18, then the state of the Q1 first model is the first state; for example, if the values ​​of some or all of the sixth switching delays in R6 sixth switching delays (such as less than or equal to R1 sixth switching delays) are in the value range #19, then the state of the R first models is the first state; for example, if the values ​​of some or all of the third reasoning delays in R2 third inference delays (such as less than or equal to the third reasoning delay) are in the value range #19, then the state of the L1 first model is the first state.

[0605] S1903: The terminal device sends fourteenth information to the model management device. Correspondingly, the model management device receives the fourteenth information from the terminal device. The fourteenth information may indicate the status of some or all of the one or more first models.

[0606] The terminal device may determine that the state of at least one of the one or more first models is updated after the first time window, or after the second time window, and send the fourteenth information to the model management device. In this case, some or all of the first models may be the at least one first model, or may be one or more first models. Alternatively, the terminal device uses the first period as the period for sending the state of the first model. Therefore, when the terminal device determines that the first moment has been reached, it can send the fourteenth information. In this case, some or all of the first models may also be the first models whose states are updated in one or more first models, or may be one or more first models. The embodiments of the present application do not limit this.

[0607] Optionally, the length of the first cycle may be greater than or equal to the first time window, and may also be greater than or equal to the second time window.

[0608] S1903 is an example of the terminal device actively sending the fourteenth information to the model management device. In another possible implementation, the terminal device may notify the model management device of the state of part or all of the first model when the model management device triggers it. This possible implementation may include the contents of S1904 and S1905, and these two steps are introduced below.

[0609] S1904: The model management device sends indication information #1 to the terminal device. Correspondingly, the terminal device receives indication information #1 from the model management device. Indication information #1 indicates reporting the status of some or all of the one or more first models.

[0610] The model management device may also execute S1904 after the first time window or the second time window, or the model management device may execute S1904 after reaching the first moment, which is not limited in the embodiment of the present application.

[0611] S1905. The terminal device sends the fifteenth information to the model management device. Accordingly, the model management device receives the fifteenth information from the terminal device. The fifteenth information may indicate the status of some or all of the first models in one or more first models. The part or all of the first models may be at least one first model whose status is updated in the one or more first models, or may be one or more first models, which is not limited in the embodiments of the present application.

[0612] It should be noted that the terminal device can actively report the status of part or all of the first model using the above-mentioned S1903 method, or can also use S1904-S1905 methods to report the status of part or all of the first model. This embodiment of the application does not limit this.

[0613] It is understandable that in order to implement the functions in the above embodiments, the model management device and the terminal device include hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and method steps of each example described in the embodiments disclosed in this application, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application scenario and design constraints of the technical solution.

[0614] Fig. 20 and Fig.21 The schematic diagram of the structure of possible communication devices provided in the embodiments of the present application. These communication devices can be used to implement the functions of the terminal device or the model management device in the above method embodiments, and thus can also achieve the beneficial effects possessed by the above method embodiments. In the embodiments of the present application, the communication device can be as follows: Figure 1-Figure 3 Any terminal device or software or hardware module in the terminal device, or the foregoing Figure 4 The model application device or the software or hardware module in the model application device may also be Figure 4 The model management device shown or the software or hardware model in the model management device.

[0615] like Fig. 20 As shown, the communication device 2000 includes a processing module 2010 and a transceiver module 2020 .

[0616] Exemplarily, the communication device 2000 may be used to implement the above Figure 6-Figure 11 and Figure 18-19 The functions of the terminal device or the model management device in any of the method embodiments shown.

[0617] In the first embodiment, the communication device 2000 is used to implement the above Figure 6 The model application device in the method embodiment shown, such as the terminal device, has the function, and accordingly, the communication device can also implement the above Figure 6 The steps performed by a model application devic...

Claims

1. A communication method, characterized in that: The method is performed by a model application device or a chip used for the model application device, and includes: First information is sent to a model management device, wherein the first information indicates that a state of one or more first models is a first state, the first state indicates that the model application device supports the use of the one or more first models, and the first state is valid within a first time window.

2. The method according to claim 1, characterized in that The method further comprises: After the first time window, determine that the states of M first models among the one or more first models are updated to a second state, where the second state indicates that the model application device does not support the use of the M first models, and M is a positive integer; Sending second information to the model management device, where the second information indicates that the states of the M first models / functions are updated to the second state.

3. The method according to claim 2, characterized in that Before sending the second information to the model management device, the method further includes: It is determined that a first moment has been reached, wherein a time interval between the first moment and the moment of sending the first information satisfies a first period, and the first period is a period for the model application device to report the status of the M first models.

4. The method according to claim 3, characterized in that The length of the first time window is less than or equal to the length of the first period.

5. The method according to claim 1, characterized in that The method further comprises: After the first time window, and at a first moment, it is determined that the states of M first models among the one or more first models are not updated, wherein the time interval between the first moment and the moment of sending the first information satisfies a first period, and the first period is a period for the model application device to report the states of the M first models, and M is a positive integer; Determine a state of skipping sending the M first models to the model management device at the first moment.

6. The method according to claim 1, characterized in that The method further comprises: After the first time window, receiving first indication information from the model management device, the first indication information indicating reporting status of some or all of the one or more first models; Sending third information to the model management device, wherein the third information indicates a state of the part or all of the first model.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Fourth information is sent to the model management device, wherein the fourth information indicates that the status of one or more second models is a third status, and the third status indicates that the model application device does not support the use of the one or more second models.

8. The method according to claim 7, characterized in that The third state is valid within a second time window, wherein the second time window is the same as or different from the first time window.

9. The method according to claim 7, characterized in that: The method further comprises: Send fifth information to the model management device, wherein the fifth information indicates that the status of some or all of the one or more second models is updated to a fourth status, and the fourth status indicates that the model application device supports the use of some or all of the second models, wherein the difference between the time when the model application device sends the fifth information and the time when the model application device sends the first information is less than the length of the first time window.

10. The method according to claim 9, characterized in that The fourth state is valid within a third time window, wherein the third time window is the same as or different from the first time window.

11. The method according to any one of claims 1 to 10, characterized in that: The method further comprises: sending second indication information to the model management device, where the second indication information indicates the first time window; or, Third indication information is received from the model management device, where the third indication information indicates the first time window.

12. The method according to any one of claims 1 to 11, characterized in that: The first information includes first delay information, the first delay information indicating one or more delays associated with all or part of the one or more first models; Part or all of the first delay information indicates the first state.

13. The method according to claim 12, characterized in that The first information also includes: an identification of the part or all of the first model; or, Configuration information related to the part or all of the first model.

14. The method according to claim 12 or 13, characterized in that The first delay information includes at least one of the following: first activation delay information, where the first activation delay information indicates K1 first activation delays, wherein one first activation delay indicates a delay for activating one of the K1 first models; First inference delay information, wherein the first inference delay information indicates K2 first inference delays, wherein one first inference delay indicates a delay for inferring one of the K2 first models; First fallback delay information, the first fallback delay information indicating K3 first fallback delays, wherein one first fallback delay indicates a delay for falling back one of the K3 first models to a delay for implementing a first function without using a model, and the one first model can implement the first function; First monitoring delay information, wherein the first monitoring delay information indicates K4 first monitoring delays, wherein one first monitoring delay indicates a delay for monitoring one first model among the K4 first models; first switching delay information, wherein the first switching delay information indicates K5 first switching delays, wherein one first switching delay indicates a delay for switching one of the K5 first models to a third model; second switching delay information, where the second switching delay information indicates K6 second switching delays, wherein one second switching delay indicates switching the fourth model to one of the K6 first models; or, first deactivation delay information, where the first deactivation delay information indicates K7 first deactivation delays, wherein one first deactivation delay indicates a delay for deactivating one of the K7 first models; Wherein, K1, K2, K3, K4, K5, K6 and K7 are all positive integers.

15. The method according to claim 14, characterized in that Part or all of the first delay information indicates the first state, including at least one of the following: The first activation delay information indicates that the states of P first models among the K1 first models are the first state, where P is a positive integer; The first switching delay information indicates that the states of Q first models among the K5 first models are the first state, where Q is a positive integer; The second switching delay information indicates that the states of R first models among the K5 first models are the first state, where R is a positive integer; or, The first inference delay information indicates that the states of L first models among the K2 first models are the first state, where L is a positive integer.

16. The method according to claim 15, characterized in that If values ​​of some or all of the K1 first activation delays are within the first value range, then the states of the P first models corresponding to the some or all of the first activation delays are the first state; If values ​​of some or all of the first switching delays in the K5 pieces of first switching delay information are within the second value range, then the states of the Q first models corresponding to the some or all of the first switching delays are the first states; If the values ​​of some or all of the second switching delays in the K5 second switching delays are within the third value range, then the states of the R first models corresponding to the some or all of the second switching delays are the first states; If the values ​​of some or all of the K2 first inference delays are within the fourth value range, then the states of the L first models corresponding to the some or all of the first inference delays are the first states.

17. The method according to claim 16, characterized in that The length of the first time window satisfies at least one of the following conditions: is greater than the maximum value within the first value range; is greater than the maximum value in the second value range; is greater than the maximum value in the third value range; or, Greater than the maximum value within the fourth value range.

18. The method according to any one of claims 1 to 17, characterized in that: The method further comprises: Within the first time window, sixth information is sent to the model application device, where the sixth information indicates a strategy for managing N first models among the one or more first models, where N is a positive integer.

19. The method according to any one of claims 7 to 10, characterized in that: The fourth information includes second delay information, wherein the second delay information indicates one or more delays associated with all or part of the one or more second models, wherein part or all of the second delay information also indicates the third state.

20. A communication device, characterized in that: include: A module for executing the method according to any one of claims 1-19.

Images