Compression model update methods, apparatus, systems and storage media

Abstract

This application relates to a compression model update method, apparatus, system, and storage medium. The method includes: acquiring compressed Channel State Information (CSI) reports and uncompressed CSI reports reported by a user terminal; decompressing the compressed CSI reports to obtain decompressed CSI reports; acquiring the similarity between the decompressed and uncompressed CSI reports to obtain a similarity index; and outputting a model update instruction if the similarity index is less than or equal to a first threshold. The model update instruction instructs the user terminal to update the compressed model to the updated model. Upon receiving the model update instruction, the user terminal updates its stored compressed model to the updated model, ensuring that the subsequently updated compressed CSI reports obtained through compression using the updated model better meet the needs of the current wireless scenario, thereby maintaining the wireless performance of the existing network.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a compression model update method, apparatus, system and storage medium. Background Technology

[0002] Wireless communication systems have been widely deployed for everyday voice, video, data, and SMS services. Mobile communication has evolved through several stages: 2G (GSM, Global System for Mobile Communications), 3G (TD-SCDMA, UMTS), and 4G (LTE, Long Term Evolution), and is now in the research and deployment phase of 5G (NR, New Radio).

[0003] The 3GPP NR (New Radio) protocol supports three service types: eMBB (Enhanced Mobile Broadband), URLLC (Ultra-reliable and Low Latency Communications), and eMTC (Massive Machine Type Communication). In Release 17, 3GPP studied the performance and framework of high-level AI (Artificial Intelligence), and in Release 18, it approved a project for AI NR air interface technology. In the AI ​​NR air interface technology research, three directions were identified: CSI (Channel State Information) feedback enhancement, beam management, and positioning accuracy enhancement. CSI feedback enhancement mainly identified two sub-directions: CSI compression and CSI prediction.

[0004] In CSI feedback enhancement, network models are primarily trained using artificial intelligence (AI) and machine learning (ML) to compress or predict CSI. However, after the AI / ML models are deployed, as the wireless environment changes during terminal operation, the original AI / ML models will no longer be suitable for the current wireless environment, leading to a decline in the wireless performance of the existing network. Summary of the Invention

[0005] Therefore, it is necessary to provide a compressed model update method, apparatus, system, and storage medium that can maintain the existing wireless performance after the wireless scenario in which the terminal operates changes, in order to address the above-mentioned technical problems.

[0006] In a first aspect, this application provides a compression model update method, which is applied to a base station, and the method includes:

[0007] Acquire compressed channel status information (CSI) reports and uncompressed channel status information (CSI) reports reported by the user terminal; wherein the compressed channel status information (CSI) reports are obtained by the user terminal through processing of a trained compression model;

[0008] The compressed Channel State Information (CSI) report is decompressed to obtain a decompressed Channel State Information (CSI) report.

[0009] The similarity between the decompressed Channel State Information (CSI) report and the uncompressed Channel State Information (CSI) report is obtained to generate a similarity index.

[0010] If the similarity index is less than or equal to the first threshold, a model update instruction is output; wherein, the model update instruction is used to instruct the user terminal to update the compressed model to an updated model.

[0011] Secondly, this application also provides a compression model update method, which is applied to a user terminal, and the method includes:

[0012] The uncompressed Channel State Information (CSI) report is compressed using the trained compression model to obtain a compressed Channel State Information (CSI) report.

[0013] Report the compressed channel state information (CSI) report and the uncompressed channel state information (CSI) report to the base station;

[0014] If a model update instruction is received from the base station, the compressed model is updated to the updated model.

[0015] The model update instruction is output by the base station when the similarity index is less than or equal to a first threshold. The similarity index is the similarity between the decompressed Channel State Information (CSI) report and the uncompressed CSI report. The decompressed CSI report is obtained by the base station decompressing the compressed CSI report.

[0016] Thirdly, this application also provides a compression model update apparatus, which is applied to a base station, and the apparatus includes:

[0017] The data acquisition module is used to acquire compressed channel status information (CSI) reports and uncompressed channel status information (CSI) reports reported by the user terminal; wherein the compressed channel status information (CSI) reports are obtained by the user terminal through processing of a trained compression model;

[0018] The decompression module is used to decompress the compressed Channel State Information (CSI) report to obtain a decompressed Channel State Information (CSI) report.

[0019] The similarity index acquisition module is used to acquire the similarity between the decompressed channel state information (CSI) report and the uncompressed channel state information (CSI) report to obtain a similarity index.

[0020] An update instruction output module is used to output a model update instruction if the similarity index is less than or equal to a first threshold; wherein the model update instruction is used to instruct the user terminal to update the compressed model to an updated model.

[0021] Fourthly, this application also provides a compressed model update apparatus, which is applied to a user terminal, and the apparatus includes:

[0022] The compression module is used to compress the uncompressed Channel State Information (CSI) report according to the trained compression model to obtain a compressed Channel State Information (CSI) report.

[0023] The data reporting module is used to report the compressed channel status information (CSI) report and the uncompressed channel status information (CSI) report to the base station;

[0024] The model update module is used to update the compressed model to an updated model if it receives a model update instruction transmitted by the base station.

[0025] The model update instruction is output by the base station when the similarity index is less than or equal to a first threshold. The similarity index is the similarity between the decompressed Channel State Information (CSI) report and the uncompressed CSI report. The decompressed CSI report is obtained by the base station decompressing the compressed CSI report.

[0026] Fifthly, this application also provides a compressed model update system, including a base station and a user terminal connected to the base station; wherein: the base station is used to perform the steps of the method described in the first aspect; and the user terminal is used to perform the steps of the method described in the second aspect.

[0027] Sixthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0028] In the aforementioned compression model update method, apparatus, system, and storage medium, the base station obtains a decompressed Channel State Information (CSI) report based on the compressed CSI report reported by the user terminal. Simultaneously, it determines whether the compressed model conforms to the current operating wireless scenario based on the similarity between the decompressed and uncompressed CSI reports. If it does not conform to the current operating wireless scenario, it outputs a model update command. Upon receiving the model update command, the user terminal updates its stored compressed model to the updated model, ensuring that the subsequently obtained updated compressed CSI reports better meet the needs of the current operating wireless scenario, thereby maintaining the wireless performance of the existing network. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a diagram illustrating the application environment of the compressed model update method in one embodiment;

[0031] Figure 2 This is a flowchart illustrating a compression model update method implemented from the perspective of a base station in one embodiment;

[0032] Figure 3 This is a flowchart illustrating a compression model update method implemented from the base station's perspective in another embodiment;

[0033] Figure 4 This is a flowchart illustrating the compression model update method implemented from the base station's perspective in yet another embodiment;

[0034] Figure 5 This is a schematic diagram of the time domain configuration for CSI reporting during the model monitoring period in one embodiment;

[0035] Figure 6 This is a schematic diagram illustrating the resource scheduling process from the perspective of a base station in one embodiment;

[0036] Figure 7 This is a schematic diagram illustrating resource scheduling from the perspective of a base station in another embodiment;

[0037] Figure 8 This is a schematic diagram of the modules for interaction between the base station and the user terminal in one embodiment;

[0038] Figure 9This is a flowchart illustrating a compression model update method implemented from the user terminal's perspective in one embodiment;

[0039] Figure 10 This is a block diagram of a compression model update apparatus implemented from the perspective of a base station in one embodiment;

[0040] Figure 11 This is a block diagram of a compression model update device implemented from the perspective of a user terminal in one embodiment;

[0041] Figure 12 This is a schematic diagram illustrating the signaling interaction between the base station and the user terminal triggered by the base station-side model monitoring period in one embodiment.

[0042] Figure 13 This is a schematic diagram illustrating the signaling interaction between the base station and the user terminal triggered by a monitoring event on the base station side in one embodiment. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0044] In IRS (Intelligent Reflecting Surface) assisted frequency division duplex multiple-input multiple-output (MIMO) communication systems that employ orthogonal frequency division multiplexing as a multi-carrier modulation scheme, in order to improve the system gain, the user equipment (UE) needs to send channel state information (CSI) for multiple channels to the base station (BS) via a feedback link. Therefore, compared to traditional MIMO systems, the amount of CSI data and the feedback overhead in this system are enormous.

[0045] Related technologies propose using deep learning techniques to compress and reconstruct CSI, thereby reducing the amount of CSI data and feedback overhead. For example, the compression model can be a deep learning network such as CNN, RNN, Transformer, ResNet, CsiNet, CRNet, DS-NLCsiNet, and QuanCsiNet. However, after model deployment, as the user terminal's working wireless scenario changes, the original compression model may no longer be suitable for the current wireless environment, leading to a decline in the wireless performance of the existing network.

[0046] Based on this, this application proposes a compression model update method, apparatus, system, and storage medium, which can update the original compression model in a timely manner when the wireless scenario changes, preventing the degradation of the wireless performance of the existing network.

[0047] In one embodiment, such as Figure 2 As shown, a compressed model update method is provided, which is applied to... Figure 1 Taking base station 104 as an example, the explanation includes the following steps:

[0048] Step S110: Obtain the compressed channel status information (CSI) report and the uncompressed channel status information (CSI) report reported by the user terminal; wherein, the compressed channel status information (CSI) report is obtained by the user terminal through processing of the trained compression model.

[0049] Specifically, the compression model in this embodiment is a pre-trained AI / ML model. The same compression model exists in both the user terminal 102 and the base station 104. The compression model in the user terminal 102 is used to compress the uncompressed Channel State Information (CSI) report (non-AI / ML based CSI report) to obtain a compressed Channel State Information (CSI) report (AI / ML based CSI report). The compression model in the base station 104 is used to decompress the compressed CSI report to restore the CSI report. It is understood that the restored CSI report includes: PMI (Precoding Matrix Indicator), CQI (Channel Quality Indicator), and RI (Rank Indicator). For downlink data, PMI is used to determine the frequency domain scheduling precoder, CQI is used to determine the modulation and coding scheme (MCS), and RI is used to determine the layer number. The base station 104 notifies the user terminal 102 of the MCS used in the downlink DCI format.

[0050] Base station 104 can notify user terminal 102 via RRC (Radio Resource Control) signaling, causing user terminal 102 to report compressed channel state information (CSI) and uncompressed channel state information (CSI). The compressed CSI report includes compressed PMI, CQI, and RI, while the uncompressed CSI report includes raw PMI, CQI, and RI.

[0051] Step S120: Decompress the compressed Channel State Information (CSI) report to obtain a decompressed Channel State Information (CSI) report. Specifically, base station 104 decompresses the compressed CSI report using its built-in compression model to obtain a decompressed CSI report. The decompressed CSI report includes decompressed PMI, CQI, and RI.

[0052] Step S130: Obtain the similarity between the decompressed Channel State Information (CSI) report and the uncompressed Channel State Information (CSI) report to obtain the similarity index.

[0053] Specifically, the similarity GCS between the decompressed Channel State Information (CSI) report and the uncompressed Channel State Information (CSI) report is calculated based on the obtained reports, and the similarity GCS is used as a similarity index. The formula for calculating GCS is as follows:

[0054]

[0055] Among them, W i j It is the j-th feature vector of resource element i in the uncompressed Channel State Information (CSI) report, where K is the rank; W~ i j Let $\mathbf{i}$ be the j-th feature vector of resource element $i$ in the decompressed Channel State Information (CSI) report, $N$ be the total number of resource elements, and $E{·}$ define the average output of multiple samples. The closer the similarity index is to 1, the higher the similarity between the decompressed and uncompressed CSI reports, and the better the compression and decompression performance of the compression model. Conversely, the closer the similarity index is to 0, the lower the similarity between the decompressed and uncompressed CSI reports, and the worse the compression and decompression performance of the compression model. In this case, the compression model does not meet the needs of the current wireless scenario.

[0056] Step S140: If the similarity index is less than or equal to the first threshold, then output a model update instruction; wherein, the model update instruction is used to instruct the user terminal to update the compressed model to the updated model.

[0057] Specifically, the first threshold can be set to 85%. When the calculated similarity index is less than or equal to the first threshold, it indicates that the similarity between the decompressed Channel State Information (CSI) report and the uncompressed CSI report is low, and the compressed model does not meet the needs of the current wireless scenario. In this case, base station 104 can output a model update command via RRC signaling. After receiving the model update command, user terminal 102 will update the original compressed model, changing it to the updated model. It is understandable that, in the absence of a suitable updated model, user terminal 102 can switch to the traditional CSI reporting mode. Furthermore, when user terminal 102 updates the compressed model to the updated model, the corresponding compressed model in base station 104 will also change to the updated model.

[0058] In the aforementioned compression model update method, the base station obtains the decompressed Channel State Information (CSI) report based on the compressed CSI report reported by the user terminal. Simultaneously, it determines whether the compressed model conforms to the current wireless scenario based on the similarity between the decompressed and uncompressed CSI reports. If it does not conform to the current wireless scenario, it outputs a model update command. Upon receiving the model update command, the user terminal updates its stored compressed model to the updated model, ensuring that the subsequently obtained updated compressed CSI reports better meet the needs of the current wireless scenario, thereby maintaining the wireless performance of the existing network.

[0059] In one embodiment, such as Figure 3 As shown, the compressed model update method also includes:

[0060] Step S150: Statistically analyze the received negative responses (NACK) to obtain the negative response probability; whereby the negative response probability is the proportion of negative responses (NACK) to response signals.

[0061] Specifically, the response signal includes ACK (Acknowledgement) and NACK (Negative Acknowledgement). An ACK is a positive feedback mechanism; the receiver replies with a message to the sender after receiving data. A NACK is a negative feedback mechanism; the receiver only notifies the sender if no data has been received. Base station 104 can count the number of NACKs sent by user terminal 102 by analyzing the received NACKs. Simultaneously, based on the number of ACKs received, base station 104 can calculate the proportion of NACKs in the total response signal, thus obtaining the probability of a negative ACK.

[0062] In step S160, if the probability of a negative response is greater than the second threshold, then output a model update instruction.

[0063] Specifically, when the probability of a negative response is high, it indicates that user terminal 102 has repeatedly failed to accurately demodulate the data sent by base station 104. This suggests that the current network's wireless performance is poor and the compression model needs to be updated. Base station 104 will then output a model update command. For example, if the first threshold is set to 85% and the second threshold is set to 10%, and the currently calculated similarity index is less than 85% while the negative response probability is greater than 10%, base station 104 will output a model update command to user terminal 102. This embodiment of the application uses a combination of negative response probability to determine whether the original compression model has failed, resulting in higher accuracy. In some other embodiments, the compression model can also be determined to be failed by calculating the proportion of positive ACKs to the response signal; when the proportion is less than a certain threshold.

[0064] In one embodiment, such as Figure 4 As shown, before the steps of obtaining the compressed channel state information (CSI) report and the uncompressed channel state information (CSI) report reported by the user terminal, the compression model update method further includes:

[0065] Step S170: Obtain the model monitoring triggering method information configured by the network management system. Specifically, base station 104 first obtains the model monitoring triggering method information configured by the network management system. The model detection triggering method is used to notify base station 104 to select the triggering method for user terminal 102 to report CSI reports.

[0066] Step S180: Obtain target triggering method information based on model monitoring triggering method information; wherein, the target triggering method information includes periodic triggering information. Specifically, the target triggering method information also includes event triggering information. After receiving the model monitoring triggering method information, base station 104 determines whether the conditions for periodic triggering or event triggering are met. If the conditions for periodic triggering are met, the target triggering method information is determined to be periodic triggering information; if the conditions for event triggering are met, the target triggering method information is determined to be event triggering information. When the target triggering method information is determined to be event triggering information, a model update request will only be triggered when a specific event meets the conditions, effectively reducing the overhead of uplink feedback resources in the periodic triggering method.

[0067] Step S190: Output target triggering method information. Specifically, after determining the target triggering method information, base station 104 sends the target triggering method information to user terminal 102 via RRC signaling. Upon receiving the corresponding target triggering method information, user terminal 102 changes its own compressed model monitoring event triggering method. For example, when the target triggering method information is periodic triggering information, user terminal 102 will report compressed channel status information (CSI) and uncompressed channel status information (CSI) at fixed intervals; when the target triggering method information is event triggering information, user terminal 102 will start reporting compressed channel status information (CSI) and uncompressed channel status information (CSI) when it detects the occurrence of the corresponding event. Users can configure the model monitoring triggering method information through network management and select the specific triggering method (periodic triggering or event triggering) for the user terminal to report CSI reports to meet the update requirements of the compressed model in different scenarios.

[0068] In one embodiment, the periodic triggering information includes: a model monitoring period and a monitoring window duration. The user terminal reports uncompressed Channel State Information (CSI) reports during the monitoring window duration and compressed Channel State Information (CSI) reports during other times of the model monitoring period. Specifically, the monitoring window duration is a period within the model monitoring period. The user terminal 102 reports CSI reports cyclically according to the model monitoring period, that is, it reports uncompressed Channel State Information (CSI) reports during the monitoring window duration and compressed Channel State Information (CSI) reports during other times of the model monitoring period that are not within the monitoring window duration. (Refer to...) Figure 5 This is a schematic diagram illustrating the time-domain configuration of CSI reporting during the model monitoring period in one embodiment. The periodic triggering information is sent via RRC signaling, and the following are configured in the CSI-ReportConfig of the RRC signaling: Offset (0, ..., T-1), model monitoring period T (e.g., 4, 5, 8, 10, 16, 20, ..., 160, 320, 640, 1280 slots, etc.), and monitoring window length L (e.g., 100, 200, ..., 800 slots, etc.). The monitoring window length L represents the duration of the monitoring window. User terminal 102 reports uncompressed channel state information (CSI) within the monitoring window duration and compressed channel state information (CSI) within the non-monitoring window duration of the model monitoring period T. It can be understood that the monitoring window can be located at any other position within the model monitoring period T.

[0069] In one embodiment, the model update instruction includes a model deactivation instruction and a model switching instruction. The user terminal is used to stop the operation of the compressed model according to the model deactivation instruction, and the user terminal is used to switch the compressed model to the updated model according to the model switching instruction. Specifically, the model update instruction output by the base station 104 includes a model deactivation instruction and a model switching instruction. After receiving the model update instruction, the user terminal 102 stops the operation of the current compressed model according to the model deactivation instruction, and at the same time, the user terminal 102 switches the compressed model to the updated model according to the model switching instruction. It can be understood that if the user terminal 102 cannot switch to a suitable updated model, the user terminal 102 automatically falls back to the traditional CSI reporting mode.

[0070] In one embodiment, refer to Figure 6 After the step of outputting model update instructions, the compressed model update method also includes:

[0071] Step S111: Obtain the updated compressed channel state information (CSI) report reported by the user terminal; wherein, the updated compressed channel state information (CSI) report is obtained by the user terminal processing the uncompressed channel state information (CSI) report using an update model.

[0072] Specifically, after base station 104 outputs a model update command, user terminal 102 updates the compressed model to an updated model, and then uses the updated model to compress the uncompressed Channel State Information (CSI) report to obtain an updated compressed CSI report, which is then reported to base station 104. Base station 104 can then obtain the updated compressed CSI report obtained by user terminal 102 using the updated model.

[0073] Step S112: Decompress the updated compressed channel state information (CSI) report to obtain a decompressed updated channel state information (CSI) report.

[0074] Specifically, when base station 104 notifies user terminal 102 to update the compression model to the updated model, the compression model stored internally by base station 104 is also updated to the updated model, meaning that the compression models in base station 104 and user terminal 102 are the same. After receiving the updated compressed channel state information (CSI) report, base station 104 decompresses the updated compressed channel state information (CSI) report using the updated model to obtain the decompressed updated channel state information (CSI) report.

[0075] Step S113: Resource scheduling is performed based on the decompressed and updated Channel State Information (CSI) report. Specifically, when switching from a compressed model to an updated model, base station 104 selects the AI / ML compression model that best suits the current wireless scenario, which can maintain the existing network's wireless performance relatively well. Therefore, base station 104 can directly perform resource scheduling based on the decompressed and updated CSI report, achieving good wireless performance in the current wireless scenario. It is understood that in some other embodiments, the similarity between the decompressed and updated CSI reports and the uncompressed CSI reports can be used to determine whether the updated model meets the needs of the current wireless scenario, and a new updated model can be continuously switched until the obtained similarity index meets the requirements.

[0076] In one embodiment, refer to Figure 7 The steps for resource scheduling based on the decompressed and updated Channel State Information (CSI) report include:

[0077] Step S114: Obtain the scheduling time interval; wherein, the scheduling time interval is the time between the base station obtaining the decompressed and updated Channel State Information (CSI) report and the scheduler using the decompressed and updated CSI report.

[0078] Specifically, since the wireless scenario changes dynamically over time, if the base station 104 calculates and obtains the decompressed and updated Channel State Information (CSI) report but does not use it for a long period of time, the decompressed and updated CSI report may not be consistent with the current wireless scenario. Therefore, the base station 104 first needs to obtain the time between the decompressed and updated CSI report and the timer's use of the decompressed and updated CSI report to obtain the scheduling time interval.

[0079] Step S115: If the scheduling time interval is greater than the third threshold, a predicted channel state information (CSI) report is obtained based on the decompressed updated CSI report. Specifically, after obtaining the scheduling time interval, the scheduling time interval is compared with the third threshold. For example, if the third threshold is set to 4 milliseconds, when the scheduling time interval is greater than the third threshold, the base station 104 will input the decompressed updated CSI report into the update model for prediction, obtaining the predicted CSI report after the corresponding time interval; when the scheduling time interval is less than or equal to the third threshold, the base station 104 will perform resource scheduling based on the decompressed updated CSI report.

[0080] Step S116: Perform resource scheduling based on the predicted channel state information (CSI) report. Specifically, after obtaining the predicted channel state information (CSI) report through base station 104, resource scheduling is performed based on the predicted channel state information (CSI) report, which can effectively improve the efficiency of resource scheduling.

[0081] Specific examples, such as Figure 8 As shown, base station 104 includes a dataset collection module, a model training module, a model inference module, a model monitoring module, and a model update module. The dataset collection module collects compressed channel state information (CSI) reports, uncompressed CSI reports, traditional CSI data, and AI / ML capability information supported by user terminal 102, all reported by user terminal 102. The model training module trains a suitable AI / ML compressed model on the network side based on the model library supported by user terminal 102 and the reported traditional CSI data. Network-side model training perfectly matches the changing characteristics of the wireless environment surrounding the base station. The model inference module recovers the compressed CSI data through inference based on the model deployment. Model monitoring is used to monitor the accuracy of the recovered compressed CSI data and updates the compressed model through the model update module if inaccurate data is found.

[0082] In one embodiment, such as Figure 9 As shown, a compressed model update method is provided, which is applied to... Figure 1 Taking user terminal 102 as an example, the explanation includes the following steps:

[0083] Step S210: Compress the uncompressed Channel State Information (CSI) report according to the trained compression model to obtain a compressed Channel State Information (CSI) report.

[0084] Specifically, base station 104 trains a model based on information reported by user terminal 102 and sends the trained compressed model to user terminal 102. The compressed model is a trained AI / ML model, and the same compressed model exists in both user terminal 102 and base station 104. The compressed model in user terminal 102 is used to compress the uncompressed Channel State Information (CSI) report to obtain a compressed CSI report. The compressed model in base station 104 is used to decompress the compressed CSI report to restore the CSI report. It is understood that the restored CSI report includes PMI, CQI, and RI. For downlink data, PMI is used to determine the frequency domain scheduling precoder, CQI is used to determine the modulation and coding scheme (MCS), and RI is used to determine the layer. Base station 104 notifies user terminal 102 of the MCS used in the downlink DCI format.

[0085] Step S220: Report Compressed Channel State Information (CSI) report and Uncompressed Channel State Information (CSI) report to the base station. Specifically, base station 104 can notify user terminal 102 via RRC signaling to report both compressed and uncompressed channel state information (CSI) reports. The compressed channel state information (CSI) report includes compressed PMI, CQI, and RI, while the uncompressed channel state information (CSI) report includes raw PMI, CQI, and RI.

[0086] Step S230: If a model update instruction is received from the base station, the compressed model is updated to an updated model; wherein, the model update instruction is output by the base station when the similarity index is less than or equal to the first threshold, the similarity index is the similarity between the decompressed channel state information (CSI) report and the uncompressed channel state information (CSI) report, and the decompressed channel state information (CSI) report is obtained by the base station decompressing the compressed channel state information (CSI) report.

[0087] Specifically, if user terminal 102 receives a model update instruction from the base station, it will update its stored compressed model to the updated model. After obtaining the compressed Channel State Information (CSI) report and the uncompressed CSI report, base station 104 decompresses the compressed CSI report using its built-in compression model to obtain the decompressed CSI report. The decompressed CSI report includes decompressed PMI, CQI, and RI. Then, base station 104 calculates the similarity GCS between the decompressed and uncompressed CSI reports and uses the GCS as a similarity index. The formula for calculating GCS is as follows:

[0088]

[0089] Among them, W i j It is the j-th feature vector of resource element i in the uncompressed Channel State Information (CSI) report, where K is the rank; W~ i jLet $\mathbf{i}$ be the j-th feature vector of resource element $i$ in the decompressed Channel State Information (CSI) report, $N$ be the total number of resource elements, and $E{·}$ define the average output of multiple samples. The closer the similarity index is to 1, the higher the similarity between the decompressed and uncompressed CSI reports, and the better the compression and decompression performance of the compression model. Conversely, the closer the similarity index is to 0, the lower the similarity between the decompressed and uncompressed CSI reports, and the worse the compression and decompression performance of the compression model. In this case, the compression model does not meet the needs of the current wireless scenario.

[0090] For a specific example, the first threshold can be set to 85%. When the calculated similarity index is less than or equal to the first threshold, it indicates that the similarity between the decompressed Channel State Information (CSI) report and the uncompressed CSI report is low, and the compressed model does not meet the needs of the current wireless scenario. In this case, base station 104 can output a model update command via RRC signaling. After receiving the model update command, user terminal 102 will update the original compressed model, changing it to the updated model. It is understandable that, in the absence of a suitable updated model, user terminal 102 can switch to the traditional CSI reporting mode. Furthermore, when user terminal 102 updates the compressed model to the updated model, the corresponding compressed model in base station 104 will also change to the updated model.

[0091] In the aforementioned compression model update method, the base station obtains the decompressed Channel State Information (CSI) report based on the compressed CSI report reported by the user terminal. Simultaneously, it determines whether the compressed model conforms to the current wireless scenario based on the similarity between the decompressed and uncompressed CSI reports. If it does not conform to the current wireless scenario, it outputs a model update command. Upon receiving the model update command, the user terminal updates its stored compressed model to the updated model, ensuring that the subsequently obtained updated compressed CSI reports better meet the needs of the current wireless scenario, thereby maintaining the wireless performance of the existing network.

[0092] In one embodiment, the model update instruction is output by the base station when the negative acknowledgment probability is greater than the second threshold. The negative acknowledgment probability is obtained by the base station by statistically analyzing the received negative acknowledgments (NACKs), where the negative acknowledgment probability is the proportion of negative acknowledgments (NACKs) to acknowledgment signals.

[0093] Specifically, the response signal includes ACK (Acknowledgement) or NACK (Negative Acknowledgement). An ACK is a positive feedback mechanism; the receiver replies to the sender after receiving and accurately demodulating the data. A NACK is a negative feedback mechanism; the receiver only notifies the sender if it has not accurately demodulated and received the data. Base station 104 can count the number of NACKs sent by user terminal 102 by statistically analyzing the received NACKs. Simultaneously, based on the sum of the expected ACK and NACK counts, base station 104 can calculate the proportion of NACKs in the overall response signal, thus obtaining the probability of a negative ACK.

[0094] When the probability of a negative response is high, it indicates that user terminal 102 has repeatedly failed to accurately demodulate the data sent by base station 104. This suggests that the current network's wireless performance is poor and the compression model needs to be updated. Base station 104 will then output a model update command. Specifically, if the first threshold is set to 85% and the second threshold is set to 10%, and the currently calculated similarity index is less than 85% while the negative response probability is greater than 10%, base station 104 will output a model update command to user terminal 102. This embodiment of the application uses a combination of negative response probabilities to determine whether the original compression model has failed, resulting in higher accuracy. In some other embodiments, the compression model can also be determined to be failed by calculating the proportion of positive ACKs to the response signal; when this proportion is less than a certain threshold.

[0095] In one embodiment, before the step of compressing the uncompressed Channel State Information (CSI) report according to the trained compression model, the compression model update method further includes: obtaining target triggering mode information output by the base station; wherein, the target triggering mode information includes periodic triggering information, the target triggering mode information is obtained by the base station based on the model monitoring triggering mode information, and the model monitoring triggering mode information is configured by the network management system.

[0096] Specifically, base station 104 first obtains the model monitoring triggering method information configured by the network management system. This triggering method is used to notify base station 104 to select the triggering method for user terminal 102 to report CSI. The target triggering method information also includes event triggering information. After receiving the model monitoring triggering method information, base station 104 determines whether the conditions for periodic triggering or event triggering are met. If the conditions for periodic triggering are met, the target triggering method information is determined to be periodic triggering information; if the conditions for event triggering are met, the target triggering method information is determined to be event triggering information. When the target triggering method information is determined to be event triggering information, a model update request will only be triggered when a specific event meets the conditions, effectively reducing the overhead of uplink feedback resources in the periodic triggering method.

[0097] After determining the target triggering method information, base station 104 sends the target triggering method information to user terminal 102 via RRC signaling. Upon receiving the corresponding target triggering method information, user terminal 102 changes its own compressed model monitoring event triggering method. Specifically, when the target triggering method information is periodic triggering, user terminal 102 will report compressed channel status information (CSI) and uncompressed channel status information (CSI) at fixed intervals; when the target triggering method information is event triggering, user terminal 102 will start reporting compressed channel status information (CSI) and uncompressed channel status information (CSI) when the corresponding event is detected. Users can configure the model monitoring triggering method information through network management and choose the specific triggering method (periodic triggering or event triggering) for CSI report reporting to meet the update requirements of the compressed model in different scenarios.

[0098] In one embodiment, the periodic triggering information includes: the model monitoring period and the monitoring window duration. After the step of obtaining the target triggering method information output by the base station, the compressed model update method further includes: reporting an uncompressed channel state information (CSI) report during the monitoring window duration; and reporting a compressed channel state information (CSI) report during other times of the model monitoring period.

[0099] Specifically, the monitoring window duration is a period within the model monitoring cycle. User terminal 102 cyclically reports CSI reports according to the model monitoring cycle; that is, it reports uncompressed channel state information (CSI) reports during the monitoring window duration and compressed channel state information (CSI) reports at other times within the model monitoring cycle that are not part of the monitoring window duration. (Refer to...) Figure 5 This is a schematic diagram illustrating the time-domain configuration of CSI reporting during the model monitoring period in one embodiment. The periodic triggering information is sent via RRC signaling, and the following are configured in the CSI-ReportConfig of the RRC signaling: Offset (0, ..., T-1), model monitoring period T (e.g., 4, 5, 8, 10, 16, 20, ..., 160, 320, 640, 1280 slots, etc.), and monitoring window length L (e.g., 100, 200, ..., 800 slots, etc.). The monitoring window length L represents the duration of the monitoring window. User terminal 102 reports uncompressed channel state information (CSI) within the monitoring window duration and compressed channel state information (CSI) within the non-monitoring window duration of the model monitoring period T. It can be understood that the monitoring window can be located at any other position within the model monitoring period T.

[0100] In one embodiment, the model update instruction includes a model deactivation instruction and a model switching instruction. The step of updating the compressed model to an updated model includes: stopping the operation of the compressed model according to the model deactivation instruction; and switching the compressed model to an updated model according to the model switching instruction.

[0101] Specifically, the model update command output by base station 104 includes a model deactivation command and a model switching command. After receiving the model update command, user terminal 102 stops the operation of the current compressed model according to the model deactivation command, and at the same time, user terminal 102 switches the compressed model to the updated model according to the model switching command. It can be understood that if user terminal 102 cannot switch to a suitable updated model, user terminal 102 will automatically fall back to the traditional CSI reporting mode.

[0102] In one embodiment, after updating the compressed model to the updated model, the method further includes: compressing the uncompressed Channel State Information (CSI) report according to the updated model to obtain an updated compressed Channel State Information (CSI) report; and outputting the updated compressed Channel State Information (CSI) report; wherein the base station is used to decompress the updated compressed Channel State Information (CSI) report to obtain a decompressed updated Channel State Information (CSI) report, and to perform resource scheduling based on the decompressed updated Channel State Information (CSI) report.

[0103] Specifically, after base station 104 outputs a model update command, user terminal 102 updates its compressed model to the updated model, then uses the updated model to compress the uncompressed Channel State Information (CSI) report, obtaining an updated compressed CSI report, which is then reported to base station 104. Base station 104 can then obtain the updated compressed CSI report obtained by user terminal 102 using the updated model. When base station 104 notifies user terminal 102 to update its compressed model to the updated model, the compressed model stored internally by base station 104 is also updated accordingly; that is, the compressed models in base station 104 and user terminal 102 are the same. After receiving the updated compressed CSI report, base station 104 decompresses the updated compressed CSI report using the updated model, obtaining a decompressed updated CSI report. When switching from a compressed model to an updated model, base station 104 selects the AI / ML compression model that best suits the current wireless scenario, which can better maintain the wireless performance of the existing network. Therefore, base station 104 can directly perform resource scheduling based on the decompressed updated Channel State Information (CSI) report, achieving better wireless performance in the current wireless scenario. It is understood that in some other embodiments, the similarity between the decompressed and uncompressed CSI reports can be used to determine whether the update model meets the needs of the current wireless scenario, and a new update model can be continuously applied until the obtained similarity index meets the requirements.

[0104] In one embodiment, the base station is used to obtain a scheduling time interval; wherein, the scheduling time interval is the time between obtaining the decompressed updated channel state information (CSI) report and the scheduler using the decompressed updated CSI report; the base station is used to obtain a predicted channel state information (CSI) report based on the decompressed updated CSI report when the scheduling time interval is greater than a third threshold, and to perform resource scheduling based on the predicted CSI report.

[0105] Specifically, since the wireless scenario changes dynamically over time, if the base station 104 calculates and then does not use the decompressed updated Channel State Information (CSI) report for an extended period, the CSI report may become outdated and no longer reflect the current wireless scenario. Therefore, the base station 104 first needs to obtain the time between obtaining the decompressed updated CSI report and the scheduler's use of it, thus determining the scheduling time interval. After obtaining the scheduling time interval, it compares it with a third threshold. For example, with the third threshold set to 4 milliseconds, when the scheduling time interval is greater than the third threshold, the base station 104 inputs the decompressed updated CSI report into the update model for prediction, obtaining a predicted CSI report after the corresponding time interval. When the scheduling time interval is less than or equal to the third threshold, the base station 104 performs resource scheduling based on the decompressed updated CSI report. Obtaining the predicted CSI report through the base station 104 and then performing resource scheduling based on it effectively improves resource scheduling efficiency.

[0106] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0107] In one embodiment, such as Figure 10 As shown, a compressed model update device is provided, applied to base station 104. The device includes: a data acquisition module 310, a decompression module 320, a similarity index acquisition module 330, and an update instruction output module 340, wherein:

[0108] The data acquisition module 310 is used to acquire the compressed channel state information (CSI) report and the uncompressed channel state information (CSI) report reported by the user terminal; wherein the compressed channel state information (CSI) report is obtained by the user terminal through the processing of the trained compression model.

[0109] The decompression module 320 is used to decompress the compressed Channel State Information (CSI) report to obtain a decompressed Channel State Information (CSI) report.

[0110] The similarity index acquisition module 330 is used to acquire the similarity between the decompressed channel state information (CSI) report and the uncompressed channel state information (CSI) report to obtain the similarity index.

[0111] The update instruction output module 340 is used to output a model update instruction if the similarity index is less than or equal to the first threshold; wherein, the model update instruction is used to instruct the user terminal to update the compressed model to the updated model.

[0112] Specific limitations regarding the compression model update device implemented from the perspective of base station 104 can be found in the above-described limitations regarding the compression model update method implemented from the perspective of base station 104, and will not be repeated here. Each module in the aforementioned compression model update device implemented from the perspective of base station 104 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module. It should be noted that the module division in this embodiment is illustrative and only represents a logical functional division; other division methods may exist in actual implementation.

[0113] In one embodiment, such as Figure 11 As shown, a compressed model update device is provided, applied to a user terminal 102. The device includes: a compression module 410, a data reporting module 420, and a model update module 430, wherein:

[0114] Compression module 410 is used to compress the uncompressed Channel State Information (CSI) report according to the trained compression model to obtain a compressed Channel State Information (CSI) report.

[0115] The data reporting module 420 is used to report compressed channel status information (CSI) reports and uncompressed channel status information (CSI) reports to the base station.

[0116] The model update module 430 is used to update the compressed model to an updated model if it receives a model update instruction transmitted by the base station.

[0117] The model update instruction is output by the base station when the similarity index is less than or equal to the first threshold. The similarity index is the similarity between the decompressed Channel State Information (CSI) report and the uncompressed CSI report. The decompressed CSI report is obtained by the base station decompressing the compressed CSI report.

[0118] Specific limitations regarding the compression model update device implemented from the user terminal 102's perspective can be found in the above-described limitations regarding the compression model update method implemented from the user terminal 102's perspective, and will not be repeated here. Each module in the aforementioned compression model update device implemented from the user terminal 102's perspective can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module. It should be noted that the module division in this embodiment is illustrative and only represents a logical functional division; other division methods may exist in actual implementation.

[0119] In one embodiment, a compression model update system is provided, including a base station and a user terminal connected to the base station; wherein: the base station is used to perform the steps of the compression model update method implemented from the perspective of the base station; and the user terminal is used to perform the steps of the compression model update method implemented from the perspective of the user terminal.

[0120] The compression model update system of this application is described in detail below with a specific embodiment. Figure 12 The diagram illustrates the signaling interaction between the base station and the user terminal triggered during the base station-side model monitoring cycle, which may include the following steps:

[0121] Step S500: Resource scheduling is performed based on the compressed CSI report. Base station 104 decompresses the current compressed CSI report to obtain specific CSI information (PMI, CQI, and RI) and uses this information for resource scheduling.

[0122] Step S501, Configure periodic trigger information. Base station 104 configures periodic trigger information for user terminal 102 through CSI-ReportConfig in RRCReconfiguration message. The periodic trigger information includes: model monitoring period and monitoring window duration.

[0123] Step S502, Periodically Compress Channel State Information (CSI) Report. User terminal 102 reports periodically compressed Channel State Information (CSI) reports, including compressed PMI, CQI, and RI.

[0124] Step S503, Periodic Uncompressed Channel State Information (CSI) Report. User terminal 102 reports periodic uncompressed channel state information (CSI) reports, including: raw PMI, CQI, and RI.

[0125] Step S504: Calculate the similarity index and the negative response probability of downlink service feedback. Base station 104 calculates the similarity index based on the received periodic compressed channel state information (CSI) reports and periodic uncompressed channel state information (CSI) reports, and obtains the negative response probability based on downlink service feedback.

[0126] In step S505, the similarity index is less than or equal to the first threshold, and the negative response probability is greater than the second threshold. Base station 104 determines whether the similarity index is less than or equal to the first threshold (e.g., 85%) and whether the negative response probability is greater than the second threshold (e.g., 10%). If the decision conditions are met, proceed to step S506.

[0127] Step S506: Deactivate the compressed model and switch to the updated model. Base station 104 notifies user terminal 102 via RRC signaling to deactivate the compressed model and switch to the updated model. Optionally, if no suitable updated model is available, it falls back to the traditional CSI reporting mode.

[0128] Step S507: Perform measurements based on the updated model. User terminal 102 performs measurements based on the updated model notified by base station 104.

[0129] Step S508: Report the updated Compressed Channel Status Information (CSI) report. User terminal 102 reports the updated Compressed Channel Status Information (CSI) report based on the updated model.

[0130] Step S509: Decompress the updated compressed Channel State Information (CSI) report and perform resource scheduling based on the decompressed CSI report. After receiving the updated compressed CSI report, base station 104 first decompresses the updated compressed CSI report to obtain the decompressed updated CSI report, and then performs resource scheduling based on the PMI, CQI, and RI information in the decompressed updated CSI report. Optionally, model prediction can be initiated after obtaining the decompressed updated CSI report to further improve measurement accuracy.

[0131] Step S510: Data transmission based on resource scheduling. Base station 104 performs resource scheduling for downlink services by decompressing and updating the PMI, CQI, and RI information in the Channel State Information (CSI) report, and sends the downlink data to user terminal 102 on the scheduled resources.

[0132] The compressed model update system of this application is described in detail below with reference to another specific embodiment. Figure 13 The diagram illustrates the signaling interaction between the base station and the user terminal triggered by a base station-side model monitoring event. Specifically, it may include the following steps:

[0133] Step S600: Data transmission is scheduled based on the compressed CSI report. Base station 104 decompresses the current compressed CSI report to obtain specific CSI information (PMI, CQI, and RI) for resource scheduling.

[0134] Step S601, Event Trigger Information Configuration. Base station 104 adds compression model switching events and trigger parameters to the RRCReconfiguration message. Switching events include events such as downlink service throughput falling below a certain threshold and downlink service bit error rate exceeding a certain threshold. Trigger parameters include specific thresholds for service throughput (e.g., tens to hundreds of Mbps) and specific thresholds for downlink service bit error rate (e.g., 10%).

[0135] Step S602: Based on the received compression model switching event and trigger parameters, determine whether the model switching conditions are met. User terminal 102 determines whether the current model switching conditions are met based on the received model switching event and trigger parameters. If met, proceed to step S603.

[0136] Step S603, Compression model switching request. User terminal 102 initiates a compression model switching request to base station 104.

[0137] Step S604: Whether to perform model switching and select the updated model. Base station 104 determines whether to perform model switching based on the compressed model switching request reported by user terminal 102; if model switching is required, it selects the updated model to be switched to and proceeds to step S605.

[0138] Step S605: Deactivate the compressed model and switch to the updated model. Base station 104 notifies user terminal 102 via RRC signaling to deactivate the compressed model and switch to the updated model selected by base station 104. Optionally, if no suitable updated model is available, it falls back to the traditional CSI reporting mode.

[0139] Step S606: Measurement is performed based on the updated model. User terminal 102 performs measurements based on the updated model notified by base station 104.

[0140] Step S607: Report the updated Compressed Channel State Information (CSI) report based on the updated model. User terminal 102 reports the updated Compressed Channel State Information (CSI) report obtained based on the updated model.

[0141] Step S608: Decompress the updated compressed Channel State Information (CSI) report and perform resource scheduling. After receiving the updated compressed CSI report, base station 104 first decompresses the updated compressed CSI report to obtain the decompressed updated CSI report, and then performs resource scheduling based on the PMI, CQI, and RI information in the decompressed updated CSI report. Optionally, model prediction can be initiated after obtaining the decompressed updated CSI report to further improve measurement accuracy.

[0142] Step S609: Data is transmitted according to resource scheduling. Base station 104 performs resource scheduling for downlink services by decompressing and updating the PMI, CQI, and RI information in the Channel State Information (CSI) report, and sends the downlink data to user terminal 102 on the scheduled resources.

[0143] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0144] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0145] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A compressed model update method, characterized by, The method is applied to a base station, and the method comprises: obtaining a compressed channel state information (CSI) report and a non-compressed channel state information (CSI) report reported by a user terminal; wherein the compressed channel state information (CSI) report is obtained by a trained compression model through processing of the user terminal; decompressing the compressed channel state information (CSI) report to obtain a decompressed channel state information (CSI) report; obtaining a similarity between the decompressed channel state information (CSI) report and the non-compressed channel state information (CSI) report to obtain a similarity index; if the similarity index is less than or equal to a first threshold value, outputting a model update instruction; wherein the model update instruction is used to instruct the user terminal to update the compression model to an updated model.

2. The method of claim 1, wherein, The method further comprises: counting a received negative acknowledgement (NACK) to obtain a negative acknowledgement probability; wherein the negative acknowledgement probability is a proportion of negative acknowledgements (NACK) to acknowledgement signals; if the negative acknowledgement probability is greater than a second threshold value, outputting the model update instruction.

3. The method of claim 1, wherein, Before the step of obtaining the compressed channel state information (CSI) report and the non-compressed channel state information (CSI) report reported by the user terminal, the method further comprises: obtaining model monitoring trigger mode information configured by a network management system; obtaining target trigger mode information according to the model monitoring trigger mode information; wherein the target trigger mode information comprises periodic trigger information; outputting the target trigger mode information.

4. The method of claim 3, wherein, The periodic trigger information comprises a model monitoring period and a monitoring window duration, and the user terminal is used to report the non-compressed channel state information (CSI) report within the monitoring window duration and report the compressed channel state information (CSI) report at other times of the model monitoring period.

5. The method of claim 1, wherein, The model update instruction comprises a model deactivation instruction and a model switching instruction, and the user terminal is used to stop the compression model from running according to the model deactivation instruction, and the user terminal is used to switch the compression model to an updated model according to the model switching instruction.

6. The method according to any one of claims 1 to 5, characterized in that, After the step of outputting the model update instruction, the method further comprises: obtaining an updated compressed channel state information (CSI) report reported by the user terminal; wherein the updated compressed channel state information (CSI) report is obtained by processing the non-compressed channel state information (CSI) report by the user terminal using the updated model; decompressing the updated compressed channel state information (CSI) report to obtain a decompressed updated channel state information (CSI) report; performing resource scheduling according to the decompressed updated channel state information (CSI) report.

7. The method of claim 6, wherein, The step of performing resource scheduling according to the decompressed updated channel state information (CSI) report comprises: obtaining a scheduling time interval; wherein the scheduling time interval is a time between obtaining the decompressed updated channel state information (CSI) report by the base station and using the decompressed updated channel state information (CSI) report by a scheduler; if the scheduling time interval is greater than a third threshold value, obtaining a predicted channel state information (CSI) report according to the decompressed updated channel state information (CSI) report; Resource scheduling is performed according to the predicted channel state information (CSI) report.

8. A compression model update method characterized by, The method is applied to a user terminal, and the method comprises: compressing a non-compressed channel state information (CSI) report according to a trained compression model to obtain a compressed channel state information (CSI) report; reporting the compressed channel state information (CSI) report and the non-compressed channel state information (CSI) report to a base station; if a model update instruction transmitted by the base station is received, updating the compression model to an updated model; wherein the model update instruction is output by the base station when a similarity index is less than or equal to a first threshold, the similarity index is a similarity between a decompressed channel state information (CSI) report and the non-compressed channel state information (CSI) report, and the decompressed channel state information (CSI) report is obtained by decompressing the compressed channel state information (CSI) report by the base station.

9. The method of claim 8, wherein, The model update instruction is output by the base station when a negative acknowledgement probability is greater than a second threshold, the negative acknowledgement probability is obtained by the base station by counting a negative acknowledgement (NACK), and the negative acknowledgement probability is a proportion of negative acknowledgements (NACK) to acknowledgement signals.

10. The method of claim 8, wherein, Before the step of compressing a non-compressed channel state information (CSI) report according to a trained compression model, the method further comprises: obtaining target triggering mode information output by the base station; wherein the target triggering mode information comprises periodic triggering information, and the target triggering mode information is obtained by the base station according to model monitoring triggering mode information, and the model monitoring triggering mode information is obtained by network management.

11. The method of claim 10, wherein, The periodic triggering information comprises a model monitoring period and a monitoring window duration, and after the step of obtaining the target triggering mode information output by the base station, the method further comprises: reporting the non-compressed channel state information (CSI) report within the monitoring window duration; reporting the compressed channel state information (CSI) report at other times of the model monitoring period.

12. The method of claim 8, wherein, The model update instruction comprises a model deactivation instruction and a model switching instruction, and the step of updating the compression model to an updated model comprises: stopping the compression model from running according to the model deactivation instruction; switching the compression model to an updated model according to the model switching instruction.

13. The method according to any one of claims 8 to 12, characterized in that, After the step of updating the compression model to an updated model, the method further comprises: compressing a non-compressed channel state information (CSI) report according to the updated model to obtain an updated compressed channel state information (CSI) report; outputting the updated compressed channel state information (CSI) report; wherein the base station is configured to decompress the updated compressed channel state information (CSI) report to obtain a decompressed updated channel state information (CSI) report, and perform resource scheduling according to the decompressed updated channel state information (CSI) report.

14. The method of claim 13, wherein, The base station is configured to obtain a scheduling time interval, wherein the scheduling time interval is a time interval between obtaining the decompressed update channel state information (CSI) report and using the decompressed update channel state information (CSI) report for scheduling; and the base station is configured to obtain a predicted channel state information (CSI) report according to the decompressed update channel state information (CSI) report when the scheduling time interval is greater than a third threshold, and perform resource scheduling according to the predicted channel state information (CSI) report.

15. A compressed model update apparatus characterized by comprising: The device is applied to a base station, and the device comprises: a data obtaining module configured to obtain a compressed channel state information (CSI) report and a non-compressed channel state information (CSI) report reported by a user terminal, wherein the compressed channel state information (CSI) report is obtained by a trained compression model and the user terminal; a decompression module configured to decompress the compressed channel state information (CSI) report to obtain a decompressed channel state information (CSI) report; a similarity index obtaining module configured to obtain a similarity between the decompressed channel state information (CSI) report and the non-compressed channel state information (CSI) report to obtain a similarity index; an update instruction output module configured to output a model update instruction if the similarity index is less than or equal to a first threshold, wherein the model update instruction is used to instruct the user terminal to update the compression model to an updated model.

16. A compressed model update apparatus characterized by comprising: The device is applied to a user terminal, and the device comprises: a compression module configured to compress a non-compressed channel state information (CSI) report according to a trained compression model to obtain a compressed channel state information (CSI) report; a data reporting module configured to report the compressed channel state information (CSI) report and the non-compressed channel state information (CSI) report to a base station; a model updating module configured to update the compression model to an updated model if a model update instruction transmitted by the base station is received; wherein the model update instruction is output by the base station when a similarity index is less than or equal to a first threshold, the similarity index is a similarity between a decompressed channel state information (CSI) report and the non-compressed channel state information (CSI) report, and the decompressed channel state information (CSI) report is obtained by decompressing the compressed channel state information (CSI) report by the base station.

17. A compressed model update system, comprising: The device comprises a base station and a user terminal connected to the base station, wherein: the base station is configured to perform steps of the method in any one of claims 1 to 7; the user terminal is configured to perform steps of the method in any one of claims 8 to 14.

18. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is configured to implement steps of the method in any one of claims 1 to 14 when executed by a processor.

Images