Communication device and method for packet true value reporting for artificial intelligence / machine learning
By determining the CSI-RS measurement timing in the user equipment (UE) and reporting the real AI/ML value in packet form, the problem of large overhead of reporting the real value to the base station is solved, and the effect of reducing system overhead and improving communication performance is achieved.
Patent Information
- Application Number
- CN202280097436.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2022-08-12
- Publication Date
- 2025-02-11
Smart Images
Figure CN119422333A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of wireless communication systems, and more specifically, to communication devices and methods for grouped true value reporting of artificial intelligence (AI) / machine learning (ML). For example, the present application is related to the new study item description (SID) for artificial intelligence / machine learning for new radio (NR) air interface in Release 18 established by the 94e plenary meeting of the 3rd generation partnership project (3GPP) radio access network (RAN) in December 2022. In particular, the present application relates to enhanced channel state information (CSI) report feedback, beam management and positioning, wherein during data collection and model monitoring, the user equipment (UE) reports the true value to the base station (such as gNB). Background Art
[0002] Artificial Intelligence (AI) / Machine Learning (ML) has been applied in 3GPP RAN1, which decided to study multiple use cases, namely channel state information (CSI) feedback enhancement, beam management and positioning. As stated in 3GPP's new Study Item Description (SID), although specific AI / ML algorithms and models may be studied for evaluation purposes, they are implementation-specific and are not expected to be standardized. Currently, packet reporting of layer 1 reference signal received power (L1-RSRP) is supported in RAN1, but it has not yet been applied to ground truth reporting. In addition, reporting of the ground truth by the UE to the gNB consumes a lot of resources and signaling, and incurs scheduling overhead. In addition, the quantization method of the ground truth also needs further study.
[0003] Therefore, a communication device and method for artificial intelligence (AI) / machine learning (ML) true value group reporting are needed to solve the problems in the prior art, save the overhead of UE reporting true values to gNB, adopt a new quantization method to reduce the size of the report content, reduce system overhead, and provide good communication performance and / or high reliability. Summary of the invention
[0004] The purpose of this application is to propose a communication device and method for artificial intelligence (AI) / machine learning (ML) true value group reporting to solve the problems in the prior art, save the overhead of UE reporting true values to gNB, adopt a new quantization method to reduce the size of the report content, reduce system overhead, and provide good communication performance and / or high reliability.
[0005] In a first aspect of the present application, a method for artificial intelligence (AI) / machine learning (ML) true value group reporting performed by a user equipment (UE) includes: the UE determines a channel state information reference signal (CSI-RS) measurement timing for generating an AI / ML true value group report, and reports the AI / ML group true value to a base station in a periodic, semi-persistent and / or non-periodic manner on an uplink channel.
[0006] In a second aspect of the present application, a user equipment (UE) includes a memory, a transceiver, and a processor connected to the memory and the transceiver. The processor is configured to execute the above method.
[0007] In a third aspect of the present application, a method for group reporting of artificial intelligence (AI) / machine learning (ML) true values performed by a base station includes: configuring a CSI-RS measurement timing of a UE so that the UE generates a report of a group true value of the AI / ML, and controlling the UE to report the group true value of the AI / ML on an uplink channel in a periodic, semi-continuous and / or non-periodic manner.
[0008] In a fourth aspect of the present application, a base station includes a memory, a transceiver, and a processor connected to the memory and the transceiver. The processor is configured to execute the above method.
[0009] In a fifth aspect of the present application, a non-transitory machine-readable storage medium stores instructions thereon, and when a computer executes the instructions, the instructions cause the computer to execute the above method.
[0010] In a sixth aspect of the present application, a chip includes a processor configured to call and run a computer program stored in a memory so that a device equipped with the chip executes the above method.
[0011] In a seventh aspect of the present application, a computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the above method.
[0012] In an eighth aspect of the present application, a computer program product comprises a computer program, wherein the computer program enables a computer to execute the above method.
[0013] In a ninth aspect of the present application, a computer program is provided, wherein the computer program enables a computer to execute the above method. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the embodiments of the present application or related technologies, the following briefly introduces the drawings to be described in the embodiments. Obviously, the drawings are only some embodiments of the present application, and ordinary technicians in this field can obtain other drawings based on these drawings without paying any price.
[0015] Figure 1 is a schematic diagram of an example of a basic autoencoder model for enhancing CSI feedback according to an embodiment of the present application.
[0016] Figure 2 It is a block diagram of communication between one or more user equipments (UE) and a base station (such as a gNB) in a communication network system according to an embodiment of the present application.
[0017] Figure 3 This is a wireless communication flow chart for artificial intelligence (AI) / machine learning (ML) true value group reporting performed by a UE according to an embodiment of the present application.
[0018] Figure 4 This is a wireless communication flow chart for artificial intelligence (AI) / machine learning (ML) true value group reporting performed by a base station according to an embodiment of the present application.
[0019] Figure 5 is a schematic diagram of an example of a basic autoencoder model for enhancing CSI feedback according to an embodiment of the present application.
[0020] Figure 6 It is a schematic diagram of an example of independent quantification of each item according to an embodiment of the present application.
[0021] Figure 7 It is a schematic diagram of an example of differential quantization between entries according to an embodiment of the present application.
[0022] Figure 8 is a schematic diagram of the difference between vectors v2 and v1 according to an embodiment of the present application.
[0023] Fig. 9 It is a schematic diagram of an adaptive grouping report example according to an embodiment of the present application.
[0024] Fig.10 It is a flowchart of an example of configuring a group report according to an embodiment of the present application.
[0025] Fig.11 is a schematic diagram of selecting a vector v(n / 2) as a reference vector according to an embodiment of the present application.
[0026] Fig.12 It is a schematic diagram of a CSI measurement opportunity selected after a group true value report indication according to an embodiment of the present application.
[0027] Fig.13 It is a schematic diagram of a CSI measurement opportunity after a group true value report indication and processing delay selection according to an embodiment of the present application.
[0028] Fig.14 It is a schematic diagram of a CSI measurement opportunity before a group true value reporting opportunity and a processing delay selection according to an embodiment of the present application.
[0029] Fig.15 is a block diagram of a wireless communication system according to an embodiment of the present application. DETAILED DESCRIPTION
[0030] The technical content, structural features, objectives and effects of the present application are described in detail below in conjunction with the accompanying drawings. Specifically, the terms in the embodiments of the present application are only used to describe the purpose of specific embodiments, and are not intended to limit the present application.
[0031] Introducing Artificial Intelligence (AI) / Machine Learning (ML) in the Physical Layer (PHY) and Medium Access Control Layer (MAC) to enhance system performance. In 3GPP RAN1, it was decided to study several use cases, namely channel state information (CSI) feedback compression, beam management and positioning. The machine learning model can be trained online or offline.
[0032] For machine learning models, both online and offline training, as well as model monitoring, ground truth data is required. Usually, ground truth data is reported from user equipment (UE) to base stations (such as gNB) with high precision. In order to maintain the accuracy of the ground truth and reflect the real characteristics, the reported ground truth needs to have high precision, but this consumes a lot of resources and signaling, and causes scheduling overhead, which burdens the normal operation of the UE. Therefore, it is necessary to reduce the reporting overhead of the ground truth. A natural way is to buffer or store the ground truth collected by the UE and report it in batches / groups. The reporting frequency of the ground truth can be reduced. At the same time, it is also necessary to study the quantization method of the ground truth. On the one hand, the quantization should have high precision, which requires more bits; on the other hand, the overall size of the report content should be reduced as much as possible.
[0033] Some embodiments of the present application discuss use cases for CSI feedback enhancement. Figure 1 This is a schematic diagram of an example of a basic autoencoder model for enhancing CSI feedback according to an embodiment of the present application. Figure 1 It is shown that in some embodiments, the basic model of the autoencoder is as follows: the encoder compresses the original CSI-RS value (referred to as original CSI) / maximum eigenvector and reports its output to the gNB, which will decompress it. The new CSI report is a report containing CSI feedback enhanced by the AI / ML model.
[0034] Figure 2 A block diagram is shown in which one or more user equipments (UEs) 10 and a base station (such as a gNB) 20 communicate in a communication network system 40 according to one embodiment of the present application. The communication network system 40 includes one or more UEs 10 and a base station 20 (for example, a first base station or a second base station). One or more UEs 10 may include a memory 12, a transceiver 13, and a processor 11 coupled to the memory 12 and the transceiver 13. The base station 20 may include a memory 22, a transceiver 23, and a processor 21 coupled to the memory 22 and the transceiver 23. The processor 11 or 21 may be configured to implement the functions, procedures, and / or methods described herein. The layers of the wireless interface protocol may be implemented in the processor 11 or 21. The memory 12 or 22 is operatively coupled to the processor 11 or 21 for storing a variety of information to operate the processor 11 or 21. The transceiver 13 or 23 is operatively coupled to the processor 11 or 21 for sending and / or receiving wireless signals.
[0035] The processor 11 or 21 may include an application specific integrated circuit (ASIC), other chipsets, logic circuits and / or data processing devices. The memory 12 or 22 may include a read-only memory (ROM), a random access memory (RAM), flash memory, a memory card, a storage medium and / or other storage devices. The transceiver 13 or 23 may include a baseband circuit to process radio frequency signals. When the embodiment is implemented by software, the technology described herein may be implemented by a module (e.g., a program, a function, etc.) that performs the function. The module may be stored in the memory 12 or 22 and executed by the processor 11 or 21. The memory 12 or 22 may be implemented within the processor 11 or 21, or may be implemented outside the processor, in which case it may be communicatively coupled to the processor 11 or 21 in various known ways.
[0036] In some embodiments, the processor 11 is configured to determine the CSI-RS timing for reporting the true value of the AI / ML group, and the transceiver 13 is configured to report the true value of the AI / ML group to the base station 20 in a periodic, semi-persistent and / or aperiodic manner on the uplink channel. This can solve the problems in the prior art, reduce the true value reporting overhead from the UE to the gNB, reduce the size of the report content under the new quantization method, reduce system overhead, and provide good communication performance and / or high reliability.
[0037] In some embodiments, the processor 21 is configured to configure the CSI-RS timing for AI / ML group true value reporting to the UE 10, and control the UE 10 to report the AI / ML group true value on the uplink channel in a periodic, semi-persistent and / or aperiodic manner. This can solve the problems in the prior art, reduce the true value reporting overhead from the UE to the gNB, reduce the size of the report content under the new quantization method, reduce the system overhead, and provide good communication performance and / or high reliability.
[0038] Figure 3 The present invention describes a wireless communication method 300 for reporting the true value of AI / ML packets performed by a UE according to an embodiment of the present application. In some embodiments, the method 300 includes: step 302, the UE determines the CSI-RS measurement timing for generating the true value report of the AI / ML packets; step 304, the UE reports the true value of the AI / ML packets to the base station in a periodic, semi-persistent and / or non-periodic manner on the uplink channel. This can solve the problems in the prior art, reduce the overhead of the true value report, optimize the allocation of system resources, and provide high reliability and good communication performance.
[0039] Figure 4 The present invention describes a wireless communication method 400 for reporting the true value of AI / ML packets performed by a base station according to an embodiment of the present application. In some embodiments, the method 400 includes: step 402, the base station configures the CSI-RS measurement opportunity to the UE to generate the true value report of the AI / ML packets; step 404, the base station controls the UE to report the true value of the AI / ML packets in a periodic, semi-continuous and / or non-periodic manner on the uplink channel. This can solve the problems in the prior art, reduce the overhead of the true value report, optimize the allocation of system resources, and provide high reliability and good communication performance.
[0040] In some embodiments, the packet true value is used to enhance at least one of CSI feedback, predictive beam management, and positioning. In some embodiments, the collection of AI / ML packet true values can be activated or deactivated by at least one of the following signaling: radio resource control (RRC) signaling, media access control control element (MAC-CE), and downlink control information (DCI). In some embodiments, the packet true value of AI / ML can preconfigure the packet size through RRC signaling and activate it through MAC-CE or DCI. In some embodiments, the periodic, semi-persistent, and / or non-periodic manner of AI / ML packet true value reporting is associated with UE capabilities.
[0041] In some embodiments, the uplink channel for uploading the packet true value report includes a physical uplink control channel (PUCCH) or a physical uplink shared channel (PUSCH). In some embodiments, the PUCCH includes PUCCH format 3 and / or PUCCH format 4. In some embodiments, if the packet true value of the AI / ML is not allowed to be transmitted via the PUCCH, the UE will not report the packet true value on the PUCCH.
[0042] In some embodiments, the CSI measurement opportunity for generating the AI / ML packet true value report is selected after the packet true value indication. In some embodiments, the CSI measurement opportunity for reporting the AI / ML packet true value is selected after the packet true value indication plus the UE's processing delay. In some embodiments, there is UE processing time before the reporting opportunity of the AI / ML packet true value. In some embodiments, the CSI measurement opportunity is selected in a sparse manner for reporting the AI / ML packet true value. For example, an even or odd CSI measurement opportunity can be selected to generate a packet true value report.
[0043] In some embodiments, if sufficient true value entries are not collected from the CSI measurement opportunity for reporting the AI / ML group true value, the UE will not report the group true value of AI / ML. In some embodiments, if sufficient true value entries are not collected from the CSI measurement opportunity, the UE saves the group true value of AI / ML in a buffer and reports it after collecting sufficient true value entries at the next reporting opportunity. In some embodiments, the original true value is in the form of a vector, multiple vectors are packaged together as multiple entries to form a group true value, and the vectors in the group true value are quantized separately. In some embodiments, when the group true value of AI / ML is used to enhance CSI feedback, the group true value is the original CSI-RS value or the maximum eigenvector.
[0044] In some embodiments, when the grouped true value of AI / ML is used for predictive beam management, the grouped true value contains multiple measured L1 reference signal received powers (Layer 1-Reference Signal Received Power, L1-RSRP), and each reported L1-RSRP is paired with a beam index or location information to form each entry. In some embodiments, when the grouped true value of AI / ML is used for positioning, the grouped true value is location information containing [coordinates, RS], [(x, y), RS] or [label, RS], where the label is a mark or label of a reference node, so that each entry contains amplitude quantization and phase quantization of RS.
[0045] In some embodiments, a true value or a portion of a true value is used as a vector, and the entries in the grouped true values of the AI / ML are quantized based on the reference entries and are quantized in a differential manner. In some embodiments, when the grouped true values of the AI / ML are used to enhance CSI feedback, the grouped true values are the original CSI-RS values. In some embodiments, when the grouped true values of the AI / ML are used for predictive beam management, the grouped true values are the measured L1-RSRP, and each reported L1-RSRP is paired with a beam index or position information to form each entry, and is quantized in a differential manner according to the maximum L1-RSRP. In some embodiments, when the grouped true values of the AI / ML are used for positioning, the grouped true values are position reference signals (PRS), and in each entry, the amplitude of the element is quantized with a configured step value or a fixed step value, and the phase is quantized with one of {8PSK, QPSK, BPSK}.
[0046] In some embodiments, after the base station configures the grouped true value of AI / ML, if the UE capability supports grouped true value reporting, the UE will report the grouped true value of AI / ML; otherwise, if the UE capability does not support grouped true value reporting, the UE will report a single true value as the input of AI / ML at the reporting time. In some embodiments, it is the default configuration for the UE to report a single true value as the input of AI / ML. In some embodiments, the reference vector in the grouped true value of AI / ML can be one of the following: the first vector in time, the last vector in time, the middle vector in time, the vector with the maximum Euclidean distance, the vector closest to the median Euclidean distance value, or a vector indicated by the base station.
[0047] In some embodiments, the number of entries in the grouped true value of AI / ML is configured by the base station via RRC signaling, MAC-CE or DCI. In some embodiments, the UE reports the grouped true value of AI / ML to the base station based on the UE capability.
[0048] Figure 5 It is a schematic diagram of an example of a basic autoencoder model for enhancing CSI feedback according to an embodiment of the present application.
[0049] Figure 5 It is shown that in some embodiments, there are two designs of the autoencoder model depending on the input. One input is the original CSI (original CSI-RS value), and the other input is the eigenvector corresponding to the maximum eigenvalue after the channel matrix is decomposed. In some embodiments, the eigenvector is called the maximum eigenvector. Examples for multiple CSI sub-use cases (with / without CSI prediction) can be added to this description.
[0050] Example:
[0051] Embodiment 1:
[0052] Figure 6 It is a schematic diagram illustrating the independent quantification of each item according to an embodiment of the present application. Figure 6 It is shown that in some embodiments, the real value can be in the form of a vector, such as v. In some examples, the vector v includes several elements. As an example, the entries (vectors) in the grouped real value are quantized separately, and multiple entries are packed together.
[0053] In some examples, for enhanced CSI feedback, the true value can be the original CSI-RS value or the largest eigenvector. These can be treated as complex vectors and quantized. For example, for each vector, the element with the largest amplitude is selected, and the other elements in the entry are quantized according to this largest element. The amplitude and phase of each element can be quantized separately. The composition of an entry includes: the index of the largest element, the indication (index) of the non-zero elements, the amplitude of the non-zero elements quantized according to the maximum amplitude, and the quantized phase of these non-zero elements.
[0054] In some examples, for predicted beam management, the true value can be the measured L1-RSRP. Multiple measured values can be quantized according to the value with the largest L1-RSRP. The difference is that each reported L1-RSRP is paired with a beam index. The beam index can be the index of the UE receiving beam, or the L1-RSRP is paired with location information. This L1-RSRP pairing constitutes an entry.
[0055] In some examples, for positioning, the real value can be location information, such as [coordinate, RS], [(x, y), RS], or [label, RS]. The label is a mark or label of the reference node. The composition of an entry includes the amplitude quantization and phase quantization of RS.
[0056] Embodiment 2:
[0057] Figure 7It is a schematic diagram illustrating differential quantization across entries according to an embodiment of the present application. Figure 7 It is shown that, in some embodiments, if the true value or part of the true value can be described by a vector v, the entries in the group report can be quantized based on the reference entries and quantized in a differential manner. Figure 8 is a schematic diagram illustrating the difference between vectors v2 and v1. Figure 8 It is shown that, in some embodiments, the quantization of vector v2-v1 adopts a lower level or a coarse quantization level to save quantization bits. For a slow fading channel, the quantization of vector v2-v1 saves more bits than directly quantizing v2. Figure 8 The difference between vectors v1 and v2 is shown in , which is a vector with smaller magnitude.
[0058] In some examples, v1 is a complex vector. For enhanced CSI feedback, v1 is the original CSI-RS value (vector); for positioning, v1 is the position reference signal (RS). The element amplitude is quantized with 4 bits, and the phase is quantized with one of the following: {8PSK, QPSK, BPSK}. For example, the range from 0 to the maximum value is quantized into 16 levels using 4 bits. In addition, a bitmap is used to indicate non-zero amplitude elements, v2-v1 is quantized with 2 bits, and the phase is quantized with one of the following: {8PSK, QPSK, BPSK}.
[0059] In some examples, the quantization of the magnitude of the elements in the vector v2-v1 follows the following table. "00" means that the magnitude of an element in the vector v2-v1 is the magnitude of the corresponding element in v1. v1=[v1,v2,…,v m ].if The magnitude can be quantified as It is represented as "00".
[0060] sheet:
[0061]
[0062] Alternatively, a fixed step size value can be used for quantization. For example, if If the amplitude is less than a step size (value) of |v2|, then The amplitude difference is quantized to "00".
[0063] In some examples, the grouped L1-RSRP reports for predictive beam management are differentially quantized based on the maximum L1-RSRP. Other information, such as location information or beam index in the entry, is still paired with the L1-RSRP. In this case, an entry consists of two parts: [beam index, L1-RSRP] or [location information, L1-RSRP].
[0064] Adaptive Grouping Report:
[0065] Whether to adopt embodiment 1 or embodiment 2 is configured by the gNB. Fig. 9 An example of adaptive group reporting is shown. In some embodiments, whether to adopt embodiment 1 or embodiment 2 is determined by the UE. If the difference between the two vectors is large, resulting in too low quantization accuracy, the UE may select embodiment 1. For example, if the measurement result (such as the Euclidean distance) is greater than a threshold, the UE may select embodiment 1; otherwise, embodiment 2 is selected. The threshold is indicated by the gNB. In particular, one bit in the group report can indicate the specific scheme adopted by the UE, for example, "0" represents embodiment 1 and "1" represents embodiment 2.
[0066] Fig.10 An example of a group reporting configuration is shown. In some embodiments, after the gNB configures group reporting, the UE may report the grouped true value if it supports group reporting within its capabilities. Otherwise, if the UE does not support group reporting, the UE may report a single true value in one reporting occasion. In some examples, it is the default configuration for the UE to report a single true value in one reporting occasion. If the gNB is not configured with any reporting method, the UE may also report a single true value in one reporting occasion.
[0067] Fig.11 is a schematic diagram showing an example of selecting a vector v(n / 2) as a reference vector according to an embodiment of the present application. Fig.11 As shown, in some embodiments, the first time vector in the grouped true value report is selected as the reference vector. In other examples, the last time vector in the grouped true value report is selected as the reference vector. In some examples, the middle vector is selected as the reference vector. When n is an even number, the vector v is selected n / 2 As a reference vector; when n is an odd number, select the vector as a reference vector. In some examples, the reference vector is selected as the vector with the largest Euclidean distance and is indicated by a binary value. For example, "01" indicates that the second vector / entry in the group report is the reference vector. In other examples, the vector closest to the median Euclidean distance is selected as the reference vector. In one example, the true value is the original CSI-RS value or the largest eigenvector. The quantization of each entry includes two parts: amplitude and phase. In some examples, the number of entries in the group true value report can be configured by the gNB through RRC signaling, DCI field, or MAC-CE signaling. In some examples, the reference vector in the group report is indicated by the gNB, specifically through RRC signaling, DCI field, or MAC-CE signaling. In some examples, whether the UE can perform group true value reporting depends on its capabilities.
[0068] Reporting of true values of groups:
[0069] The packet true value of enhanced CSI feedback based on machine learning (ML) is defined as a new reporting quantity / type. In some examples, it is reported through RRC signaling, MAC-CE or DCI. The packet true value of predictive beam management is defined as a new reporting quantity / type. In some examples, it is reported through RRC signaling, MAC-CE or DCI. The positioning packet true value based on AI / ML is defined as a new reporting quantity / type. In some examples, it is reported through RRC signaling, MAC-CE or DCI. In some examples, the reporting of the packet true value can be activated or deactivated through RRC signaling, MAC-CE or DCI. For example, the packet size can be preconfigured through RRC signaling and activated through MAC-CE or DCI. In one example, the packet true value report is transmitted on PUSCH in a periodic and / or semi-persistent and / or aperiodic manner. In another example, the packet true value report is transmitted in a periodic and / or semi-persistent and / or aperiodic manner through PUCCH format 3 and / or PUCCH format 4. In another example, the grouped true value reports are transmitted over all PUCCH formats in a periodic and / or semi-persistent and / or aperiodic manner.
[0070] In some examples, transmission or reporting of packet true values on a physical uplink control channel (PUCCH) or at least one PUCCH format is a capability of a user equipment (UE) that can be performed in a periodic, semi-persistent and / or aperiodic manner. For example, the UE reports the capability to the gNB (base station). In one example, the packet true value report is not allowed to be transmitted on the PUCCH, and the UE does not report the packet true value on the PUCCH.
[0071] Determine the CSI measurement timing for collecting ground truth values:
[0072] Fig.12 2 is a schematic diagram illustrating that according to an embodiment of the present application, the CSI measurement opportunity selected after the group true value indication is used for group reporting. These collected true values will be packaged and reported as a group in one reporting opportunity. Fig.12 It is shown that, in some examples, a CSI-RS opportunity after a group true value report indication is selected for collecting true values for group true value reports. Fig.12 An example in shows that three CSI-RS measurement opportunities are considered for collecting true values, and the grouping size for grouping the true values is set to 3.
[0073] Fig.13It is a schematic diagram illustrating that according to an embodiment of the present application, the CSI measurement opportunity selected after the group true value indication plus the processing delay is used for group reporting. Fig.13 It is shown that in some embodiments, when there is a UE processing delay after configuring the group true value, the CSI-RS measurement opportunity can be counted after the UE processing delay. Fig.13 As shown, in some examples, the CSI-RS measurement opportunity immediately following the group true value reporting indication, plus UE processing delay, is selected for group true value reporting. Fig.13 An example in shows that three CSI-RS measurement opportunities are considered for collecting true values, and the grouping size for grouping the true values is set to 3.
[0074] Fig.14 It is a schematic diagram illustrating that according to an embodiment of the present application, a CSI measurement opportunity selected before a group true value reporting opportunity plus a processing delay is used for group reporting. Fig.14 It is shown that in some examples, CSI measurement opportunities are selected in a sparse manner. For example, even / odd CSI measurement opportunities are selected for collecting true values to generate group true value reports. When there is another UE processing time before the group true value reporting opportunity, the CSI-RS measurement opportunity can be counted after the UE processing delay. Fig.14 As shown, in some examples, the CSI-RS measurement opportunity immediately preceding the grouped true value reporting indication, plus UE processing delay, is selected for grouped true value reporting. Fig.14 An example in shows that three CSI-RS measurement opportunities are considered for collecting true values, and the grouping size for grouping the true values is set to 3.
[0075] In some examples, if not enough true value entries are collected at the reporting opportunity of the grouped true value, the grouped true value may not be reported at this reporting opportunity. In some examples, if not enough true value entries are collected at the reporting opportunity of the grouped true value, the grouped true value will be saved in a buffer and reported immediately at the next reporting opportunity until enough true value entries are collected.
[0076] In summary, some embodiments of the present application provide a grouping method for reporting true values. As is known to all, for machine learning, true values are necessary for model training or model monitoring. Frequent reporting of true values may consume a lot of resources and signaling, resulting in a large overhead. In order to reduce this overhead, the present application proposes to report the true values in the form of groups. For example, it is recommended to report the true values in the form of groups for enhanced CSI feedback, predicted beam management and / or positioning. Specifically, the grouping true value reporting method for artificial intelligence / machine learning (AI / ML) performed by the user equipment (UE) includes: the UE determines the channel state information reference signal (CSI-RS) measurement timing for collecting true values to generate grouped true values of AI / ML, and reports the grouped true values of AI / ML to the base station through the uplink channel in a periodic, semi-persistent and / or non-periodic manner. Some embodiments of the present application have at least one of the following inventive effects: 1. The overhead of reporting true values from the UE to the gNB can be saved. 2. The size of the report content can be reduced through a new quantization method.
[0077] Fig.15 7 is a block diagram of an exemplary wireless communication system 700 according to an embodiment of the present application. The embodiments described herein may be implemented into a system using any appropriately configured hardware and / or software. Fig.15 A system 700 is shown, which includes a radio frequency (RF) circuit 710, a baseband circuit 720, an application circuit 730, a memory / storage device 740, a display 750, a camera 760, a sensor 770, and an input / output (I / O) interface 780, which are coupled to each other at least as shown. The application circuit 730 may include circuits, such as but not limited to one or more single-core or multi-core processors. The processor may include any combination of general-purpose processors and special-purpose processors, such as a graphics processor, an application processor. The processor may be coupled to a memory / storage device and configured to execute instructions stored in the memory / storage device to enable various applications and / or operating systems to run on the system.
[0078] The reporting of the packet true value is not limited to the report from the UE to the gNB. In one example, the gNB can send the packet true value to the UE via a downlink (e.g., PDSCH). In this case, the configuration or enabling of sending the packet report is completed by RRC signaling, MAC-CE, or DCI. In another example, the gNB or UE can send the packet true value to a third node. The packet true value included in the report can be organized by any quantization method and / or CSI measurement timing selection method in the present application. In this way, the signaling overhead between the sending node and the receiving node is reduced. In some examples, the gNB or UE can act as a sending node and the third node can act as a receiving node.
[0079] While the present disclosure has been described in connection with what is considered to be the most practical and preferred embodiment, it is to be understood that the present disclosure is not limited to the disclosed embodiment, but is intended to cover various arrangements made without departing from the scope of the broadest interpretation of the appended claims.
Claims
1. A method for reporting true value of artificial intelligence AI / machine learning ML grouping, executed by user equipment UE, characterized in that: include: The UE determines a channel state information reference signal CSI-RS measurement timing for generating an AI / ML packet true value report; as well as The actual value of the AI / ML packets is reported to the base station in a periodic, semi-persistent and / or aperiodic manner on an uplink channel.
2. The wireless communication method according to claim 1, characterized in that: The grouped true values are used for at least one of enhanced CSI feedback, predicted beam management, and positioning.
3. The method according to claim 1 or 2, characterized in that: The collection of the true value of the group of the AI / ML is activated or deactivated through at least one of radio resource configuration RRC signaling, media access control-control element MAC-CE and downlink control information DCI.
4. The method according to claim 3, characterized in that The actual value of the grouping of the AI / ML preconfigures a group size through the RRC signaling and is activated by the MAC-CE or the DCI.
5. The method according to any one of claims 1 to 4, characterized in that Reporting the actual value of the group of the AI / ML to the base station on the uplink channel is associated with UE capabilities in the periodic, semi-persistent and / or aperiodic manner.
6. The method according to any one of claims 1 to 5, characterized in that The uplink channel includes a physical uplink control channel PUCCH or a physical uplink shared channel PUSCH.
7. The method according to claim 6, characterized in that The PUCCH includes PUCCH format 3 and / or PUCCH format 4.
8. The method according to claim 6 or 7, characterized in that: If the packet true value of the AI / ML is not allowed to be transmitted through the PUCCH, the UE does not report the packet true value through the PUCCH.
9. The method according to any one of claims 1 to 8, characterized in that The CSI measurement opportunity is selected after the group true value indication for reporting the group true value of the AI / ML.
10. The method according to any one of claims 1 to 8, characterized in that The CSI measurement opportunity plus the UE processing delay is selected to generate and report the true value of the group of AI / ML.
11. The method according to claim 9 or 10, characterized in that: There is UE processing time before the CSI measurement opportunity to report the packet true value of the AI / ML.
12. The method according to any one of claims 9 to 11, characterized in that The CSI measurement occasions are selected in a sparse manner for reporting the grouped true values of the AI / ML.
13. The method according to any one of claims 1 to 12, characterized in that If sufficient true value entries are not collected for the CSI measurement occasion for reporting the group true value of the AI / ML, the UE does not report the group true value of the AI / ML at the CSI measurement occasion.
14. The method according to any one of claims 1 to 13, characterized in that If sufficient true value entries are not collected for the CSI measurement opportunity to report the group true value of the AI / ML, the UE saves the group true value of the AI / ML in a buffer and reports the group true value at the next reporting opportunity until sufficient true value entries are collected.
15. The method according to any one of claims 2 to 14, characterized in that A real value exists in the form of a vector, and a plurality of vectors are packed together as a plurality of entries to form the grouped real value, and the vectors in the grouped real value are quantized respectively.
16. The method according to claim 15, characterized in that When the grouped true value of the AI / ML is used to enhance CSI feedback, the grouped true value of the AI / ML is the original CSI-RS value or the maximum eigenvector.
17. The method according to claim 15, characterized in that When the grouped true value of the AI / ML is used for predictive beam management, the grouped true value of the AI / ML is the measured layer 1 reference signal received power L1-RSRP, and each reported L1-RSRP is paired with a beam index or position information to form each entry.
18. The method according to claim 15, characterized in that When the grouped true value of the AI / ML is used for positioning, the grouped true value of the AI / ML is location information including [coordinate, RS], [(x, y), RS] or [label, RS], where the label is a mark / label of a reference node, and each entry includes an RS amplitude quantization and phase quantization.
19. The method according to any one of claims 2 to 14, characterized in that A true value or a portion of the true value is used as a vector, and entries in the grouped true values of the AI / ML are differentially quantized relative to reference entries.
20. The method according to claim 19, characterized in that When the grouped true value of the AI / ML is used to enhance CSI feedback, the grouped true value of the AI / ML is the original CSI-RS value.
21. The method according to claim 19, characterized in that When the grouped true value of the AI / ML is used to predict beam management, the grouped true value of the AI / ML is a plurality of measured L1-RSRPs, each reported L1-RSRP is paired with a beam index or position information to form each entry, and each reported L1-RSRP and the beam index or position information are quantized according to the maximum L1-RSRP difference.
22. The method according to claim 19, characterized in that When the grouped true value of the AI / ML is used for positioning, the grouped true value of the AI / ML is a position reference signal, in which the element amplitude is quantized with a configured step value or a fixed step value, and the phase is quantized with at least one of {8PSK, QPSK, BPSK}.
23. The method according to any one of claims 1 to 22, characterized in that After the base station configures the grouped true value of the AI / ML, if the UE supports the grouped true value of the AI / ML in the UE capabilities, the UE reports the grouped true value of the AI / ML; otherwise, if the UE does not support the grouped true value of the AI / ML in the UE capabilities, the UE reports a single AI / ML true value at the reporting timing.
24. The method according to any one of claims 1 to 23, characterized in that It is a default configuration for the UE to report a single AI / ML true value at one reporting opportunity.
25. The method according to any one of claims 19 to 24, characterized in that The first vector in time, the last vector in time, the middle vector in time, the vector with the maximum Euclidean distance, the vector closest to the middle Euclidean distance value, or the vector indicated by the base station is used as a reference vector in the group true value of the AI / ML.
26. The method according to any one of claims 19 to 25, characterized in that The number of entries in the grouped true value of the AI / ML is configured by the base station through RRC signaling, MAC-CE or DCI.
27. The method according to any one of claims 1 to 26, characterized in that The UE reports the grouped true value of the AI / ML to the base station based on the capability of the UE.
28. A method for AI / ML packet true value reporting performed by a base station, characterized in that: include: Configuring a CSI-RS measurement opportunity of a UE so that the UE generates a report of a group true value of AI / ML; as well as The UE is controlled to report the grouped true value of the AI / ML on an uplink channel in a periodic, semi-persistent and / or aperiodic manner.