Communication method and communication device

Abstract

The invention provides a communication method and a communication device, which are applied to the technical field of communication. For example, the communication method is adapted to an NTN scene, and the method comprises: a network device issues information related to a first AI model to a UE, so that the UE predicts channel information (such as time delay and / or Doppler frequency shift) at a future moment based on the first AI model. Compared with a mode in the prior art that the UE calculates the channel information (including time delay and / or Doppler frequency shift) through a geometric calculation mode, the first AI model in the embodiment of the invention considers historical measurement data (or historical rules), and the UE predicts the channel information (including time delay and / or Doppler frequency shift) at the future moment through the first AI model; the accuracy of the prediction result can be improved, so that it is ensured that the time and / or frequency of the transmitted signal arriving at the satellite are / is synchronous.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication, and in particular, to a communication method and a communication apparatus. BACKGROUND

[0002] With the development of communication technology, communication networks (such as the fifth generation communication network, the sixth generation communication network) can be integrated with satellite communication systems (such as non-terrestrial networks (NTN)), to solve the coverage blind area problem (such as mountainous areas, oceans or deserts, etc.) of ground networks, and to provide better communication services for users. The space-air-ground integrated communication network for 6G has become a research hotspot in the communication industry. NTN link prediction is an important cornerstone for realizing future space-air-ground integrated communication networks. In the existing NTN link prediction method, before transmitting a signal, the UE uses a geometric-based pre-compensation method to calculate compensation information based on ephemeris data and other information, such as calculating the propagation delay and the Doppler shift, so as to offset the propagation delay and the Doppler shift before transmitting the signal, to ensure that the transmitted signal is synchronized in time and frequency when it reaches the satellite.

[0003] However, for complex and variable NTN network environments, the UE using the geometric calculation method has certain limitations, such as the results obtained are not accurate enough, which can cause a series of problems (such as the uplink transmitted signal being out of synchronization in time and / or frequency when it reaches the satellite). SUMMARY

[0004] Therefore, the present application provides a communication method, a communication apparatus, a chip system, a computer readable storage medium, a computer program product and a communication system, which can improve the accuracy of the prediction result by predicting the channel information (including the time delay and / or the Doppler shift) at the future time through the first AI model, so as to ensure that the transmitted signal is synchronized in time and / or frequency when it reaches the satellite.

[0005] In a first aspect, a communication method is provided, which can be executed by a user equipment (UE) or a component (such as a circuit, a chip or a chip system, etc.) configured in the UE, and can also be implemented by a logic module or software capable of realizing all or part of the functions of the UE. The present application does not limit this.

[0006] Specifically, the method comprises: receiving, by the UE, a first message from the network device, the first message comprising information related to a first AI model, the information related to the first AI model comprising one or more of: a network structure of the first AI model, a layer connection relationship of the first AI model, a model parameter of the first AI model; the first AI model being used to output predicted information at a time t+k from input information at the time t, wherein ; determining, based on the input information and the first AI model, a prediction result, the prediction result comprising one or more of: a time delay and a Doppler shift; performing pre-compensation processing based on the prediction result to obtain a compensated transmission time and / or a compensated transmission frequency; and transmitting a signal based on the compensated transmission time and / or the compensated transmission frequency to the network device (for example, transmitting a signal with a time delay predicted by the first AI model, the time delay corresponding to a time length in advance).

[0007] Based on the above technical solution, the UE receives information related to the first AI model issued by the network device, so that the UE predicts channel information (such as time delay and / or Doppler shift) at a future time based on the first AI model. Compared with the way in which the UE calculates channel information (including time delay and / or Doppler shift) by geometric calculation in the prior art, the first AI model in the embodiment of the application takes into account historical measurement data (or historical regularity), and the UE predicts channel information (including time delay and / or Doppler shift) at a future time through the first AI model, which can improve the accuracy of the prediction result, thereby ensuring that the transmitted signal is synchronized in time and / or frequency when it reaches the satellite.

[0008] It should be noted that, in order to realize uplink pre-compensation (Uplink Pre-compensation), the UE can use the time delay predicted by the first AI model to replace the time delay calculated by the UE using a geometric algorithm in the traditional method, and adjust the uplink transmission time based on the replaced time delay; and / or use the Doppler shift predicted by the first AI model to replace the Doppler shift calculated by the UE using a geometric algorithm in the traditional method, and adjust the uplink transmission carrier frequency based on the replaced Doppler shift, to meet the synchronization requirements of NTN transmission in 3GPP.

[0009] Optionally, the prediction result further comprises a predicted value of a Doppler rate. Since there is a certain time interval (such as performing inference every T milliseconds) in the inference of the first AI model, the UE can utilize as the slope, and combine The UE can perform linear extrapolation on the frequency compensation value using the Doppler rate of change, thereby performing pre-compensation using the frequency compensation value. In this way, it can be ensured that the frequency compensation curve applied to the physical layer is continuous and smooth, eliminating the jump error caused by the traditional step compensation.

[0010] To input the above input information into the first AI model, the UE can first construct a corresponding feature vector based on the input information, which can be in the form of a time series. In one possible implementation, the UE determines the prediction result according to the input information and the first AI model, including: constructing an input feature vector, and inputting the input feature vector into the first AI model, the input feature vector at least including a first feature sequence, the first feature sequence including one or more of the following sequences: a propagation delay observation value sequence, a Doppler shift sequence, and a Doppler rate of change sequence; obtaining the prediction result output by the first AI model.

[0011] Illustratively, the UE constructs an input feature vector at the t-th moment, and inputs the input feature vector at the t-th moment into the first AI model, and the first AI model outputs prediction information at the t+k-th moment.

[0012] In one possible implementation, the input feature vector further includes a second feature sequence and / or a third feature sequence; the second feature sequence is a feature sequence related to the motion of the UE; and the third feature sequence is a feature sequence related to the spatial angle.

[0013] Optionally, the second feature sequence includes a velocity vector sequence and / or an acceleration vector sequence.

[0014] Optionally, the third feature sequence includes an included angle sequence and / or a satellite elevation angle sequence.

[0015] It can be understood that the prediction result output by the first AI model may or may not be accurate. For the UE, the UE can continuously compare the difference between the prediction result output by the first AI model and the actual channel measurement value (or true measurement value) to evaluate the performance of the first AI model; or the UE can also evaluate the performance of the first AI model through the uncertainty parameter corresponding to the prediction result. The UE can decide whether to continue using the first AI model based on the performance evaluation result of the first AI model.

[0016] Optionally, after the first AI model outputs the prediction result, a sampling algorithm can be used to calculate the uncertainty parameter corresponding to each prediction item, the uncertainty parameter being used to quantify the uncertainty of the first AI model, or in other words, to quantify the accuracy of the prediction result output by the first AI model.

[0017] In a possible implementation, the method further includes: determining, by the UE, whether the performance of the first AI model meets a first preset condition according to the prediction result and an actual measurement value of the channel; in a case where it is determined that the performance of the first AI model does not meet the first preset condition, performing, by the UE, the pre-compensation processing based on a geometric algorithm to calculate the channel information; in a case where it is determined that the performance of the first AI model meets the first preset condition, performing, by the UE, the pre-compensation processing using the prediction result output by the first AI model.

[0018] The embodiments of the present application do not limit the specific implementation of whether the first AI model meets the first preset condition. The following is described in combination with a prediction error value.

[0019] In a possible implementation, the performance of the first AI model does not meet the first preset condition, including: the prediction error value exceeds an error threshold value, and a duration for which the prediction error value exceeds the error threshold value exceeds a first duration threshold value, wherein the prediction error value is determined based on a sequence predicted by the first AI model and a sequence actually measured by the UE; and the performance of the first AI model meets the first preset condition, including: the prediction error value exceeds the error threshold value, and a duration for which the prediction error value exceeds the error threshold value does not exceed the first duration threshold value; or the prediction error value does not exceed the error threshold value.

[0020] In another possible implementation, the performance of the first AI model does not meet the first preset condition, including one or more of the following: a first uncertainty parameter related to a time delay output by the first AI model exceeds a first confidence threshold value for M consecutive times; a second uncertainty parameter related to a Doppler shift output by the first AI model exceeds a second confidence threshold value for W consecutive times; and a third uncertainty parameter related to a Doppler rate output by the first AI model exceeds a third confidence threshold value for Q consecutive times.

[0021] If the UE determines that the performance of the first AI model does not meet the first preset condition, the UE can switch to a traditional geometric algorithm to calculate the time delay and / or the Doppler shift, so as to ensure the link reliability; and the UE can also feed back an error to the network device.

[0022] Optionally, the method further includes: in a case where the performance of the first AI model does not meet the first preset condition, sending, by the UE, feedback information to the network device, the feedback information being used to notify the network device that the performance of the first AI model does not meet the first preset condition. Therefore, the present application further provides an error feedback mechanism, so that the network device can learn in time when the performance of the first AI model deteriorates.

[0023] Further, the UE can continue to run the first AI model in the background and verify the prediction result output by the first AI model using the new actual channel measurement value (i.e. the real channel measurement value measured by the UE after switching to calculating the channel information based on the geometric algorithm) to decide whether the prediction result of the first AI model can be continued to be used.

[0024] Optionally, in the case where the UE calculates the channel information based on the geometric algorithm and performs the pre-compensation processing, the method further comprises: the UE determines that the first AI model satisfies a second preset condition, and performs the pre-compensation processing again using the output result of the first AI model. Therefore, by introducing the second preset condition, the UE can resume using the prediction result of the first AI model at the appropriate time (i.e. the performance of the first AI model has recovered), thereby improving the accuracy of the prediction result.

[0025] Optionally, the first AI model satisfying the second preset condition comprises one or more of the following: a prediction error value of the time delay does not exceed a fourth error threshold, and a duration in which the prediction error value of the time delay does not exceed the fourth error threshold exceeds a second duration threshold; a prediction error value of the Doppler shift does not exceed a fifth error threshold, and a duration in which the prediction error value of the Doppler shift does not exceed the fifth error threshold exceeds the second duration threshold; a prediction error value of the Doppler rate does not exceed a sixth error threshold, and a duration in which the prediction error value of the Doppler rate does not exceed the sixth error threshold exceeds the second duration threshold.

[0026] In a possible implementation, in the case where the result output by the first AI model comprises a predicted value of the time delay, the method further comprises: the UE determines a first uncertainty parameter based on the predicted value of the time delay output by the first AI model; the first uncertainty parameter is an uncertainty parameter related to the predicted value of the time delay, and the first uncertainty parameter is used to represent the accuracy of the predicted value of the time delay output by the first AI model; and a duration of an SSB measurement time configuration (SMTC) window is determined based on the first uncertainty parameter. Exemplarily, determining the duration of the SMTC window based on the first uncertainty parameter comprises: when the first uncertainty parameter does not exceed a preset threshold, reducing the duration of the SMTC window; and when the first uncertainty parameter exceeds the preset threshold, maintaining the duration of the SMTC window. Therefore, when the first uncertainty parameter does not exceed the preset threshold, the UE shortens the duration of the SMTC window, i.e. a shorter and more accurate measurement window can be used, which helps to reduce the radio frequency energy consumption.

[0027] Optionally, before receiving the first message from the network device, the method further comprises: the UE reporting historical data to the network device; the historical data is used for training the first AI model; the historical data comprises one or more of the following information: channel state information, kinematic information, and spatial angle information; wherein the channel state information comprises one or more of the following: reference signal received power (RSRP), measurement delay, Doppler shift; the kinematic information comprises velocity and / or acceleration; the spatial angle information comprises satellite elevation angle and / or beam angle calculated based on historical time. Thus, the UE reports the historical data to the network device, so that the network device trains the first AI model based on the historical data, which can learn statistical rules (such as nonlinear dynamic rules) from the historical data, such as deep learning of nonlinear changes caused by satellite micro-vibration, attitude adjustment, and UE non-uniform motion, etc.; thereby capturing the nonlinearity, randomness of delay and Doppler jitter (such as rapid and frequent Doppler shift caused by fast movement of the UE) in complex network environment (or dynamic or unpredictable harsh communication environment, including but not limited to: atmospheric disturbance, or high-speed or non-uniform motion of the UE, or GNSS signal limited or interfered scenarios), which helps to improve the prediction accuracy of the first AI model.

[0028] The network structure of the first AI model is not specifically limited in the embodiments of the present application. Optionally, the first AI model comprises an input layer, a feature extraction layer, and a multi-head output layer; wherein a Dropout layer is inserted in the feature extraction layer; the input layer is connected to the input end of the feature extraction layer; and the output end of the feature extraction layer is connected to the multi-head output layer.

[0029] Exemplarily, the first AI model is an LSTM model based on a Bayesian model; the feature extraction layer comprises an LSTM layer 1 and an LSTM layer 2; a Dropout layer is included between the LSTM layer 1 and the LSTM layer 2; the input end of the LSTM layer 1 is connected to the input layer, and the output end of the LSTM layer 1 is connected to the Dropout layer; the input end of the LSTM layer 2 is connected to the Dropout layer, and the output end of the LSTM layer 2 is connected to the multi-head output layer; the multi-head output layer comprises a fully connected output layer 1 and a fully connected output layer 2; the fully connected output layer 1 is used for outputting the predicted value of the delay and / or the predicted value of the Doppler shift; and the fully connected output layer 2 is used for outputting the predicted value of the Doppler rate. Thus, using the LSTM model based on the Bayesian model as the first AI model, and using LSTM as the first AI model, not only can capture nonlinear rules, such as learning and compensating nonlinear delays caused by atmospheric factors, etc., which cannot be covered by traditional models; but also can achieve short-term robustness, such as when the GNSS signal is temporarily lost, the first AI model can still rely on the historical "inertia" pattern learned to continue to provide effective short-term prediction within a certain time.

[0030] In a second aspect, a communication method is provided, which can be performed by a network device, or by a component (such as a circuit, a chip or a chip system, etc.) configured in the network device, or by a logic module or software capable of implementing all or part of the functions of the network device. The present application does not limit this.

[0031] Specifically, the method comprises: training, by the network device, a first AI model according to a training data set, the first AI model being configured to output predicted information at a time t+k based on input information at the time t, the predicted information comprising one or more of: a time delay, a Doppler shift, a Doppler rate, wherein ; sending a first message, the first message comprising information related to the first AI model, the information related to the first AI model comprising one or more of: a network structure of the first AI model, a layer connection relationship of the first AI model, a model parameter of the first AI model.

[0032] Based on the above technical solution, the network device transmits information related to the first AI model, so that the UE predicts the channel information (such as time delay and / or Doppler shift) at the future time based on the first AI model. Compared with the way in the prior art that the UE calculates the channel information (including time delay and / or Doppler shift) by geometric calculation, the first AI model in the present application considers historical measurement data (or historical regularity), and the UE predicts the channel information (including time delay and / or Doppler shift) at the future time through the first AI model, which can improve the accuracy of the prediction result, thereby ensuring that the transmitted signal is synchronized in time and / or frequency when it reaches the satellite.

[0033] Optionally, the method further comprises: receiving, by the network device, an uplink signal sent by the UE based on the compensated transmission time and / or the compensated transmission frequency.

[0034] In a possible implementation, the method further includes: receiving, by the network device, historical data from one or more UEs, the historical data being used to train the first AI model; the historical data including one or more of the following: channel state information, kinematic information, and spatial angle information; the channel state information including one or more of the following: reference signal received power (RSRP), measurement delay, and Doppler shift; the kinematic information including velocity and / or acceleration; the spatial angle information including a satellite elevation angle and / or a beam angle calculated based on a historical time; and constructing a training data set according to the historical data. Thus, the network device can learn statistical rules (for example, nonlinear dynamic rules) from the historical data by receiving the historical data reported by the UE, for example, can learn the nonlinear changes caused by satellite micro-vibration, attitude adjustment, and UE non-uniform motion, and the like; and can capture the nonlinear and random time delay and Doppler jitter (for example, the Doppler shift caused by the rapid movement of the UE) in a complex network environment (or a dynamic or unpredictable poor communication environment, including but not limited to: atmospheric disturbance, or high-speed or non-uniform motion of the UE, or a GNSS signal limited or interfered scene), thereby improving the prediction accuracy of the first AI model.

[0035] The network structure of the first AI model is not specifically limited in the embodiments of the present application. Optionally, the first AI model includes an input layer, a feature extraction layer, and a multi-head output layer; the Dropout layer is inserted in the feature extraction layer; the input layer is connected to the input end of the feature extraction layer; and the output end of the feature extraction layer is connected to the multi-head output layer.

[0036] For example, the first AI model is an LSTM model based on a Bayesian model; the feature extraction layer includes an LSTM layer 1 and an LSTM layer 2; the Dropout layer is included between the LSTM layer 1 and the LSTM layer 2; the input end of the LSTM layer 1 is connected to the input layer, and the output end of the LSTM layer 1 is connected to the Dropout layer; the input end of the LSTM layer 2 is connected to the Dropout layer, and the output end of the LSTM layer 2 is connected to the multi-head output layer; the multi-head output layer includes a fully connected output layer 1 and a fully connected output layer 2; the fully connected output layer 1 is used to output the predicted value of the time delay and / or the predicted value of the Doppler shift; and the fully connected output layer 2 is used to output the predicted value of the Doppler rate of change.

[0037] Optionally, before sending the first message to the UE, the method further includes: performing compression processing on the first AI model by the network device to obtain information related to the first AI model. In this way, not only the signaling overhead can be saved, but also the computing power or memory space of the UE can be adapted.

[0038] In a possible implementation, the method further includes: receiving, by the network device, feedback information, the feedback information being used to inform the network device that the performance of the first AI model does not satisfy the first preset condition. In this way, the network device can determine whether to adjust the parameters of the first AI model or optimize the first AI model based on the feedback information of the UE, so as to improve the prediction accuracy of the first AI model.

[0039] Optionally, the performance of the first AI model not satisfying the first preset condition includes: a prediction error value exceeding an error threshold value, and a duration for which the prediction error value exceeds the error threshold value exceeding a first duration threshold value, wherein the prediction error value is determined based on a sequence predicted by the first AI model and a sequence actually measured by the UE.

[0040] Alternatively, optionally, the performance of the first AI model not satisfying the first preset condition includes one or more of the following: a first uncertainty parameter exceeding a first confidence threshold value for M consecutive times, the first uncertainty parameter being a parameter related to a time delay output by the first AI model; a second uncertainty parameter exceeding a second confidence threshold value for W consecutive times, the second uncertainty parameter being a parameter related to a Doppler shift output by the first AI model; and a third uncertainty parameter exceeding a third confidence threshold value for Q consecutive times, the third uncertainty parameter being a parameter related to a Doppler rate output by the first AI model.

[0041] The second aspect is a network device side implementation corresponding to the first aspect. The explanations (for example, including but not limited to: implementation of the first AI model, implementation of the first preset condition, and the like) and beneficial effect descriptions of the first aspect also apply to the second aspect, and will not be repeated here.

[0042] In a third aspect, a communication apparatus is provided, which includes various modules or units for performing the methods in any of the possible implementation manners of the first aspect.

[0043] In one design, the communication apparatus can include a module corresponding to each of the methods / operations / steps / actions described in the various aspects described above. The module can be a hardware circuit, or software, or a combination of hardware circuit and software.

[0044] In one design, the communication apparatus is a communication chip, which can include an input circuit or interface for transmitting information or data, and an output circuit or interface for receiving information or data.

[0045] In another design, the communication apparatus is a communication device, which can include a transmitter for transmitting information or data, and a receiver for receiving information or data.

[0046] In another design, the communication apparatus can be configured in the network device, or the communication apparatus itself can be the network device, and the communication apparatus is configured to perform the method in any possible implementation of the method in the second aspect.

[0047] In a fourth aspect, a communication apparatus is provided. The communication apparatus can include various modules or units configured to perform the method in any possible implementation of the method in the second aspect.

[0048] In one design, the communication apparatus can include a module corresponding to each of the methods / operations / steps / actions described in the various aspects above. The module can be hardware circuitry, software, or a combination of hardware circuitry and software.

[0049] In one design, the communication apparatus can be a communication chip. The communication chip can include an input circuit or interface for transmitting information or data, and an output circuit or interface for receiving information or data.

[0050] In another design, the communication apparatus can be a communication device. The communication device can include a transmitter for transmitting information or data, and a receiver for receiving information or data.

[0051] In another design, the communication apparatus can be configured in the network device, or the communication apparatus itself can be the network device, and the communication apparatus is configured to perform the method in any possible implementation of the method in the second aspect.

[0052] Optionally, the network device can be a satellite, or an access network device (e.g., a gNB), or the network device can be a network element in a core network.

[0053] In a fifth aspect, a communication apparatus is provided. The communication apparatus can include a processor. The processor can be coupled to a memory and can be configured to execute instructions or data in the memory to implement the method in any possible implementation of the method in the first aspect. Optionally, the communication apparatus can further include the memory. Optionally, the communication apparatus can further include a communication interface, and the processor can be coupled to the communication interface.

[0054] In one implementation, the communication interface can be a transceiver, or an input / output interface.

[0055] In another implementation, the communication apparatus can be a chip configured in the UE. When the communication apparatus is a chip configured in the UE, the communication interface can be an input / output interface.

[0056] In a sixth aspect, a communication apparatus is provided, which includes a processor. The processor is coupled with a memory and is configured to execute instructions or data stored in the memory to implement the method in any possible implementation of the second aspect. Optionally, the communication apparatus further includes the memory. Optionally, the communication apparatus further includes a communication interface, and the processor is coupled with the communication interface.

[0057] In an implementation form, the communication interface can be a transceiver, or an input / output interface.

[0058] In another implementation form, the communication apparatus is a chip configured in a network device. When the communication apparatus is a chip configured in a network device, the communication interface can be an input / output interface.

[0059] In a seventh aspect, a processor is provided, which includes an input circuit, an output circuit and a processing circuit. The processing circuit is configured to receive a signal through the input circuit and transmit a signal through the output circuit, so that the processor performs the method in any possible implementation of any aspect.

[0060] In a specific implementation process, the processor can be one or more chips, 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 flip-flop and various logic circuits, etc. The input signal received by the input circuit can be received and input by, for example but not limited to, a receiver, the output signal output by the output circuit can be output to and transmitted by, for example but not limited to, a transmitter, and the input circuit and the output circuit can be the same circuit which is used as the input circuit and the output circuit at different times. The embodiments of the present application do not limit the specific implementation of the processor and various circuits.

[0061] In an eighth aspect, a communication apparatus is provided, which includes a processor and a memory. The processor is configured to read instructions stored in the memory, and is configured to receive a signal through a receiver and transmit a signal through a transmitter to perform the method in any possible implementation of any aspect.

[0062] Optionally, the processor is one or more, and the memory is one or more.

[0063] Optionally, the memory can be integrated with the processor, or the memory and the processor are separately arranged.

[0064] In a practical implementation process, the memory can be a non-transitory memory, for example, a read only memory (ROM), which can be integrated on the same chip as the processor, or separately arranged on different chips. The embodiments of the present application do not limit the type of memory and the arrangement mode of the memory and the processor.

[0065] It should be understood that the related data interaction process, for example, the sending of the indication information, can be a process of outputting the indication information from the processor, and the receiving of the capability information can be a process of receiving the input capability information by the processor. Specifically, the data output by the processor can be output to a transmitter, and the input data received by the processor can come from a receiver. The transmitter and the receiver can be collectively referred to as a transceiver.

[0066] The processing device in the eighth aspect can be one or more chips. The processor in the processing device can be implemented by hardware or software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented by software, the processor can be a general-purpose processor, which is implemented by reading software codes stored in a memory. The memory can be integrated in the processor or exist independently outside the processor.

[0067] In a ninth aspect, a computer program product is provided, which includes a computer program (also referred to as code or instructions), which, when executed, causes a computer to perform the method in any possible implementation manner of any one of the aspects.

[0068] In a tenth aspect, a computer readable storage medium is provided, which stores a computer program (also referred to as code or instructions), which, when executed on a computer, causes the computer to perform the method in any possible implementation manner of any one of the aspects.

[0069] In an eleventh aspect, the embodiments of the present application provide a chip system, which includes one or more processors for calling and executing instructions stored in a memory, so that the method in each aspect or any possible implementation manner of each aspect is executed. The chip system can be composed of a chip, or can include a chip and other discrete devices.

[0070] The chip system can include an input circuit or interface for sending information or data, and an output circuit or interface for receiving information or data.

[0071] In a twelfth aspect, a communication system is provided, including the UE and the network device described above. The UE is configured to perform any possible implementation of the first aspect described above. The network device is configured to perform any possible implementation of the second aspect described above.

[0072] Optionally, the communication system can further include other devices in communication with the UE and / or the network device. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 FIG. 1 is an example diagram of a communication system; Figure 2 FIG. 2 is an example diagram of a satellite communication system to which the embodiments of the present application are applied; Figure 3 FIG. 3 is an example diagram of a communication method according to the embodiments of the present application; Figure 4 FIG. 4 is an example diagram of a network structure of a first AI model according to the embodiments of the present application; Figure 5 FIG. 5 is a schematic block diagram of a communication device according to the embodiments of the present application; Figure 6 FIG. 6 is another schematic block diagram of a communication device according to the embodiments of the present application. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the present application will be described below with reference to the accompanying drawings.

[0075] In the embodiments of the present application, “multiple” can be understood as “at least two”, and “multiple items” can be understood as “at least two items”.

[0076] In the description of the present application, unless otherwise specified, " / " represents a "or" relationship between the objects associated in front and back, for example, A / B can represent A or B; "and / or" in the present application is only a description of the association between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent: A exists alone, A and B exist at the same time, B exists alone, and three cases, where A, B can be singular or plural. And in the description of the present application, unless otherwise specified, "multiple" means two or more than two. "At least one of the following" or its similar expression refers to any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b, or c, can represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, c can be single or multiple. In addition, in order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, "first", "second" and the like are used to distinguish the same items or similar items with basically the same function and effect. The skilled in the art can understand that "first", "second" and the like do not limit the quantity and execution order, and "first", "second" and the like do not necessarily mean different.

[0077] The present application can be applied to a communication system. The mobile communication system includes, but is not limited to, the following systems, for example: a long term evolution (LTE) system, a universal mobile telecommunication system (UMTS), a worldwide interoperability for microwave access (WiMAX) communication system, a 5th generation (5G) mobile communication system or a new radio access technology (NR), and future mobile communication systems, vehicle-to-X (V2X), which can include vehicle to network (V2N), vehicle to vehicle (V2V), vehicle to infrastructure (V2I), vehicle to pedestrian (V2P), etc., LTE-V, vehicle networking, machine type communication (MTC), internet of things (IoT), LTE-M, machine to machine (M2M), etc. Among them, the 5G mobile communication system can include non-standalone (NSA) and / or standalone (SA).

[0078] The present application can also be applied to a system in which a mobile communication system is integrated with a satellite communication system. The satellite communication system includes, but is not limited to, a non-terrestrial network (NTN) system such as a high altitude platform station (HAPS) communication, etc., for example, a global navigation satellite system (GNSS), etc. Optionally, the satellite communication system includes a geostationary earth orbit (GEO) and a non-geostationary earth orbit (NGEO); or various terrestrial network (TN) systems.

[0079] Figure 1 is a schematic diagram of a communication system 100 to which embodiments of the present application are applied. The communication system 100 can include a network device, for example, a network device 110 as shown. Figure 1 The communication system 100 can also include a terminal device, for example, a terminal device 120 as shown. Figure 1 The network device 110 and the terminal device 120 can communicate with each other through a wireless link.

[0080] Figure 1 One network device 110 and one terminal device 120 are exemplarily shown. Optionally, the communication system 100 can also include a plurality of network devices and / or a plurality of terminal devices.

[0081] The network device in the present application can be a device of a network side such as an access network, a core network device, and the like. The access network device is also sometimes referred to as an access node. The access network device has a wireless transceiving function and is used to communicate with a terminal. The access network device includes, but is not limited to, a base station in the above-mentioned communication system, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next generation NodeB (gNB) in a 5G mobile communication system, an access network device or a module of an access network device in an open RAN (ORAN) system, a satellite in an NTN communication system, a base station in a future mobile communication system, or an access node in a WiFi system, and the like. The access network device can also be a module or unit capable of realizing part of the function of a base station. The access network device can be a macro base station, a micro base station or an indoor station, a relay node or a donor node, or a wireless controller in a cloud radio access network (CRAN) scenario. Optionally, the access network device can also be a server, a wearable device, or a vehicle-mounted device, and the like. For example, the access network device in vehicle to everything (V2X) technology can be a road side unit (RSU). Multiple access network devices in a communication system can be the same type of base station or different types of base stations. The base station can communicate with the terminal directly or through a relay station. The terminal can communicate with multiple base stations in different access technologies. The embodiments of the present application do not limit the specific technology and specific device form of the access network device. In the present application, the access network device or the core network device can be referred to as a network device.

[0082] In the present application, the device for realizing the function of the network device can be a network device or a device capable of supporting the network device to realize the function, such as a processor, a circuit, a chip, or a chip system, and the like, which can be installed in the network device or connected with the network device for use. In the technical solutions provided in the present application, the device for realizing the function of the network device is taken as an example to describe the technical solutions provided in the present application.

[0083] The terminal device in the present application can be a wireless terminal device capable of receiving network device scheduling and indication information. The wireless terminal device can be a device that provides voice and / or data connectivity to a user, or a handheld device with wireless connection function, or other processing devices connected to a wireless modem. For example, the terminal device can communicate with one or more core networks or the Internet through a radio access network (RAN). The terminal device can also be referred to as a terminal, user equipment (UE), mobile station, mobile terminal, etc. The terminal device can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), internet of things (IOT), ultra-reliable low-latency communication (URLLC), virtual reality, augmented reality, industrial control, autonomous driving, remote medical treatment, smart power grid, smart furniture, smart office, smart wear, smart transportation, smart city, or satellite communication, etc. The terminal can be a mobile phone, tablet computer, computer with wireless transceiver function, wearable device, vehicle, aircraft (such as unmanned aerial vehicle, helicopter, airplane), hot air balloon, ship, robot, mechanical arm, or smart home device, etc. The embodiments of the present application do not limit the form of the terminal device.

[0084] In the present application, the device for realizing the function of the terminal device can be a terminal device, or a device capable of supporting the terminal device to realize the function, such as a processor, circuit, chip, chip system, etc., which can be installed in the terminal device or connected with the terminal device for use. In the technical solutions provided in the present application, the device for realizing the function of the terminal device is taken as an example to describe the technical solutions provided in the present application.

[0085] The access network device and / or the terminal device can be fixed or mobile. The access network device and / or the terminal device can be deployed on land, including indoor or outdoor, handheld or vehicle-mounted; can also be deployed on the water surface; can also be deployed on aircraft, balloons and artificial satellites in the air. The application embodiments do not limit the application scenarios of the access network device and the terminal device. The access network device and the terminal device can be deployed in the same scenario or different scenarios, for example, the access network device and the terminal device are deployed on land at the same time; or the access network device is deployed on land and the terminal device is deployed on the water surface, etc., which will not be listed one by one.

[0086] In actual application, wireless access can be implemented by a terminal assisted by multiple network devices, and different network devices respectively implement part of the functions of a base station. For example, a network device can be a central unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU), etc. The CU and the DU can be separately arranged or can be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or a radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH).

[0087] In different systems, the CU (or CU-CP and CU-UP), DU, or RU can also have different names, but those skilled in the art can understand their meanings. For example, in an ORAN system, the CU can also be referred to as an O-CU (open CU), the DU can also be referred to as an O-DU, the CU-CP can also be referred to as an O-CU-CP, the CU-UP can also be referred to as an O-CU-UP, and the RU can also be referred to as an O-RU. Any of the CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module. The CU (or CU-CP and CU-UP), DU, and RU can implement different protocol layer functions.

[0088] To facilitate understanding of the embodiments of the present application, first, the terms involved in the present application are simply explained. For example, the explanation of some terms can also refer to the explanation in the 3rd generation partnership project (3GPP) standard protocol.

[0089] 1. Calculate the delay and Doppler shift based on the geometric algorithm In the 3GPP NTN standard, geometric-based pre-compensation is adopted to realize synchronization of the UE transmitted signal. Specifically, before transmitting a signal to a network device (such as a satellite or a base station), the UE autonomously calculates the delay and the Doppler shift, and offsets (or performs pre-compensation) the delay and the Doppler shift when transmitting the signal, so as to ensure that the transmitted signal is synchronized in time and frequency when it reaches the satellite.

[0090] Timing Advance (TA), generally used for UE uplink transmission, refers to the time delay caused by the distance between the UE and the satellite, which is used to estimate the radio frequency transmission delay caused by the distance, so as to send the data packet in advance. The purpose of introducing TA is to align the timing of uplink transmission and downlink transmission, or in other words, to align the time between the UE and the satellite. In the TN of the ground network, the TA is usually controlled by the base station, for example, the base station determines the value of the TA through the random access preamble, and notifies the UE of the TA. In the satellite communication system, the UE reports the TA to the satellite / base station.

[0091] 1) Transmission delay At present, the 3GPP NTN stipulates that the timing advance (Timing Advance, TA) of the physical layer of the UE satisfies the following formula: ; wherein, represents the common delay; is calculated by the UE according to the geometric distance; satisfies the following formula: ; wherein, represents the distance between the UE and the satellite; represents the speed of light.

[0092] It can be understood as the transmission delay (or propagation delay). The round-trip time (RTT) of the backhaul can be introduced to represent .

[0093] It should be noted that in order to ensure that the signal reaches the satellite in time synchronization, the UE will send the signal in advance time, or in other words, the transmission time is pre-compensated by time; The in the time is calculated by the UE through a geometric algorithm.

[0094] In some embodiments, the UE calculates the RTT of the service link based on its own position information (such as GNSS position, or the latitude and longitude of the position obtained through the positioning function), and the ephemeris information of the satellite obtained through the system information block (system information block, SIB) 19.

[0095] For ease of understanding, the following examples are combined with the satellite communication system shown in Figure 2 . As Figure 2As shown, the satellite communication system provides wide-area coverage communication services for multiple ground terminals. Figure 2 The image shows the satellite's coverage area at times T1, T2, T3, and T4. The satellite moves along a certain direction from time T1 to T4. The satellite's coverage area can include one or more UEs (e.g., [UEs]). Figure 2 (UE1 to UE5 are shown in the diagram). The transmission time domain between multiple UEs and the satellite can be different. As the satellite moves, the propagation distance between the UEs within the satellite's coverage area and the satellite will change. As the propagation distance between the UE and the satellite changes, the transmission delay of data between the UE and network devices (e.g., satellite / base station) will change accordingly.

[0096] like Figure 2 As shown, at time T1, the transmission delays received by the satellite from UE1, UE2, and UE3 are relatively small and similar, while the transmission delays from UE4 and UE5 are relatively large and similar. At time T2, the transmission delays received by the satellite from UE1 to UE5 are all relatively similar. As the satellite continues to move, the transmission delays of UE1, UE2, and UE3 gradually increase, while the transmission delays of UE4 and UE5 gradually decrease.

[0097] by Figure 2 Taking UE1 as an example, at time T1, the transmission delay between UE1 and the satellite is minimal; at time T3, the transmission delay between UE1 and the satellite is maximum. At time T4, after the satellite moves, its coverage area can no longer reach UE1, UE2, and UE3. At this time, the satellite stops providing services to UE1, UE2, and UE3.

[0098] As can be seen, the relative motion between the UE and the satellite causes changes in transmission latency. Furthermore, high-speed movement of the UE itself will also cause changes in transmission latency. For example, a scenario where the UE moves at high speed is a user carrying a mobile phone on a high-speed train, airplane, or other high-speed transportation.

[0099] It should be understood that the satellite mentioned in the embodiments of this application may also be called a satellite base station or a network-side device mounted on a satellite.

[0100] In order to accurately predict time delay and / or Doppler shift, embodiments of this application introduce a first AI model to predict time delay and / or Doppler shift at future times based on the input information at the current time.

[0101] 2) Doppler Shift It should be noted that, in order to ensure that the signal is synchronized on the frequency when it arrives at the satellite, the UE will subtract the Doppler frequency shift value from the transmission frequency when transmitting the uplink signal, or in other words, use the Doppler frequency shift value to perform pre-compensation on the transmission frequency.

[0102] Exemplarily, the UE subtracts the carrier frequency from the final transmission frequency , to obtain the final transmission frequency ; the UE transmits the uplink signal using the transmission frequency ; that is, ; wherein, is a Doppler shift value; satisfies the following formula: ； wherein, represents the relative speed of the satellite and the UE in the line-of-sight direction; is calculated by combining the satellite speed vector and the position vector of the UE; represents the speed of light; represents the carrier frequency (i.e., the aforementioned ). is a reference parameter for calculating the Doppler shift amount.

[0103] 2. Ephemeris information In some embodiments, the ephemeris information can be carried in SIB19. SIB19 is configuration information in NTN communication, or in other words, satellite-related configuration information. The UE can obtain the ephemeris information through SIB19.

[0104] The present embodiments do not make specific limitations on the format of the ephemeris information. Optionally, the ephemeris information includes one or more of the following: the position (such as the latitude and longitude) of the satellite, the speed vector, and the orbital parameter information. Optionally, the ephemeris information can also include the motion trajectory of the serving satellite in a period of time.

[0105] It should be understood that the ephemeris information is periodically updated. For example, the ephemeris information is updated every 15 minutes.

[0106] Currently, in the NTN standard of 3GPP, the UE calculates the delay and Doppler based on a geometric algorithm. However, due to the complex and variable environment (such as random jitter caused by atmospheric disturbance or non-uniform motion of the UE), the traditional geometric calculation method has certain limitations and cannot be applied to complex and variable network environments, for example, the calculated delay and Doppler are not suitable for the current environment.

[0107] Therefore, an embodiment of the present application proposes a communication method, which introduces an AI model to predict the time delay and Doppler shift, or in other words, to predict the future channel state, which helps to improve the accuracy of the prediction result. Further, by quantifying the uncertainty parameter of the AI model to obtain the reliability of the prediction result, it is determined whether to use the prediction result output by the AI model, which helps to balance the relationship between link reliability and energy efficiency.

[0108] The schemes provided by the present application will be described in detail below in combination with corresponding flowcharts. It can be understood that the main body of the interaction in the illustrative flowcharts provided by the present application is taken as an example to illustrate the method, for example, the device (such as a UE, a network device) in the illustrative flowchart. However, the present application does not limit the main body of the interaction. For example, the device (such as a UE, a network device) in the illustrative flowchart can also be a chip, a chip system, or a processor supporting the device to implement the method, and can also be a logic module or software capable of implementing all or part of the functions of the device.

[0109] Here, it is uniformly stated that the messages or signaling interactions involved in the interaction process of the embodiments of the present application can adopt the messages or signaling in the standard or newly introduced messages or signaling, and the embodiments of the present application do not make specific limitations.

[0110] Figure 3 is an example of a communication method of an embodiment of the present application. It can be understood that Figure 3 the UE in Figure 1 may be the terminal device 120 in Figure 1 may be the network device 110 in Figure 3 Or, Figure 2 any UE in Figure 3 the network device in Figure 2 is a satellite in Figure 3 As shown in the method at least includes the following steps: Step 310, the network device sends a first message, the first message includes information related to a first AI model, the first AI model is used to output prediction information at the t+k moment according to the input information at the t moment, wherein

[0111] It should be understood that the introduction of the t moment is a general reference to a moment, the t+k moment is used to refer to a moment after the t moment, or in other words, a future moment, which is introduced only for the convenience of description, and the embodiments of the present application are not limited thereto. Moreover, t or The value of the first AI model can depend on the specific implementation, and embodiments of the present application do not make specific limitations.

[0112] The information related to the first AI model can be understood as a model file of the first AI model. After obtaining the first AI model (such as training the first AI model), the network device can distribute the related model file of the first AI model to the UE, so that the UE uses the first AI model to perform prediction.

[0113] In some embodiments, the information related to the first AI model includes one or more of the following: network structure of the first AI model, layer connection relationship of the first AI model, model parameters (or weight parameters) of the first AI model.

[0114] For the UE, after receiving the information related to the first AI model, the UE can reconstruct the first AI model locally to use the first AI model to perform real-time inference or prediction, so as to obtain the channel information (such as time delay and / or Doppler shift) predicted for the future time.

[0115] Embodiments of the present application do not make specific limitations on the message type or signaling type of the first message. Alternatively, the first message can be a broadcast message or an RRC message. For example, the network device distributes the information related to the first AI model to the UE through a dedicated container (Container) or certain information element (IE) in the RRC reconfiguration message (RRCReconfiguration). Alternatively, the first message is a non-access layer (non-access stratum, NAS) message. Alternatively, the first message can be a newly introduced message or signaling dedicated to distributing information related to the first AI model.

[0116] The network device can distribute the information related to the first AI model to one or more UEs. Of course, the network device can perform the training process of the first AI model before distribution, or obtain the first AI model that has been trained from the cloud. Alternatively, before step 310, the method further includes: step 301, the network device trains the first AI model according to the training data set.

[0117] Optionally, since the network device side has more abundant resources, the network device can perform the training process of the first AI model. The network device trains the first AI model by constructing a training data set. For example, the first AI model is a Bayesian long short-term memory (B-LSTM) model; the network device trains the first AI model by using cloud computing power, so that the first AI model learns the delay and the non-linear variation law of the Doppler under the current satellite orbit, and prepares for subsequent output of prediction results.

[0118] It should be noted that by introducing the first AI model, the statistical law (for example, the non-linear dynamic law) can be learned from the historical data, for example, the non-linear variation caused by satellite micro-vibration, attitude adjustment, and UE non-uniform motion, and the like can be deeply learned; thereby capturing the non-linear and random delay and Doppler jitter (for example, the Doppler shift caused by the rapid movement of the UE) under a complex network environment (or a dynamic or unpredictable poor communication environment, including but not limited to: atmospheric disturbance, or UE high-speed movement or non-uniform movement, or, GNSS signal limited or interfered scenarios).

[0119] It can be understood that the embodiments of the present application do not specifically limit the execution subject of training the first AI model, and the embodiments of the present application are only described by taking the network device as an example of the execution subject of training the first AI model. The embodiments of the present application are not limited thereto. In fact, the execution subject of training the first AI model can be other training devices. For example, the training process of the first AI model is performed in the cloud. For another example, if the resources of the UE side are sufficient with the development or technical evolution of the UE side, the training process of the first AI model can also be performed by the UE side, and of course, the training process of the first AI model can refer to the process of training the first AI model by the network device of the embodiments of the present application.

[0120] It can also be understood that the training process of the first AI model can be an offline training manner or an online training manner, and the embodiments of the present application do not specifically limit this.

[0121] In some embodiments, before performing the training of the first AI model, the network device can construct a training data set in advance. The training data set can be constructed by the network device based on the historical measurement data reported by the UE. Optionally, before step 301, the method further includes: step 300, the UE reports historical data to the network device. Correspondingly, the network device receives the historical data.

[0122] For the UE, the UE can continuously collect or calculate the historical data. Optionally, the historical data comprises one or more of: channel state information, kinematic information, and spatial angle information; wherein the channel state information comprises one or more of: reference signal receiving power (RSRP), measurement delay, Doppler shift; the kinematic information comprises velocity and / or acceleration; and the spatial angle information comprises satellite elevation angle and / or beam angle calculated based on a historical time.

[0123] Embodiments of the present application do not make specific limitations on the message type in which the historical data is located. Optionally, the historical data is carried in a measurement report, such as a Minimization Drive Test (MDT); or the historical data is carried in UE Assistance Information (UAI) signaling. For example, the UE reports the historical data including the time delay sequence, the Doppler shift sequence, and the kinematic characteristics, etc. to the network device through the MDT or the UE assistance information.

[0124] Correspondingly, the network device can pre-collect historical measurement data reported from one or more UEs (or different UEs), and construct a training data set based on the historical measurement data. The network device trains the first AI model using the training data set.

[0125] After the network device completes the training of the first AI model, the network device sends information related to the first AI model to the UE, so that the UE performs prediction based on the first AI model. In some embodiments, before sending the first message to the UE, the method further comprises: the network device performing compression processing on the first AI model (or the model file of the first AI model) to obtain the information related to the first AI model.

[0126] It should be noted that if the network device directly issues the model file of the first AI model, it may cause an increase in signaling overhead or affect the transmission efficiency of the transmission model file. Therefore, after the network device trains the first AI model, the network device can perform compression processing on the first AI model, and issue the model file after compression processing to the UE. In this way, not only the signaling overhead can be saved, but also the computing power or memory space of the UE can be adapted. That is, the purpose of the "compression processing" is to perform lightweight compression processing (such as quantization and pruning processing) on the first AI model, so as to adapt to the computing power of the UE or the memory space of the UE. Quantization and pruning processing can be understood as a model optimization technique, and specific descriptions can be referred to in related technologies, and will not be expanded here.

[0127] At step 320, the UE determines a prediction result according to the input information and the first AI model, the prediction result including one or more of the following predicted values: time delay, Doppler shift, and Doppler rate.

[0128] The input information at least includes channel measurement information at a historical moment and channel measurement information at a current moment. For example, the input information includes channel measurement information at the t th moment and channel measurement information at the t th moment and the previous N moments. In some embodiments, the UE inputs, at the t th moment (i.e., the current moment), the channel measurement information at the t th moment and the channel measurement information at the t th moment and the previous N moments to the first AI model; obtains an output result of the first AI model, the output result including prediction information at the t th + k moment (wherein k is a positive integer).

[0129] Of course, the above is only described by taking an example that the input information at least includes channel measurement information at a historical moment and channel measurement information at a current moment, and the embodiments of the present application are not limited thereto. For example, the input information can not include channel measurement information at a current moment, but include channel measurement information at a historical moment.

[0130] For example, the channel measurement information at the t th moment includes time delay at the t th moment and / or Doppler shift at the t th moment; and the channel measurement information at the t th moment and the previous N moments includes time delay at each of the previous N moments and / or Doppler shift at each of the previous N moments.

[0131] Further, the input information further includes speed information and / or spatial angle information. The speed information of the UE includes the moving speed and / or acceleration of the UE. The spatial angle information includes satellite elevation angle and / or included angle. The included angle refers to the included angle between the UE and the center of the beam transmitted by the satellite.

[0132] For example, the input information further includes speed information at the t th moment and spatial angle information at the t th moment, and speed information at each of the previous N moments and spatial angle information at each of the previous N moments.

[0133] The embodiments of the present application do not make specific limitations on how to obtain the moving speed of the UE. Optionally, the UE includes a positioning module, and the UE obtains positioning data through the positioning module, the positioning data including the moving speed of the UE. In some embodiments, the positioning module is a GNSS module. For example, the GNSS module is a GPS or Beidou module.

[0134] ​The embodiments of the present application do not make specific limitations on how to obtain the acceleration of the UE. In some embodiments, the acceleration can be calculated by the aforementioned moving speed of the UE. Alternatively, in some embodiments, the UE includes a measurement device such as an inertial measurement unit or an accelerometer sensor or a gyroscope, that is, the UE can obtain the acceleration through the measurement device for measuring the acceleration. For example, for the scene of detecting the non-linear motion such as the sudden stop and the sudden turn of the UE, the acceleration of the UE can be obtained through the measurement device for measuring the acceleration.

[0135] Alternatively, the UE calculates the satellite elevation angle of the satellite relative to the UE at the current time through a geometric formula based on the ephemeris information issued by the network device through the system message (such as SIB19) and the position information of the UE itself. The specific calculation process can refer to the description in the related art, and for the sake of brevity, it will not be expanded here.

[0136] Alternatively, the UE performs calculation based on the beam coverage information in the satellite ephemeris information and the position information of the UE itself, and can obtain the included angle between the UE and the beam center.

[0137] In order to input the above input information into the first AI model, the UE can first construct a corresponding feature vector based on the input information, and the feature vector can be in the form of a time sequence. In some embodiments, the UE constructs an input feature vector (i.e., input information) and inputs the input feature vector into the first AI model, and the input feature vector at least includes a first feature sequence. The first feature sequence includes one or more of the following sequences: a propagation delay observation value sequence, a Doppler shift sequence, and a Doppler rate of change sequence. For example, the first feature sequence can also be referred to as a main input feature sequence; the first feature sequence is represented as .

[0138] The propagation delay observation value sequence is measured by the UE. Alternatively, the physical layer of the UE obtains an instantaneous delay measurement value by measuring the time difference of arrival (Time of Arrival, TOA) of the downlink synchronization signal issued by the network, such as SSB or CSI-RS. For example, the propagation delay observation value sequence is used to represent the delay of the UE at the current time and the previous N times, that is, a sequence of N+1 discrete values, each discrete value represents the propagation delay at the corresponding time; the propagation delay observation value sequence is represented as .

[0139] The Doppler shift sequence is calculated by the UE through the frequency of the received signal and the local frequency. For example, when the automatic frequency control (automatic frequency control, AFC) module of the UE locks the downlink carrier frequency, it will calculate the frequency deviation between the local oscillator and the received signal, and the deviation is the Doppler shift measurement value.

[0140] For example, a Doppler frequency shift sequence is used to characterize the Doppler frequency shift of a UE at the current time and the previous N time steps, i.e., a sequence of N+1 discrete values, where each discrete value represents the Doppler frequency shift at the corresponding time step; the Doppler frequency shift sequence is represented as follows: .

[0141] The Doppler rate of change sequence is calculated by the UE based on the Doppler frequency shift sequence. Optionally, the UE obtains the Doppler rate of change sequence by performing a first-order difference operation (or differentiation) on the Doppler frequency shift sequence.

[0142] For example, the UE can process the current time t and the previous N time steps. The measured values ​​are differentiated to obtain the Doppler rate of change sequence for the current time t and the previous N times; the Doppler rate of change sequence is expressed as... .

[0143] Optionally, the input feature vector may also include a second feature sequence and / or a third feature sequence. The second feature sequence is a feature sequence related to the motion of the UE. The third feature sequence is a feature sequence related to spatial angles (or spatial calibration).

[0144] The second feature sequence can also be described as a kinematic sequence or a physically driven feature. For example, the second feature sequence is represented as... In some embodiments, the second feature sequence includes a velocity vector sequence and / or an acceleration vector sequence.

[0145] The velocity vector sequence is used to characterize the UE's movement speed (e.g., ground movement speed) at the current time and N time steps prior. It consists of a sequence of N+1 discrete values, each representing the movement speed at the corresponding time step. For example, suppose... t The velocity vector sequence of the UE at the current time and the previous N time steps is represented as follows: .

[0146] An acceleration vector sequence is used to characterize the acceleration of the UE at the current time and the previous N time steps; that is, a sequence of N+1 discrete values, where each discrete value represents the acceleration at the corresponding time step. For example, suppose... t The acceleration vector sequence of the UE at the current time and the previous N time steps is represented as follows: .

[0147] The third feature sequence can also be described as a geometric sequence. For example, the third feature sequence is represented as... The third feature sequence is calculated by the UE based on the ephemeris information sent by the network device and its own location information. In some embodiments, the third feature sequence includes an included angle sequence and / or a satellite elevation angle sequence.

[0148] The angle sequence is used to characterize the angle between the UE and the center of the satellite beam at the current time and the previous N time points. It consists of a sequence of N+1 discrete values, where each discrete value represents the angle at the corresponding time point. For example, suppose... t The angle sequence between the UE at the current time and the previous N time points is represented as follows: .

[0149] A satellite elevation angle sequence is used to characterize the satellite elevation angle at the current time and the previous N time points; that is, a sequence of N+1 discrete values, where each discrete value represents the elevation angle at the corresponding time. For example, suppose... t The sequence of elevation angles of the UE at the current moment and the previous N moments is represented as follows: .

[0150] The following describes the output of the first AI model after the aforementioned input feature vector is input into it.

[0151] For example, for the first AI model, the input sequence at time t It can be represented as: At time t, the first AI model outputs the future value of the [missing value]. Time (of which) The prediction results for ( ). The time prediction results include one or more of the following: the predicted value of time delay, the predicted value of Doppler frequency shift, and the predicted trend (or the predicted value of Doppler rate of change).

[0152] For example, the first Time (of which) The prediction results include: (Indicates the predicted value of latency) (representing the predicted value of Doppler shift), and, (Used to predict inflection points).

[0153] Optionally, after the first AI model outputs the prediction result, an uncertainty parameter corresponding to each prediction item can be calculated separately using a sampling algorithm. The uncertainty parameter is used to quantify the uncertainty of the first AI model, or in other words, to quantify the accuracy of the prediction result output by the first AI model. This application embodiment does not specifically limit the type of sampling algorithm used. For example, the sampling algorithm may be a Monte Carlo (MC) sampling algorithm. In this application embodiment, when the UE uses the first AI model for inference, it can simulate Bayesianness by activating the dropout layer in the first AI model, thereby quantifying the model uncertainty (epistemic uncertainty) of the prediction result without introducing huge computational overhead.

[0154] For example, when the first AI model performs inference, all dropout layers (for reducing overfitting) in the first AI model are forced to be activated, and the dropout layers are set to training=True when the model is called in Keras. For the same input sequence , the forward propagation is repeated T times (e.g., T=50). Due to the randomness of the dropout layer, each propagation will obtain a set of prediction vectors that are different from each other . From the prediction vectors that are different from each other, the uncertainty can be calculated. For each prediction result, the uncertainty parameter corresponding to the prediction result can be calculated, which is described as follows: (1) For the predicted value of the time delay, the final prediction result is the average of T prediction results ,

[0155] Correspondingly, the uncertainty parameter related to the predicted time delay is the standard deviation of T prediction results . For example, satisfies the formula: ; where is the final prediction result of the time delay. quantifies the accuracy of the time delay output by the first AI model.

[0156] (2) For the predicted value of the Doppler shift, the final prediction result is the average of T prediction results ,

[0157] Correspondingly, the uncertainty parameter related to the predicted Doppler shift is the standard deviation of T prediction results . For example, satisfies the formula: ; where is the final prediction result of the Doppler shift. quantifies the accuracy of the Doppler shift output by the first AI model.

[0158] (3) For the predicted value of the Doppler rate (or inflection point or trend prediction value), the final prediction result is the average of T prediction results ,

[0159] Correspondingly, the uncertainty parameter related to the predicted inflection point is the standard deviation of T prediction results . For example, satisfies the formula: ; wherein, is the final prediction result of the aforementioned inflection point. quantifies the accuracy of the inflection point output by the first AI model.

[0160] As can be seen from the above example, the UE can obtain the time delay, Doppler shift and Doppler rate in advance based on the first AI model, so as to take a more robust measurement strategy in advance, which helps to reduce the possibility of link interruption.

[0161] It should be understood that the above example is only for understanding, and the embodiments of the present application are not limited thereto.

[0162] At step 330, the UE performs pre-compensation processing based on the prediction result to obtain a compensated transmission time and / or a compensated transmission frequency.

[0163] In order to realize uplink pre-compensation (Uplink Pre-compensation), the UE can replace the time delay calculated by the aforementioned UE using a geometric algorithm with the time delay predicted by the first AI model, and adjust the uplink transmission time based on the replaced time delay; and / or replace the Doppler shift calculated by the aforementioned UE using a geometric algorithm with the Doppler shift predicted by the first AI model, and adjust the uplink transmission carrier frequency based on the replaced Doppler shift, to meet the synchronization requirements of NTN transmission in 3GPP.

[0164] Compared with the situation of passive response when the rate of change occurs using a geometric algorithm, by introducing the first AI model to predict the time delay and / or the Doppler shift, and by considering historical data (such as a historical Doppler shift sequence), the first AI model can predict the time when the rate of change occurs in advance for a period of time (such as several seconds or even longer). Moreover, the first AI model can learn deeply the nonlinear transformation caused by complex factors such as satellite micro-motion, attitude adjustment or non-uniform motion of the UE, so as to realize more accurate and forward-looking channel management.

[0165] It should be noted that the prediction result output by the first AI model can include both the time delay and the Doppler shift; or include the time delay but not the Doppler shift; or include the Doppler shift but not the time delay. For the case where the first AI model does not include the time delay or the Doppler shift, the UE can calculate the value not predicted by the first AI model using a geometric calculation method.

[0166] Exemplarily, the prediction result of the first AI model includes the time delay and the Doppler shift; and the UE performs pre-compensation processing based on the time delay and the Doppler shift output by the first AI model to obtain a compensated transmission time and a compensated transmission frequency.

[0167] For example, the prediction result of the first AI model includes latency; the UE performs pre-compensation processing based on the latency output by the first AI model to obtain the compensated transmission time; and performs pre-compensation processing in combination with the Doppler frequency shift obtained by geometric calculation method to obtain the compensated transmission frequency.

[0168] For example, the prediction result of the first AI model includes Doppler frequency shift; the UE performs pre-compensation processing based on the Doppler frequency shift output by the first AI model to obtain the compensated transmission frequency; and performs pre-compensation processing in conjunction with the time delay obtained by the geometric calculation method to obtain the compensated transmission time.

[0169] The UE obtains the predicted latency value through the first AI model. In such cases, predicted values ​​can be used. Calculate transmission latency. For example, based on the prediction of the first AI model. It can be defined as one-way delay or RTT; if If it is a one-way delay, then utilize Replace the previous text In the formula Or, if If it is RTT, then utilize Replace the previous text In the formula Optionally, if the training phase of the first AI model takes into account common latency. Factors that the UE can utilize Replace the previous text .

[0170] The UE obtains the Doppler frequency shift using the first AI model. In such cases, predicted values ​​can be used. Calculate the transmission frequency. For example, using... Replace the previous text

[0171] Step 340: The UE sends a signal (e.g., an uplink transmission signal) to the network device based on the compensated transmission time and / or the compensated transmission frequency. Correspondingly, the network device receives the signal sent by the UE based on the compensated transmission time and / or the compensated transmission frequency.

[0172] For example, when a UE transmits a signal based on the compensated transmission time, it can be understood as the UE transmitting the signal in advance. Time-based signal transmission; Based on predicted values The calculation method is described above and will not be elaborated here.

[0173] For example, the UE transmits the signal based on the compensated transmission frequency can be understood as: the UE adopts the transmission frequency transmitting the uplink signal; satisfying the formula: 。

[0174] The prediction result described above includes the time delay and / or the Doppler shift. Alternatively, the prediction result can also include the Doppler rate. For the case that the prediction result includes the Doppler rate, the UE can use the Doppler rate to perform linear extrapolation on the frequency compensation value. That is, the Doppler rate can be used for fine-grained interpolation of the gap.

[0175] It should be noted that since the inference of the first AI model has a certain time interval (for example, the inference is performed every T milliseconds), during the idle period between two inferences, the UE can use the as the slope, and combine to calculate the frequency compensation value at a future time. For example, the frequency compensation value at the first time satisfies the formula: , so as to perform pre-compensation processing by using the frequency compensation value. In this way, it can be ensured that the frequency compensation curve applied to the physical layer is continuous and smooth, and the jump error caused by the traditional step compensation is eliminated.

[0176] In the embodiments of the present application, the network device sends information related to the first AI model to the UE, so that the UE predicts the channel information (such as time delay and / or Doppler shift) at a future time based on the first AI model. Compared with the way in the prior art that the UE calculates the channel information (including time delay and / or Doppler shift) by geometric calculation, the first AI model in the embodiments of the present application considers historical measurement data (or historical regularity), and the UE predicts the channel information (including time delay and / or Doppler shift) at a future time by using the first AI model, which can improve the accuracy of the prediction result, thereby ensuring that the transmitted signal is synchronized in time and / or frequency when it reaches the satellite.

[0177] Optionally, the embodiments of the present application introduce a Bayesian Neural Network (BNN) model as the first AI model. The long short-term memory (LSTM) is a kind of time recurrent neural network. By using LSTM as the first AI model, not only can the non-linear law be captured, such as learning and compensating the non-linear delay caused by factors such as the atmosphere, which cannot be covered by the traditional model; but also the short-term robustness can be realized, such as when the GNSS signal is temporarily lost, the first AI model can still rely on the historical "inertia" pattern learned to continue to provide effective short-term prediction within a certain time.

[0178] Optionally, the first AI model is a Monte Carlo (MC) based approximate Bayesian model, such as a Bayesian based B-LSTM. The principle of B-LSTM includes: for the standard neural network using Dropout during training, if the Dropout layer is still forced to be activated during inference, then the network is equivalent to an approximate Bayesian neural network.

[0179] Exemplarily, constructing the first AI model includes: the feature extraction layer uses the LSTM layer; and in order to realize the Bayesian approximation, the built-in dropout parameter and recurrent_dropout parameter in the LSTM layer are enabled, and the Dropout layer is inserted between the layers.

[0180] The network structure of the first AI model is described below. Figure 4 As shown in Figure 4 , the first AI model includes an input layer, a feature extraction layer, and a multi-head output layer. Among them, the Dropout layer (or called random dropout layer) is inserted in the feature extraction layer. The input layer is connected with the input end of the feature extraction layer; the output end of the feature extraction layer is connected with the multi-head output layer.

[0181] The input layer is the first layer of the first AI model. The input information can be transmitted to the first AI model through the input layer. The input information is the historical measurement data of the previous N moments of the current moment determined (such as calculated or measured) by the UE. For example, for the way of UE reporting historical data measurement, the input layer is used to receive the multi-dimensional time series data processed by the UE.

[0182] Exemplarily, the data stream received by the input layer is the input tensor , The data dimension of . Among them, represents the batch size, which is a core parameter in deep learning training that determines the number of input samples at a time; Time Steps representing historical observations; representing time The total number of features (including joint features such as time delay, Doppler shift, Doppler rate, velocity, acceleration, angle (such as included angle, satellite elevation angle), etc.).

[0183] The feature extraction layer is used to capture the long and short term dependencies of the sequence input by the input layer. In addition, Bayesian characteristics are introduced in the feature extraction layer. For example, the feature extraction layer can adopt Bayesian recurrent feature extraction layer (Bayesian Recurrent Layers). The feature extraction layer is composed of two stacked LSTM units.

[0184] As shown in Figure 4 , the feature extraction layer includes LSTM layer 1 and LSTM layer 2. A Dropout layer is included between LSTM layer 1 and LSTM layer 2. The input end of LSTM layer 1 is connected to the input layer, and the output end is connected to the intermediate Dropout layer. The input end of LSTM layer 2 is connected to the Dropout layer, and the output end is connected to the multi-head output layer.

[0185] Among them, LSTM layer 1 is used to preliminarily encode the input sequence and output a complete hidden state sequence with the same time step as the input. The dropout and recurrent_dropout are turned on in LSTM layer 1 to introduce randomness. For example, the configuration of LSTM layer 1 is set as follows: set the parameters .

[0186] The Dropout layer is used to further randomly discard part of the neuron connections, prevent overfitting, and enhance the random variance in Monte Carlo sampling.

[0187] LSTM layer 2 is used to receive sequence data from the Dropout layer and compress and extract the sequence data, outputting the hidden state vector of the last time step. The vector includes all the space-time features of the past time. For example, the configuration of LSTM layer 2 is set as follows: .

[0188] In order to meet the needs of numerical prediction and trend prediction at the same time, the output layer of the model adopts a "multi-task branch" structure. The multi-head output layer includes fully connected output layer 1 and fully connected output layer 2. Fully connected output layer 1 and fully connected output layer 2 are two parallel output layers. The hidden state vector is simultaneously sent to two parallel fully connected layers (Fully Connected Layers).

[0189] The input of the fully connected output layer 1 is the hidden state vector output by the LSTM layer 2 , and the output is the predicted mean vector . That is, the fully connected output layer 1 is used to fit the size of the predicted value.

[0190] The input of the fully connected output layer 2 is also the hidden state vector output by the LSTM layer 2 , and the output is the trend prediction value . That is, the fully connected output layer 2 is used to fit the rate of change of the Doppler, thereby assisting in determining whether there is a nonlinear inflection point.

[0191] In addition, after the UE obtains the prediction result based on the first AI model, it can also calculate the uncertainty parameter . In the model inference (Inference) stage, all Dropout modules in the network structure of the first AI model (including the Dropout layers inside the LSTM and between the LSTM layers) are kept in an open state (Training=True). By performing forward propagation times on the same input , the randomness of each layer Dropout is used to generate a set of output results that exist differences, thereby calculating the predicted uncertainty parameter .

[0192] The optimizer used in the training process of the first AI model in the embodiments of the present application is not specifically limited, for example, the optimizer is an Adam optimizer. The Adam optimizer is a widely used adaptive optimization algorithm in deep learning. The principle or explanation of the Adam optimizer can be referred to the description in the related art, and for the sake of brevity, it will not be expanded here.

[0193] Optionally, a loss function is introduced to determine whether the training process of the first AI model is completed, or in other words, whether the model is converged. The loss function can adopt a mean squared error (MSE).

[0194] Exemplarily, the training target of the first AI model is to minimize the following prediction loss function:

[0195] wherein, represents the loss value, represents the training sample (or true value) at the i-th moment, represents the predicted value (or output value) of the first AI model in the training process. It can be understood that for the fully connected output layer 1 and the fully connected output layer 2, the error between the predicted value and the true value can be calculated respectively.

[0196] Optionally, the determination condition of the end of the model training can adopt an "early stopping" mechanism in combination with the loss convergence criterion, where the early stopping mechanism is a regularization strategy for preventing overfitting of the model; used to monitor a certain index during the model training process, such as the loss or accuracy of the validation set; if the index does not improve for consecutive multiple training rounds, the training is terminated in advance.

[0197] In some embodiments, the network device calculates the loss value of the validation set during the training process, such as calculating the loss value of the validation set every training cycle; when the loss value of the validation set does not decrease significantly (for example, the decrease is less than a threshold δ, it is considered that there is no significant decrease), it is determined that the training model reaches the convergence state, and the training process is stopped. When there is no significant decrease (for example, the decrease is less than a threshold δ, it is considered that there is no significant decrease) in consecutive P training cycles (epochs), it is determined that the training model reaches the convergence state, and the training process is stopped. After the first AI model training is completed, the network device saves the current optimal weight parameter as a lightweight model file, so as to perform the compression and delivery process subsequently.

[0198] It can be understood that the above implementation manner of determining the end of the model training is only an exemplary description, and those skilled in the art can also use other manners in the model training or deep learning process to determine whether to end the model training process, and the embodiments of the present application do not make specific limitations. Optionally, after obtaining the prediction result of the first AI model, the UE can also determine the performance parameter to decide whether to use the prediction result (or output result) of the first AI model. The following is described by taking the performance parameter as an uncertainty parameter.

[0199] Optionally, the method further includes: determining, by the UE, a length of an SSB measurement timing configuration (SMTC) window based on the first uncertainty parameter, the first uncertainty parameter being an uncertainty parameter related to the time delay prediction value, and the first uncertainty parameter being used to represent the accuracy of the time delay prediction value output by the first AI model.

[0200] The SMTC window is introduced to avoid the blindness of terminal device measurement. The UE can perform cell switching, cell reselection or cell selection by measuring a reference signal (such as a synchronization signal or a physical broadcast channel block (SSB)). The network device can send SMTC configuration information to the UE; the UE can determine the SMTC window based on the SMTC configuration information; the UE only needs to measure the reference signal within the SMTC window, thereby avoiding the blindness of measurement and helping to reduce the measurement power consumption of the UE. In the embodiments of the present application, the UE can dynamically adjust the window length of the SMTC based on the first uncertainty parameter, thereby further saving the power consumption of the UE.

[0201] For example, the first parameter is represented as The specific calculation method of the first parameter can refer to the foregoing description, and will not be described here again.

[0202] In some embodiments, the UE determines the duration of the SMTC window based on the first uncertainty parameter, including: when the first uncertainty parameter does not exceed a preset threshold, reducing the duration (i.e., the window duration) of the SMTC window; when the first uncertainty parameter exceeds the preset threshold, maintaining the duration of the SMTC window. Maintaining the duration of the SMTC window can be understood as using the standard window length (i.e., using the SMTC window length configured by the network device), without changing the duration of the SMTC window. When the first uncertainty parameter does not exceed the preset threshold, the UE shortens the duration of the SMTC window, that is, a shorter and more accurate measurement window can be used, which helps to reduce the radio frequency energy consumption.

[0203] The preset threshold can be configured by the network device to the UE, or can be predefined by the protocol, and is not limited in this regard.

[0204] It can be understood that the prediction result output by the first AI model can be accurate or inaccurate. For the UE, the UE can continuously compare the difference between the prediction result output by the first AI model and the actual channel measurement value (or true measurement value) to evaluate the performance of the first AI model; or the UE can also evaluate the performance of the first AI model through the uncertainty parameter corresponding to the prediction result. The UE can decide whether to continue to use the first AI model based on the performance evaluation result of the first AI model; or use the first AI model to predict the time delay and / or Doppler shift at the future time, or use the traditional geometric algorithm to calculate the time delay and / or Doppler shift at the future time. The following will be described in detail in combination with different implementation manners.

[0205] ​Optionally, the method further comprises: in step 321, determining, by the UE, whether the performance of the first AI model meets a first preset condition according to the prediction result and an actual measurement value of the channel.

[0206] In a case where it is determined that the performance of the first AI model does not meet the first preset condition, the UE calculates channel information (delay and / or Doppler shift) based on a geometric algorithm and performs pre-compensation processing; In a case where it is determined that the performance of the first AI model meets the first preset condition, the UE performs pre-compensation processing using the prediction result output by the first AI model.

[0207] That is, in a case where the performance of the first AI model does not meet the first preset condition, the UE can trigger a fallback mechanism, i.e., switching to a traditional geometric calculation mode. Optionally, the UE can also expand the SMTC window to ensure the stability of the connection. Of course, if the performance of the first AI model meets the first preset condition, the UE adopts the prediction result output by the first AI model.

[0208] As to whether the performance of the first AI model meets the first preset condition, different implementation manners can be adopted. The following describes modes one and two in detail.

[0209] Mode one, the performance of the first AI model does not meet the first preset condition, including: a prediction error value exceeds an error threshold (or a performance threshold), and a duration for which the prediction error value exceeds the error threshold exceeds a first duration threshold. Correspondingly, the performance of the first AI model meets the first preset condition, including: the prediction error value exceeds the error threshold, and a duration for which the prediction error value exceeds the error threshold does not exceed the first duration threshold; or the prediction error value does not exceed the error threshold.

[0210] The prediction error value is determined based on a sequence predicted by the first AI model (which can be referred to as a predicted sequence) and a sequence actually measured by the UE. The prediction error value can be an average absolute error or a root mean square error or other indicators for characterizing the prediction error in a sliding time window.

[0211] Exemplarily, the size of the sliding time window is N, the sequence actually measured by the UE is: , and the sequence predicted by the first AI model is: In the sliding time window, the average absolute error calculated by the UE satisfies the formula: .

[0212] Exemplarily, the size of the sliding time window is N, the sequence actually measured by the UE is: , and the sequence predicted by the first AI model is: In the sliding time window, the root mean square error calculated by the UE satisfies the formula: .

[0213] It should be noted that the actually measured sequence and the predicted sequence in the above examples can be a sequence for a certain feature vector, for example, an actually measured sequence for a time delay and a predicted sequence for a time delay; for example, an actually measured sequence for a Doppler shift and a predicted sequence for a Doppler shift; for example, an actually measured sequence for a Doppler rate and a predicted sequence for a Doppler rate.

[0214] The error threshold and / or the first duration threshold can be configured by the network device (for example, the network device can deliver the error threshold and / or the first duration threshold together when delivering information related to the AI model), or can be predefined by a protocol, which is not limited specifically.

[0215] It can be understood that if the prediction result output by the first AI model includes multiple categories, an error threshold can be set for the feature vector of each category of prediction result, and the corresponding error threshold can be set to be the same or different, which is not limited specifically.

[0216] Taking the prediction result output by the first AI model as an example, which includes a time delay, a Doppler shift and a Doppler rate, and the performance of the first AI model does not meet the first preset condition, including one or more of the following: The prediction error value of the time delay exceeds the first error threshold, and the duration that the prediction error value of the time delay exceeds the first error threshold exceeds the first duration threshold; The prediction error value of the Doppler shift exceeds the second error threshold, and the duration that the prediction error value of the Doppler shift exceeds the second error threshold exceeds the first duration threshold; The prediction error value of the Doppler rate exceeds the third error threshold, and the duration that the prediction error value of the Doppler rate exceeds the third error threshold exceeds the first duration threshold.

[0217] Correspondingly, the performance of the first AI model meets the first preset condition, including the following (1) to (3): (1) The prediction error value of the time delay exceeds the first error threshold, and the duration that the prediction error value of the time delay exceeds the first error threshold does not exceed the first duration threshold; or, the prediction error value of the time delay does not exceed the first error threshold; (2) The prediction error value of the Doppler shift exceeds the second error threshold, and the duration that the prediction error value of the Doppler shift exceeds the second error threshold does not exceed the first duration threshold; or, the prediction error value of the Doppler shift does not exceed the second error threshold; (3) the prediction error value of the Doppler rate exceeds the third error threshold, and the duration for which the prediction error value of the Doppler rate exceeds the third error threshold does not exceed the first duration threshold; or the prediction error value of the Doppler rate does not exceed the third error threshold.

[0218] It should be understood that the above is described by taking the prediction result output by the first AI model as an example, which includes the time delay, the Doppler shift, and the Doppler rate. The embodiments of the present application are not limited thereto. For example, if the prediction result output by the first AI model includes one or more of the time delay, the Doppler shift, and the Doppler rate, the related implementation mode for evaluating the performance of the first AI model can also refer to the above examples.

[0219] Mode two, the performance of the first AI model does not satisfy the first preset condition, including one or more of the following: the first uncertainty parameter exceeds the first confidence threshold for M consecutive times; the second uncertainty parameter exceeds the second confidence threshold for W consecutive times; the third uncertainty parameter exceeds the third confidence threshold for Q consecutive times.

[0220] The first uncertainty parameter is a parameter related to the time delay output by the first AI model, for example, the first uncertainty parameter is represented as The second uncertainty parameter is a parameter related to the Doppler shift output by the first AI model, for example, the second uncertainty parameter is represented as The third uncertainty parameter is a parameter related to the Doppler rate output by the first AI model, for example, the third uncertainty parameter is represented as The description of how to calculate the uncertainty parameter can refer to the formula in the foregoing description, which will not be expanded here. M, W, and Q are integers; the values can be the same or different, and no specific limitation is made thereto.

[0221] Correspondingly, the performance of the first AI model satisfies the first preset condition, including: the first uncertainty parameter does not exceed the first confidence threshold for M consecutive times, the second uncertainty parameter does not exceed the second confidence threshold for W consecutive times; and the third uncertainty parameter does not exceed the third confidence threshold for Q consecutive times.

[0222] That is, the UE can determine the performance of the first AI model based on the calculated uncertainty parameter; that is, the first preset condition can be set based on the uncertainty parameter.

[0223] Here, it should be uniformly stated that the "exceeds" mentioned in the embodiments of the present application can be understood as "greater than or equal to", which is represented by the corresponding mathematical symbol ; it can also be understood as "greater than", which is represented by the corresponding mathematical symbol "not less than". That is, the embodiments of the present application do not specifically limit whether the boundary value belongs to the range of "not less than" in the case of "equal to".

[0224] Correspondingly, the "not more than" mentioned in the embodiments of the present application can be understood as "less than or equal to", represented by the corresponding mathematical symbol ". It can also be understood as "less than", represented by the corresponding mathematical symbol ". It can also be understood as "not greater than". That is, the embodiments of the present application do not specifically limit whether the boundary value belongs to the range of "not more than" in the case of "equal to".

[0225] It can be understood that, in combination with the above description of "more than" and "not more than", when "more than" is understood as "greater than or equal to", "not more than" is understood as "less than"; when "more than" is understood as "greater than", "not more than" is understood as "less than or equal to".

[0226] Regardless of whether the foregoing way one or way two is adopted, if the UE judges that the performance of the first AI model does not satisfy the first preset condition, the UE can switch to a traditional geometric algorithm to calculate the delay and / or Doppler shift to ensure link reliability; and the UE can also feed back the error to the network device. That is, the present application also provides an error feedback mechanism, so that the network device can learn in time when the performance of the first AI model deteriorates.

[0227] Optionally, in the case where the performance of the first AI model does not satisfy the first preset condition, the method further includes: step 322, the UE sends feedback information to the network device, the feedback information being used to inform the network device that the performance of the first AI model does not satisfy the first preset condition. Correspondingly, the network device receives the feedback information.

[0228] The embodiments of the present application do not specifically limit the content included in the feedback information. For example, the UE can report the calculated error value to the network device through the feedback information. For another example, the UE only reports the feedback information to inform the network device that the performance of the first AI model does not meet the requirements.

[0229] For the network device, after receiving the feedback information of the UE, the network device can decide whether to optimize the first AI model based on the actual situation, for example, whether to adjust the parameters of the first AI model, so as to improve the prediction accuracy of the first AI model. If the network device optimizes the first AI model, the network device can send the model file of the optimized first AI model (such as the model parameters of the optimized first AI model) to the UE again, so that the UE performs prediction based on the optimized first AI model.

[0230] As mentioned above, if the UE judges that the first AI model does not satisfy the first preset condition, the UE can switch (or fallback) to calculating the channel information based on the geometric algorithm and performing the pre-compensation processing, that is, switch to the traditional calculation method which does not rely on the first AI model. However, the UE can continue to run the first AI model in the background and verify the prediction result output by the first AI model using the new actual channel measurement value (i.e. the real channel measurement value measured by the UE after switching to calculating the channel information based on the geometric algorithm) to decide whether the prediction result of the first AI model can be continued to be used.

[0231] Optionally, in the case where the UE calculates the channel information based on the geometric algorithm and performs the pre-compensation processing, the method further comprises: the UE determines that the first AI model satisfies a second preset condition, and again performs the pre-compensation processing using the output result of the first AI model. That is, if the UE judges that the first AI model has recovered to be stable, the UE can reuse the prediction result of the first AI model. Of course, if the UE judges that the first AI model still does not satisfy the second preset condition, the UE can continue to calculate the channel information based on the geometric algorithm and perform the pre-compensation processing to ensure the link reliability.

[0232] In some embodiments, the first AI model satisfying the second preset condition comprises one or more of the following: (1) a prediction error value of the delay does not exceed a fourth error threshold value, and a duration that the prediction error value of the delay does not exceed the fourth error threshold value exceeds a second duration threshold value; (2) a prediction error value of the Doppler shift does not exceed a fifth error threshold value, and a duration that the prediction error value of the Doppler shift does not exceed the fifth error threshold value exceeds the second duration threshold value; (3) a prediction error value of the Doppler rate does not exceed a sixth error threshold value, and a duration that the prediction error value of the Doppler rate does not exceed the sixth error threshold value exceeds the second duration threshold value.

[0233] The one or more threshold values involved in the above-mentioned second preset condition can be configured by the network device or predefined by the protocol, and no specific limitation is made thereto.

[0234] Therefore, by introducing the second preset condition, the UE can again use the prediction result of the first AI model at an appropriate time (i.e. the performance of the first AI model has recovered), thereby improving the accuracy of the prediction result. In addition, by introducing the above-mentioned first preset condition and the second preset condition, it is helpful to balance the relationship between the link reliability and the energy efficiency.

[0235] It should also be understood that, Figures 1 to 4 The flowcharts or scenario charts shown are only for ease of understanding and are not intended to limit the embodiments of the present application to the examples shown in the charts. In fact, those skilled in the art can make modifications to the embodiments of the present application based on the charts.Figures 1 to 4 The foregoing embodiments and / or examples of the present application are further

[0236] The communication method provided by the embodiments of the present application is described in detail above. Figures 1 to 4 The communication method provided by the embodiments of the present application is described in detail above. Figure 5 The apparatus embodiments of the present application are described in detail below. Figure 6 It should be understood that the communication apparatus of the embodiments of the present application can perform the various communication methods of the foregoing embodiments of the present application, i.e., the specific working processes of the following various products can refer to the corresponding processes in the foregoing method embodiments.

[0237] In the foregoing embodiments, the UE can perform some or all of the steps in the embodiments, and the network device can perform some or all of the steps in the embodiments. These steps or operations are merely examples, and the embodiments of the present application can also perform other operations or variations of the various operations. In addition, the various steps can be performed in different orders from the order presented in the embodiments, and it is possible that not all of the operations in the embodiments of the present application are performed. Moreover, the magnitude of the serial number of the steps does not mean the order of execution, and the execution order of the various processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0238] Figure 5 is a schematic block diagram of the communication apparatus provided by the embodiments of the present application. As shown in Figure 5 The communication apparatus 1500 can include a communication module 1520. The communication module 1520 can implement a corresponding communication function, which can be an internal communication function of the communication apparatus 1500, or a communication function of the communication apparatus 1500 and other apparatuses. Optionally, the communication module 1520 can also be referred to as a communication interface or a transceiver module. Optionally, the communication apparatus 1500 further includes a processing module 1510. The processing module 1510 can implement a corresponding processing function.

[0239] Optionally, the communication apparatus 1500 further includes a storage module, which can be used to store instructions and / or data; the processing module 1510 can read the instructions and / or data in the storage module, so that the communication apparatus 1500 implements the foregoing method embodiments.

[0240] In a possible design, the communication apparatus 1500 can correspond to the UE in the foregoing method embodiments, or a component (such as a circuit, a chip or a chip system, etc.) configured in the UE. The communication apparatus 1500 can be used to perform the steps or processes performed by the UE in any of the foregoing method embodiments.

[0241] In a possible design, the communication module 1520 is configured to receive a first message from a network device, where the first message comprises information related to a first AI model, and the information related to the first AI model comprises one or more of the following: a network structure of the first AI model, a layer connection relationship of the first AI model, and a model parameter of the first AI model; the first AI model is used to output predicted information at a time t+k according to input information at the time t, where ; The processing module 1510 is configured to determine a prediction result according to the input information and the first AI model, where the prediction result comprises one or more of the following predicted values: a time delay and a Doppler shift. The processing module 1510 is further configured to perform pre-compensation processing based on the prediction result, to obtain a compensated transmission time and / or a compensated transmission frequency. The processing module 1510 is further configured to send a signal to the network device based on the compensated transmission time and / or the compensated transmission frequency.

[0242] Optionally, as an embodiment, the prediction result further comprises a predicted value of a Doppler rate of change.

[0243] Optionally, as an embodiment, the processing module 1510 is configured to determine the prediction result according to the input information and the first AI model, including: constructing an input feature vector, and inputting the input feature vector to the first AI model, where the input feature vector comprises at least a first feature sequence, and the first feature sequence comprises one or more of the following sequences: a sequence of observed time delay values, a sequence of Doppler shift values, and a sequence of Doppler rates of change; and obtaining the prediction result output by the first AI model.

[0244] Optionally, as an embodiment, the input feature vector further comprises a second feature sequence and / or a third feature sequence; the second feature sequence is a feature sequence related to a motion of a UE; and the third feature sequence is a feature sequence related to a spatial angle.

[0245] Optionally, as an embodiment, the second feature sequence comprises a sequence of velocity vectors and / or a sequence of acceleration vectors.

[0246] Optionally, as an embodiment, the third feature sequence comprises a sequence of included angles and / or a sequence of satellite elevation angles.

[0247] Optionally, as an embodiment, the processing module 1510 is further configured to determine whether the performance of the first AI model meets a first preset condition according to the prediction result and an actual measurement value of the channel; in a case where it is determined that the performance of the first AI model does not meet the first preset condition, perform pre-compensation processing based on a geometric algorithm to calculate channel information; in a case where it is determined that the performance of the first AI model meets the first preset condition, perform pre-compensation processing using the prediction result output by the first AI model.

[0248] Optionally, as an embodiment, the performance of the first AI model not meeting the first preset condition comprises: a prediction error value exceeding an error threshold value, and a duration for which the prediction error value exceeds the error threshold value exceeding a first duration threshold value, wherein the prediction error value is determined based on a sequence predicted by the first AI model and a sequence actually measured by the UE; wherein the performance of the first AI model meeting the first preset condition comprises: the prediction error value exceeding the error threshold value, and a duration for which the prediction error value exceeds the error threshold value not exceeding the first duration threshold value; or the prediction error value not exceeding the error threshold value.

[0249] Optionally, as an embodiment, the performance of the first AI model not meeting the first preset condition comprises one or more of the following: a first uncertainty parameter related to a time delay output by the first AI model exceeding a first confidence threshold value for M consecutive times; a second uncertainty parameter related to a Doppler shift output by the first AI model exceeding a second confidence threshold value for W consecutive times; a third uncertainty parameter related to a Doppler rate output by the first AI model exceeding a third confidence threshold value for Q consecutive times.

[0250] Optionally, as an embodiment, the communication module 1520 is further configured to, in a case where the performance of the first AI model does not meet the first preset condition, send feedback information to a network device, the feedback information being used to inform the network device that the performance of the first AI model does not meet the first preset condition.

[0251] Optionally, as an embodiment, the processing module 1510 is further configured to, in a case where the UE calculates channel information based on a geometric algorithm and performs pre-compensation processing, determine that the first AI model meets a second preset condition, and perform pre-compensation processing again using an output result of the first AI model.

[0252] Optionally, as an embodiment, the first AI model satisfying the second preset condition comprises one or more of the following: a prediction error value of the time delay does not exceed a fourth error threshold, and a duration for which the prediction error value of the time delay does not exceed the fourth error threshold exceeds a second duration threshold; a prediction error value of the Doppler shift does not exceed a fifth error threshold, and a duration for which the prediction error value of the Doppler shift does not exceed the fifth error threshold exceeds the second duration threshold; a prediction error value of the Doppler rate does not exceed a sixth error threshold, and a duration for which the prediction error value of the Doppler rate does not exceed the sixth error threshold exceeds the second duration threshold.

[0253] Optionally, as an embodiment, the processing module 1510 is further configured to, in a case where the result output by the first AI model comprises a predicted value of the time delay, determine a first uncertainty parameter based on the predicted value of the time delay output by the first AI model; the first uncertainty parameter is an uncertainty parameter related to the predicted value of the time delay, and is used to represent the accuracy of the predicted value of the time delay output by the first AI model; and determine the duration of the SSB measurement time configuration (SMTC) window based on the first uncertainty parameter.

[0254] Optionally, as an embodiment, the processing module 1510 is configured to determine the duration of the SMTC window based on the first uncertainty parameter, comprising: when the first uncertainty parameter does not exceed a preset threshold, reducing the duration of the SMTC window; and when the first uncertainty parameter exceeds the preset threshold, maintaining the duration of the SMTC window.

[0255] Optionally, as an embodiment, the communication module 1520 is further configured to report historical data to a network device; the historical data is used to train the first AI model; and the historical data comprises one or more of the following information: channel state information, kinematic information, and spatial angle information; wherein the channel state information comprises one or more of the following: reference signal received power (RSRP), measured time delay, and Doppler shift; the kinematic information comprises velocity and / or acceleration; and the spatial angle information comprises a satellite elevation angle and / or a beam angle calculated based on a historical time.

[0256] The above is only an example, and detailed steps or processes can refer to the description of the foregoing embodiments.

[0257] Alternatively, in another possible design, the communication apparatus 1500 can correspond to the network device in the foregoing method embodiments, or be configured in a component (such as a circuit, a chip, or a chip system, etc.) in the network device. The communication apparatus 1500 can be configured to perform the steps or processes performed by the network device in any of the foregoing method embodiments.

[0258] Exemplarily, the processing module 1510 is configured to train a first AI model according to a training data set, the first AI model being configured to output prediction information at a time instant t+k based on input information at the time instant t, the prediction information comprising one or more of: a time delay, a Doppler shift, a Doppler rate, wherein ; The communication module 1520 is configured to send a first message, the first message comprising information related to the first AI model, the information related to the first AI model comprising one or more of: a network structure of the first AI model, a layer connection relationship of the first AI model, a model parameter of the first AI model.

[0259] Optionally, as an embodiment, the communication module 1520 is further configured to receive historical data from one or more UEs, the historical data being used for training the first AI model; the historical data comprising one or more of: channel state information, kinematic information, and spatial angle information; wherein the channel state information comprises one or more of: a reference signal received power (RSRP), a measurement time delay, a Doppler shift; the kinematic information comprises a velocity and / or an acceleration; the spatial angle information comprises a satellite elevation angle and / or a beam angle calculated based on a historical time instant; the processing module 1510 is further configured to construct the training data set according to the historical data.

[0260] Optionally, as an embodiment, the first AI model comprises an input layer, a feature extraction layer, and a multi-head output layer; wherein a Dropout layer is inserted in the feature extraction layer; the input layer is connected to an input end of the feature extraction layer; an output end of the feature extraction layer is connected to the multi-head output layer.

[0261] Optionally, as an embodiment, the first AI model is an LSTM model based on a Bayesian model; the feature extraction layer comprises an LSTM layer 1 and an LSTM layer 2; A Dropout layer is included between the LSTM layer 1 and the LSTM layer 2; an input end of the LSTM layer 1 is connected to the input layer, and an output end of the LSTM layer 1 is connected to the Dropout layer; an input end of the LSTM layer 2 is connected to the Dropout layer, and an output end of the LSTM layer 2 is connected to the multi-head output layer; The multi-head output layer comprises a fully connected output layer 1 and a fully connected output layer 2; the fully connected output layer 1 is configured to output a predicted value of the time delay and / or a predicted value of the Doppler shift; the fully connected output layer 2 is configured to output a predicted value of the Doppler rate.

[0262] Optionally, as an embodiment, the processing module 1510 is further configured to perform compression processing on the first AI model before sending the first message to the UE, to obtain information related to the first AI model.

[0263] Optionally, as an embodiment, the communication module 1520 is further configured to receive feedback information, which is used to notify the network device that the performance of the first AI model does not meet the first preset condition.

[0264] Optionally, as an embodiment, the performance of the first AI model does not meet the first preset condition, including: the prediction error value exceeds the error threshold, and the duration of the prediction error value exceeding the error threshold exceeds the first duration threshold, wherein the prediction error value is determined based on the sequence predicted by the first AI model and the sequence actually measured by the UE.

[0265] Optionally, as an embodiment, the performance of the first AI model does not meet the first preset condition, including one or more of the following: the first uncertainty parameter exceeds the first confidence threshold for M consecutive times, where the first uncertainty parameter is a parameter related to the time delay output by the first AI model; the second uncertainty parameter exceeds the second confidence threshold for W consecutive times, where the second uncertainty parameter is a parameter related to the Doppler frequency shift output by the first AI model; and the third uncertainty parameter exceeds the third confidence threshold for Q consecutive times, where the third uncertainty parameter is a parameter related to the Doppler rate of change output by the first AI model.

[0266] The above are merely examples; for detailed steps or procedures, please refer to the descriptions in the foregoing embodiments.

[0267] Figure 6 This is another schematic block diagram of the communication device 1600 provided in the embodiments of this application. The communication device 1600 may be a chip, chip system, or processor, etc., in a terminal device (such as a UE) or network device that implements the above methods. The communication device 1600 can be used to implement the methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0268] like Figure 6 As shown, the communication device 1600 may include one or more processors 1610, which may also be referred to as processing units or processing modules, and can implement certain control functions. The processor 1610 may be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, while the central processing unit can be used to control the communication device 1600 (e.g., a base station, baseband chip, user, user chip), execute software programs, and process data from the software programs.

[0269] In an alternative design, the processor 1610 can also store instructions and / or data, which can be executed by the processor 1610, so that the communication apparatus 1600 performs the methods described in the above method embodiments.

[0270] In another alternative design, the communication apparatus 1600 can include a communication interface 1620 for implementing receiving and sending functions. For example, the communication interface 1620 can be a transceiver circuit, an interface, an interface circuit or a transceiver, etc. The transceiver circuit, the interface, the interface circuit or the transceiver for implementing receiving and sending functions can be separate or integrated together. The above transceiver circuit, the interface, the interface circuit or the transceiver can be used for code / data reading and writing, or the above transceiver circuit, the interface, the interface circuit or the transceiver can be used for signal transmission or transfer.

[0271] Optionally, the communication apparatus 1600 can include one or more memories 1630, which can store instructions executable by the processor 1610, so that the communication apparatus 1600 performs the methods described in the above method embodiments. Optionally, the memory 1630 can also store data. Optionally, the processor 1610 can also store instructions and / or data. The processor 1610 and the memory 1630 can be separately arranged or integrated together.

[0272] It should be understood that, in a possible design, the steps in the method embodiments provided in the present application can be completed by integrated logic circuits of hardware in the processor or instructions in the form of software. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being completed by a hardware processor, or completed by a combination of hardware and software modules in the processor. The software modules can be located in random access memories, flash memories, read-only memories, programmable read-only memories or electrically erasable programmable memories, registers or other mature storage media in the art. The storage medium is located in the memory, and the processor reads information in the memory and combines the hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0273] In an implementation, the communication apparatus 1600 can correspond to the terminal device (such as a UE) in the above method embodiments, and can be used to execute the steps and / or processes performed by the terminal device (such as a UE) in the above method embodiments. The processor 1610 can be used to execute the instructions stored in the memory 1630, and when the processor 1610 executes the instructions stored in the memory, the processor 1610 is used to execute the steps and / or processes of the above method embodiments corresponding to the terminal device.

[0274] In another implementation, the communication apparatus 1600 can correspond to a network device (such as a satellite, or a base station) in the above method embodiments, and can be configured to execute various steps and / or procedures performed by the network device in the above method embodiments. The processor 1610 can be configured to execute instructions stored in the memory 1630, and when the processor 1610 executes the instructions stored in the memory, the processor 1610 is configured to execute various steps and / or procedures of the above method embodiments corresponding to the network device.

[0275] It should be understood that the above processing apparatus can be one or more chips. For example, the processing apparatus can be a field programmable gate array (FPGA), can be an application specific integrated circuit (ASIC), can be a system on chip (SoC), can be a central processor unit (CPU), can be a network processor (NP), can be a digital signal processor (DSP), can be a micro controller unit (MCU), can be a programmable logic device (PLD), or other integrated chip.

[0276] It is to be appreciated that the memory in the embodiments of the application can be volatile or nonvolatile memory, or can include both volatile and nonvolatile memory. In one example, a non-volatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory. A volatile memory can be random access memory (RAM), which acts as external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct Rambus RAM (DR RAM). It is to be appreciated that the memory described herein is intended to include, without being limited to, these and any other suitable types of memory.

[0277] According to the method provided in the embodiments of the application, the application further provides a chip system, which comprises one or more processors, and is used for calling and running instructions stored in a memory, so that the method provided in the embodiments of the application is executed. The chip system can be composed of a chip, or can comprise a chip and other discrete devices.

[0278] The chip system can comprise an input circuit or interface for sending information or data, and an output circuit or interface for receiving information or data.

[0279] According to the method provided in the embodiments of the application, the application further provides a communication system, which comprises the network device and the UE described above.

[0280] According to the method provided in the embodiments of the application, the application further provides a computer program product, which comprises computer program codes, and when the computer program codes run on a computer, the computer is caused to execute each step or flow of the network device and the UE in any of the method embodiments described above.

[0281] According to the method provided in the embodiments of the present application, the present application further provides a computer readable storage medium, which stores program codes, and when the program codes are run on a computer, the computer is caused to perform each step or flow of the network device and the UE according to any one of the method embodiments.

[0282] The computer readable storage medium can be the volatile memory or the non-volatile memory described above, or can include both the volatile memory and the non-volatile memory.

[0283] In the embodiments of the present application, each term and English abbreviation is an exemplary example given for convenience of description, and should not constitute any limitation on the present application. The present application does not exclude the possibility of defining other terms capable of achieving the same or similar functions in the existing or future protocols.

[0284] In the above embodiments, all or part of the embodiments can be realized by software, hardware, firmware or any combination thereof. When realized by software, all or part of the embodiments can be realized in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated.

[0285] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented by other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0286] It should be understood that in various embodiments of the present application, the size of the serial number of each process does not mean the execution order, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0287] In addition, the terms "system" and "network" are often used interchangeably herein. The term "and / or", only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the following three cases: A exists alone, A and B exist together, and B exists alone. In addition, the character " / " generally represents that the front and rear associated objects are an "or" relationship. For example, A / B can represent A or B.

[0288] In the embodiments of the present application, "information", "signal", "message", "channel", and "signalling" can be used interchangeably at times. It should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent. "Of", "corresponding", and "corresponding" can be used interchangeably at times. It should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent.

[0289] Here, it is uniformly stated that the specific implementation mode of "predefined" can include any of the following: protocol predefined, or manufacturer specification of the communication device, or communication operator defined, or pre-installed in the communication device when the communication device is shipped, or other agreed manner agreed in advance.

[0290] The terms (or numbers) "first", "second", and the like appearing in the embodiments of the present application are only for description purposes, that is, only to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features limited by "first", "second", and the like can explicitly or implicitly include one or more features. In the description of the embodiments of the present application, "at least one" means one or more. The meaning of "multiple" is two or more. "At least one of the following" or the like means any combination of these items, including any combination of single (item) or multiple (item).

[0291] For example, the meaning of the expression similar to "the item includes at least one of A, B, and C" in the embodiments of the present application, if not specifically stated, generally means that the item can be any one of A, B, C, A and B, A and C, B and C, A, B and C, A and A, A, A and A, A, A and B, A, A and C, A, B and B, A, C and C, B and B, B, B and B, B, B and C, C and C, C, C and C, and other combinations of A, B, and C. The above is an example of 3 elements A, B, and C to illustrate the optional entries of the item. When the expression is "the item includes at least one of A, B,..., and X", that is, the expression has more elements, the applicable entries of the item can also be obtained according to the foregoing rules.

[0292] In conclusion, the above only describes the preferred embodiments of the technical solutions of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A communication method characterized by comprising: The method is applied to a user equipment (UE), and comprises: receive a first message from a network device, the first message comprising information related to a first AI model, the information related to the first AI model comprising one or more of: a network structure of the first AI model, a layer connection relationship of the first AI model, a model parameter of the first AI model; the first AI model being used to output prediction information at a t+k time from input information at a t time, wherein ; determining a prediction result according to input information and the first AI model, the prediction result comprising one or more of the following predicted values: time delay and Doppler shift; performing pre-compensation processing based on the prediction result to obtain a compensated transmission time and / or a compensated transmission frequency; sending a signal to a network device based on the compensated transmission time and / or the compensated transmission frequency.

2. The method of claim 1, wherein, The prediction result further comprises a predicted value of a Doppler rate.

3. The method of claim 1, wherein, The determination of the prediction result according to the input information and the first AI model comprises: constructing an input feature vector and inputting the input feature vector into the first AI model, the input feature vector comprising at least a first feature sequence, the first feature sequence comprising one or more of the following sequences: a time delay observation value sequence, a Doppler shift sequence, and a Doppler rate sequence; obtaining the prediction result output by the first AI model.

4. The method of claim 3, wherein, The input feature vector further comprises a second feature sequence and / or a third feature sequence; the second feature sequence is a feature sequence related to the motion of the UE; and the third feature sequence is a feature sequence related to a spatial angle.

5. The method of claim 4, wherein, The second feature sequence comprises a velocity vector sequence and / or an acceleration vector sequence.

6. The method of claim 4, wherein, The third feature sequence comprises an included angle sequence and / or a satellite elevation angle sequence.

7. The method according to any one of claims 1 to 6, characterized in that, The method further comprises: determining whether the performance of the first AI model meets a first preset condition according to the prediction result and an actual measurement value of a channel; in a case where it is determined that the performance of the first AI model does not meet the first preset condition, calculating channel information based on a geometric algorithm and performing pre-compensation processing; in a case where it is determined that the performance of the first AI model meets the first preset condition, performing pre-compensation processing using the prediction result output by the first AI model.

8. The method of claim 7, wherein, The performance of the first AI model does not meet the first preset condition, comprising: a prediction error value exceeds an error threshold value, and a duration for which the prediction error value exceeds the error threshold value exceeds a first duration threshold value, wherein the prediction error value is determined based on a sequence predicted by the first AI model and a sequence actually measured by the UE. The performance of the first AI model meets the first preset condition, comprising: the prediction error value exceeds the error threshold value, and a duration for which the prediction error value exceeds the error threshold value does not exceed the first duration threshold value; or the prediction error value does not exceed the error threshold value.

9. The method of claim 7, wherein, The performance of the first AI model does not meet the first preset condition, comprising one or more of the following: a first uncertainty parameter exceeds a first confidence threshold value for M consecutive times, the first uncertainty parameter being a parameter related to the time delay output by the first AI model; a second uncertainty parameter exceeds a second confidence threshold value for W consecutive times, the second uncertainty parameter being a parameter related to the Doppler shift output by the first AI model; a third uncertainty parameter exceeds a third confidence threshold value for Q consecutive times, the third uncertainty parameter being a parameter related to the Doppler rate output by the first AI model.

10. The method of claim 7, wherein, The method further comprises: In a case where the performance of the first AI model does not satisfy the first preset condition, feedback information is sent to the network device, the feedback information being used to inform the network device that the performance of the first AI model does not satisfy the first preset condition.

11. The method of claim 7, wherein, In a case where the UE calculates channel information based on a geometric algorithm and performs pre-compensation processing, the method further includes: In a case where the first AI model satisfies the second preset condition, pre-compensation processing is performed again using the output result of the first AI model.

12. The method of claim 11, wherein, The first AI model satisfying the second preset condition includes one or more of the following: A prediction error value of the time delay does not exceed a fourth error threshold, and a duration for which the prediction error value of the time delay does not exceed the fourth error threshold exceeds a second duration threshold; A prediction error value of the Doppler shift does not exceed a fifth error threshold, and a duration for which the prediction error value of the Doppler shift does not exceed the fifth error threshold exceeds the second duration threshold; A prediction error value of the Doppler rate of change does not exceed a sixth error threshold, and a duration for which the prediction error value of the Doppler rate of change does not exceed the sixth error threshold exceeds the second duration threshold.

13. The method according to any one of claims 1 to 6, characterized in that, In a case where the result output by the first AI model includes a predicted value of the time delay, the method further includes: Based on the predicted value of the time delay output by the first AI model, a first uncertainty parameter is determined; the first uncertainty parameter is an uncertainty parameter related to the predicted value of the time delay, and the first uncertainty parameter is used to represent the accuracy of the predicted value of the time delay output by the first AI model; Based on the first uncertainty parameter, a duration of an SSB measurement time configuration (SMTC) window is determined.

14. The method of claim 13, wherein, The determination of the duration of the SMTC window based on the first uncertainty parameter includes: When the first uncertainty parameter does not exceed a preset threshold, the duration of the SMTC window is reduced; When the first uncertainty parameter exceeds the preset threshold, the duration of the SMTC window is maintained.

15. The method according to any one of claims 1 to 6, characterized in that, Before receiving the first message from the network device, the method further includes: Historical data is reported to the network device; the historical data is used to train the first AI model; the historical data includes one or more of the following information: channel state information, kinematic information, and spatial angle information; The channel state information includes one or more of the following: reference signal received power (RSRP), measurement time delay, and Doppler shift; the kinematic information includes velocity and / or acceleration; and the spatial angle information includes a satellite elevation angle and / or a beam angle calculated based on a historical time.

16. The method according to any one of claims 1 to 6, characterized in that, The first AI model includes an input layer, a feature extraction layer, and a multi-head output layer; a Dropout layer is inserted in the feature extraction layer; the input layer is connected to an input end of the feature extraction layer; and an output end of the feature extraction layer is connected to the multi-head output layer.

17. The method of claim 16, wherein, The first AI model is an LSTM model based on a Bayesian model; and the feature extraction layer includes an LSTM layer 1 and an LSTM layer 2. The Dropout layer is arranged between the LSTM layer 1 and the LSTM layer 2; the input end of the LSTM layer 1 is connected to the input layer, and the output end of the LSTM layer 1 is connected to the intermediate Dropout layer; the input end of the LSTM layer 2 is connected to the Dropout layer, and the output end of the LSTM layer 2 is connected to the multi-head output layer; The multi-head output layer comprises a fully connected output layer 1 and a fully connected output layer 2; the fully connected output layer 1 is used for outputting the predicted value of the time delay and / or the predicted value of the Doppler shift; and the fully connected output layer 2 is used for outputting the predicted value of the Doppler rate.

18. A method of communication, comprising: The method is applied to a network device, and the method comprises: training a first AI model according to the training data set, the first AI model being configured to output prediction information of a t+k time point based on input information of the t time point, the prediction information comprising one or more of: time delay, Doppler shift, Doppler rate, wherein ; sending a first message, wherein the first message comprises information related to the first AI model, and the information related to the first AI model comprises one or more of the following: a network structure of the first AI model, a layer connection relationship of the first AI model, and a model parameter of the first AI model.

19. The method of claim 18, wherein, The method further comprises: receiving historical data from one or more UEs, wherein the historical data is used for training the first AI model; the historical data comprises one or more of the following: channel state information, kinematic information, and spatial angle information; wherein the channel state information comprises one or more of the following: reference signal received power (RSRP), measurement time delay, and Doppler shift; the kinematic information comprises velocity and / or acceleration; and the spatial angle information comprises a satellite elevation angle and / or a beam angle calculated based on a historical time point; constructing the training data set according to the historical data.

20. The method of claim 18 or 19, wherein, The first AI model comprises an input layer, a feature extraction layer, and a multi-head output layer; wherein a Dropout layer is inserted in the feature extraction layer; the input layer is connected to the input end of the feature extraction layer; and the output end of the feature extraction layer is connected to the multi-head output layer.

21. The method of claim 20, wherein, The first AI model is an LSTM model based on a Bayesian model; and the feature extraction layer comprises an LSTM layer 1 and an LSTM layer 2. The Dropout layer is arranged between the LSTM layer 1 and the LSTM layer 2; the input end of the LSTM layer 1 is connected to the input layer, and the output end of the LSTM layer 1 is connected to the intermediate Dropout layer; the input end of the LSTM layer 2 is connected to the Dropout layer, and the output end of the LSTM layer 2 is connected to the multi-head output layer; The multi-head output layer comprises a fully connected output layer 1 and a fully connected output layer 2; the fully connected output layer 1 is used for outputting the predicted value of the time delay and / or the predicted value of the Doppler shift; and the fully connected output layer 2 is used for outputting the predicted value of the Doppler rate.

22. The method of claim 18 or 19, wherein, Before the first message is sent to the UE, the method further comprises: performing compression processing on the first AI model to obtain the information related to the first AI model.

23. The method of claim 18 or 19, wherein, The method further comprises: receiving feedback information, wherein the feedback information is used for notifying the network device that the performance of the first AI model does not satisfy a first preset condition.

24. The method of claim 23, wherein, The performance of the first AI model does not satisfy a first preset condition, including: a prediction error value exceeds an error threshold, and a duration in which the prediction error value exceeds the error threshold exceeds a first duration threshold, wherein the prediction error value is determined based on a sequence predicted by the first AI model and a sequence actually measured by the UE.

25. The method of claim 23, wherein, The performance of the first AI model does not satisfy a first preset condition, including one or more of: A first uncertainty parameter exceeds a first confidence threshold for M consecutive times, the first uncertainty parameter being related to a time delay output by the first AI model; A second uncertainty parameter exceeds a second confidence threshold for W consecutive times, the second uncertainty parameter being related to a Doppler shift output by the first AI model; A third uncertainty parameter exceeds a third confidence threshold for Q consecutive times, the third uncertainty parameter being related to a Doppler rate output by the first AI model.

26. A communications device, characterized by The communication device includes at least one processor coupled with a memory, the memory having stored therein programs or instructions, the processor executing the programs or instructions causing the communication device to perform the method of any one of claims 1-17, or causing the communication device to perform the method of any one of claims 18-25.

27. A computer-readable storage medium, characterized in that, A computer program or instructions stored thereon, wherein the computer program or instructions, when executed, cause a computer to perform the method of any one of claims 1-17, or cause the computer to perform the method of any one of claims 18-25.

28. A communication system, characterized by The communication device of claim 26.

29. A chip system, characterized by The chip system includes one or more processors for invoking and running instructions stored in a memory, such that the method of any one of claims 1-17 is performed, or the method of any one of claims 18-25 is performed.

30. A computer program product, characterised in that, Computer program instructions that cause a computer to perform the method of any one of claims 1-17; or cause the computer to perform the method of any one of claims 18-25.

Images