Method, terminal device, and network device

Abstract

The embodiments of the present disclosure relate to a method, an apparatus, and a computer-readable medium for communication. According to the embodiments of the present disclosure, a terminal device reports one or more capabilities of the terminal device to a network device. The one or more capabilities indicate that the terminal device supports an artificial intelligence / machine learning (AI / ML) model. The terminal device receives an instruction from the network device to trigger a channel state information (CSI) report for the AI / ML model. The terminal device measures a set of reference signals for the CSI report. In this way, the AI / ML model can be trained to improve the accuracy of beam management.

Description

Technical Field

[0001] Embodiments of the present disclosure generally relate to the field of telecommunications, and more particularly to communication methods, apparatuses, and computer-readable media.

Background Art

[0002] Some technologies have been proposed to improve communication performance. For example, a communication device may employ an artificial intelligence / machine learning (AI / ML) model to improve performance. In some situations, the communication device can perform beam management based on the AI / ML model.

Summary of the Invention

Problems to be Solved by the Invention

[0003] Generally, exemplary embodiments of the present disclosure provide solutions for communication.

Means for Solving the Problems

[0004] In a first aspect, a communication method is provided. The communication method includes, in a terminal device, reporting one or more capabilities of the terminal device to a network device, where the one or more capabilities at least indicate that the terminal device supports a data processing model; receiving, from the network device, one or more settings associated with the data processing model; and receiving, from the network device, an instruction for triggering a report for obtaining information associated with the data processing model.

[0005] In a second embodiment, a communication method is provided. The communication method includes a network device receiving a capability report from a terminal device, the capability of which the capability of which the capability of which the capability of which the capability of which the capability of which the terminal device supports a data processing model; transmitting to the terminal device one or more settings associated with the data processing model; and transmitting to the terminal device an instruction to trigger a report for obtaining information associated with the data processing model.

[0006] In a third embodiment, a terminal device is provided. The terminal device comprises a processing unit and a memory coupled to the processing unit for storing instructions, and when an instruction is executed by the processing unit, the terminal device performs an operation that includes reporting to a network device one or more capabilities of the terminal device, the one or more capabilities indicating at least that the terminal device supports a data processing model, receiving one or more settings associated with the data processing model from the network device, and receiving an instruction from the network device to trigger a report for obtaining information associated with the data processing model.

[0007] In a fourth embodiment, a network device is provided. The network device comprises a processing unit and a memory coupled to the processing unit for storing instructions, and when the instructions are executed by the processing unit, the network device performs an operation in the network device which includes receiving a capability report from a terminal device, the capability of which includes one or more capabilities of the terminal device, the capability of which includes at least indicating that the terminal device supports a data processing model, transmitting one or more settings associated with the AI / ML model to the terminal device, and transmitting an instruction to the terminal device to trigger a report for obtaining information associated with the data processing model.

[0008] In a fifth embodiment, a computer-readable medium is provided that, when executed on at least one processor, stores instructions causing the at least one processor to perform the method described in the first or second embodiment.

[0009] Other features of this disclosure should be easily understood from the following explanation. [Brief explanation of the drawing]

[0010] The accompanying drawings further illustrate some exemplary embodiments of this disclosure, thereby further highlighting the aforementioned and other objectives, features, and advantages of this disclosure.

[0011] [Figure 1] This is a schematic diagram of a communication environment in which the embodiments of this disclosure can be implemented.

[0012] [Figure 2] This figure shows a signaling flow for communication according to some embodiments of the present disclosure.

[0013] [Figure 3A] This is a schematic diagram illustrating fault detection in AI / ML models according to some embodiments of the present disclosure.

[0014] [Figure 3B] This is a schematic diagram illustrating fault detection in AI / ML models according to some embodiments of the present disclosure.

[0015] [Figure 4] This is a schematic diagram illustrating the training of an AI / ML model according to some embodiments of the present disclosure.

[0016] [Figure 5] This is a flowchart of an exemplary method according to an embodiment of the present disclosure.

[0017] [Figure 6] A flowchart of an exemplary method according to an embodiment of the present disclosure.

[0018] [Figure 7] A schematic block diagram of an apparatus suitable for implementing an embodiment of the present disclosure.

[0019] In the figures, the same or similar reference numerals represent the same or similar elements.

Embodiments for Carrying Out the Invention

[0020] Here, the principles of the present disclosure will be described with reference to some exemplary embodiments. It should be understood that these embodiments are described for illustrative purposes only and are intended to assist those skilled in the art in understanding and implementing the present disclosure, without suggesting any limitation on the scope of the present disclosure. The disclosure described herein can be implemented in various ways different from the methods described below.

[0021] In the following description and claims, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art of the present disclosure.

[0022] As used herein, the term “terminal device” refers to any device having wireless or wired communication capabilities. Examples of terminal devices include user equipment (UE), personal computers, desktops, mobile phones, cellular phones, smartphones, personal digital assistants (PDAs), portable computers, tablets, wearable devices, Internet of Things (IoT) devices, ultra-reliable low-latency communication (URLLC) devices, Internet of Everything (IoE) devices, machine-type communication (MTC) devices, in-vehicle devices for V2X communication where X represents pedestrians, vehicles, or infrastructure / networks, devices for integrated access and integrated access and backhaul (IAB), satellite-borne or aircraft-borne vehicles within non-terrestrial networks (NTN) including high-altitude platforms (HAP) encompassing satellites and unmanned aircraft systems (UAS), extended reality (XR) devices including different types of reality such as augmented reality (AR), mixed reality (MR), and virtual reality (VR), and unmanned aerial vehicles (UAVs), which are aircraft without human pilots and are commonly referred to as drones. This includes, but is not limited to, devices on a vehicle, a high-speed train (HST), or image acquisition devices such as digital cameras, sensor game devices, music storage and playback devices, or internet-connected home appliances that enable wireless or wired internet access and browsing. A “terminal device” may further have “multicast / broadcast” capabilities to support V2X applications, transparent IPv4 / IPv6 multicast distribution, IPTV, smart TV, wireless services, wireless software distribution, group communications, and IoT applications, where public safety and mission are of paramount importance. It may also incorporate one or more Subscriber Identity Modules (SIMs), known as multi-SIMs.The term "terminal device" may be used interchangeably with UE, mobile station, subscriber station, mobile terminal, user terminal, or wireless device. In the following description, the terms "terminal device", "communication device", "terminal", "user equipment", and "UE" may be used interchangeably.

[0023] The terminal device or network device may have the ability of artificial intelligence (AI) or machine learning. Generally, it includes a trained model from a large number of data collected for a specific function and can be used to predict some information.

[0024] The terminal device or network device may operate on some frequency ranges such as FR1 (410 MHz to 7125 MHz), FR2 (24.25 GHz to 71 GHz), frequency bands greater than 100 GHz, and terahertz (THz), etc. Furthermore, it can operate on licensed / unlicensed / shared spectrum. The terminal device may have two or more connections with the network device under a multi-radio dual connectivity (MR-DC) application scenario. The terminal device or network device can operate in full-duplex, flexible-duplex, cross-split duplex modes.

[0025] The term "network device" refers to a device that can provide or host a cell or coverage with which the terminal device can communicate. Examples of network devices include, but are not limited to, Node B (NodeB or NB), evolved Node B (eNodeB or eNB), next-generation Node B (gNB), transmit-receive point (TRP), remote radio unit (RRU), radio head (RH), remote radio head (RRH), IAB node, femto node, pico node, low-power nodes such as reconfigurable intelligent surface (RIS), etc.

[0026] In one embodiment, a terminal device can be connected to a first network device and a second network device. One of the first and second network devices may be a master node and the other a secondary node. The first and second network devices may use different radio access technologies (RATs). In one embodiment, the first network device may be a first RAT device, and the second network device may be a second RAT device. In one embodiment, the first RAT device is an eNB, and the second RAT device is a gNB. Information regarding different RATs may be transmitted to the terminal device from at least one of the first and second network devices. In one embodiment, the first information may be transmitted from the first network device to the terminal device, and the second information may be transmitted from the second network device directly or via the first network device to the terminal device. In one embodiment, information regarding the settings of the terminal device set by the second network device may be transmitted from the second network device via the first network device. Information regarding the reconfiguration of terminal devices set by the second network device may be transmitted from the second network device directly to the terminal devices or via the first network device.

[0027] The communications described herein may conform to any appropriate standard, including but not limited to New Radio Access (NR), Long-Term Evolution (LTE), LTE-Evolution, LTE-Advanced (LTE-A), Wideband Code Division Multiple Access (WCDMA®), Code Division Multiple Access (CDMA), cdma2000, and Global System for Mobile Communications (GSM). Furthermore, the communications may be performed in accordance with any generation of communication protocols currently known or to be developed in the future. Examples of communication protocols include, but are not limited to, first-generation (1G), second-generation (2G), 2.5G, 2.85G, third-generation (3G), fourth-generation (4G), 4.5G, fifth-generation (5G), and sixth-generation (6G) communication protocols. The technologies described herein can be used in the aforementioned wireless networks and technologies, as well as other wireless networks and technologies. Embodiments of this disclosure may be performed in accordance with any generation of communication protocols currently known or to be developed in the future. Examples of communication protocols include, but are not limited to, first-generation (1G), second-generation (2G), 2.5G, 2.75G, third-generation (3G), fourth-generation (4G), 4.5G, fifth-generation (5G) communication protocols, 5.5G, 5G-Advanced networks, or sixth-generation (6G) networks.

[0028] As used herein, the term “circuit” may mean a hardware circuit and / or a combination of a hardware circuit and software. For example, a circuit may be a combination of an analog and / or digital hardware circuit and software / firmware. In yet another example, a circuit may be any part of a hardware processor having a digital signal processor, software and one or more memories, which work together to cause a device such as a terminal or network device to perform various functions. In yet another example, a circuit may be a hardware circuit and / or a processor such as a microprocessor or a part thereof that requires software / firmware for operation, but the software may not be present if it is not required for operation. As used herein, the term “circuit” also includes the implementation of a hardware circuit or one or more processors alone, or a part of a hardware circuit or one or more processors and their (or their) accompanying software and / or firmware.

[0029] As used herein, the singular "one" and "the foregoing" also include the plural unless explicitly indicated in the context. The term "including" and its variations should be understood as open-ended terms meaning "including, but not limited to." The term "based on" should be understood as "at least partially based on." The terms "one embodiment" and "embodiment" should be understood as "at least one embodiment." The term "another embodiment" should be understood as "at least one other embodiment." Terms such as "first," "second," etc., may refer to different or identical subjects. The following may include other explicit and implicit definitions.

[0030] In some examples, values, procedures, or devices are referred to as “best,” “worst,” “highest,” “minimum,” “maximum,” etc. Such descriptions are intended to show that a choice can be made from among many usable functional alternatives, and it should be understood that such a choice does not need to be better, smaller, higher, or otherwise more desirable than other choices.

[0031] As mentioned above, communication equipment can perform beam management on AI / ML models. Massive MIMO (mMIMO) and beamforming are widely used in the communications industry. The terms "beamforming" and "mMIMO" may be used interchangeably in some cases. Generally speaking, beamforming uses multiple antennas to control the direction of the wavefront by appropriately weighting the amplitude and phase of the individual antenna signals in an array of multiple antennas. The most commonly seen definition is that mMIMO is a system where the number of antennas exceeds the number of users. Coverage in 5G is beam-based, not cell-based. There is no cell-level reference channel from which cell coverage can be measured. Instead, each cell has one or more Synchronization Signal Block (SSB) beams. SSB beams are static or semi-static and always point in the same direction. They form a grid of beams that cover the entire cell area. User equipment (UE) searches for and measures beams and maintains a set of candidate beams. The pair of candidate beams may include beams from multiple cells. Efficient beam management is becoming crucial, enabling directional communication with more antenna elements at 5G millimeter wave (mmWave) and providing additional beamforming gain, allowing the UE and gNB to periodically identify the optimal beam to work with at any given time.

[0032] AI / ML models may be deployed in the gNB in ​​the first stage. It is natural to discuss AI / ML for NR air interfaces starting from the gNB, as the gNB is more powerful in handling the data, models, and computing load for AI / ML. In the case of beam prediction in the spatial domain, AI / ML models trained on the gNB side are not accurate for some UEs. That is, the predicted (or inferred) optimal beam does not match the actual optimal beam from the UE side. For example, the UE can measure the quality of beams F, G, J, and K (e.g., Layer 1 reference signal received power (L1-RSRP)) and estimate the quality of all candidate beams based on the AI / ML model trained on the gNB side, and determine the optimal beam with the highest L1-RSRP (assuming beam A) based on the L1-RSRP of all candidate beams. However, in reality, beam A (i.e., the Rx beam) of the UE is covered by a shelter. If beam A is used continuously, the transmission quality will be severely affected. Offline training of the AI / ML model is not possible to account for the occurrence of all "accidents". To ensure that AI / ML models can adapt to the UE environment, offline-trained models need to be subjected to online training. For example, online training can be performed on the gNB side. However, in this scenario, the UE needs to provide a large amount of data required for training, which can cause significant UE overhead. Therefore, online training should be performed on the UE side.

[0033] According to the embodiments, a solution for improving AI / ML models is proposed. According to the embodiments of this disclosure, a terminal device reports one or more capabilities of the terminal device to a network device, which indicate that the terminal device supports an AI / ML model. The terminal device receives one or more settings associated with the AI / ML model from the network device. The terminal device receives instructions from the network device to trigger a report to obtain information associated with the AI / ML model. Thus, the AI / ML model can be trained to improve the accuracy of beam management.

[0034] Figure 1 is a schematic diagram of a communication system that can implement an embodiment of the present disclosure. The communication system 100, which is part of a communication network, comprises terminal device 110-1, terminal device 110-2, ..., terminal device 110-N, which may be collectively referred to as “terminal device 110”. The number N may be any suitable integer. The terminal devices 110 may communicate with each other. As an example only, terminal device 110-1 may be configured to have a plurality of beams, indicated as beams 131-1, 132-1, 133-1 and 134-1. The network device 120 may be configured to have a plurality of beams, indicated as beams 131-2, 132-2, 133-2 and 134-2. A beam pair may include two beams, for example, beams 131-1 and 131-2, 132-1 and 132-2, 133-1 and 133-2, and 134-1 and 134-2. It should be noted that the number of beams shown in Figure 1 is not limited to any specific configuration but is merely an example.

[0035] The communication system 100 further includes network equipment. In the communication system 100, the network equipment 120 and the terminal equipment 110 can communicate data and control information with each other. The number of terminal equipment shown in Figure 1 is for illustrative purposes only and does not imply any limitation.

[0036] Communication in the communication system 100 can be implemented in accordance with any suitable communication protocol, including but not limited to, cellular communication protocols such as first-generation (1G), second-generation (2G), third-generation (3G), fourth-generation (4G), and fifth-generation (5G), wireless local area network communication protocols such as IEEE 802.11, and / or any other protocols currently known or to be developed in the future. Furthermore, communication may utilize any suitable wireless communication technology, including but not limited to code division multiple access (CDMA), frequency division multiple access (FDMA), time division multiple access (TDMA), frequency division duplexer (FDD), time division duplexer (TDD), multi-input multiple-output (MIMO), orthogonal frequency division multiple access (OFDMA), and / or any other technologies currently known or to be developed in the future.

[0037] Embodiments of the present disclosure can be applied to any suitable scenario. For example, embodiments of the present disclosure can be implemented on NR equipment with reduced capabilities. Alternatively, embodiments of the present disclosure can be implemented within one of the following: NR multi-input multi-output (MIMO), NR sidelink enhancement, NR systems with frequencies above 52.6 GHz, extended NR operations up to 71 GHz, narrowband Internet of Things (NB-IOT) / extended machine-type communications (eMTC) on non-terrestrial networks (NTN), NTN, UE power saving enhancement, NR coverage enhancement, NB-IOT and LTE-MTC, integrated access and backhaul (IAB), NR multicast and broadcast services, or multi-radio dual connectivity enhancement.

[0038] As used herein, the term "slot" means a dynamic scheduling unit. A slot contains a predetermined number of symbols. The term "downlink (DL) subslot" may refer to a virtual subslot built upon an uplink (UL) subslot. A DL subslot may contain fewer symbols than a single DL slot. As used herein, a slot may refer to a regular slot containing a predetermined number of symbols and a subslot containing fewer symbols than said predetermined number.

[0039] Embodiments of the present disclosure are described in detail below. First, we refer to Figure 2, which shows a signaling diagram illustrating a process 200 between a terminal device and a network device according to some exemplary embodiments of the present disclosure. For illustrative purposes only, we will refer to Figure 1 to describe the process 200. The process 200 may involve terminal devices 110-1 and network device 120 in Figure 1. In some embodiments, the process 200 is applicable to fault detection in an AI / ML model. Alternatively, the process 200 is applicable to online training of an AI / ML model.

[0040] Terminal device 110-1 reports one or more capabilities of terminal device 110-1 to network device 120 (2010). The one or more capabilities indicate at least that terminal device 110-1 supports a data processing model. The data processing model may be an AI / ML model. As used herein, “AI / ML model” may mean a program or algorithm that utilizes a set of data that makes a particular pattern recognizable, thereby enabling it to reach a conclusion or make a prediction given sufficient information. Generally speaking, an AI / ML model may be a mathematical algorithm that is “trained” with data and input from human experts to reproduce the decisions that experts would make given the same information.

[0041] In some embodiments, capability may indicate the ability to support AI / ML. In some embodiments, capability may indicate the ability to support beam management based on AI / ML. Alternatively or additionally, capability may indicate the ability to support beam prediction in the spatial domain based on AI / ML. Additionally, capability may indicate that terminal device 110-1 supports the ability to support online training. For example, fine-tuning is applicable to online training. It should be noted that other suitable methods are also applicable to online training. Capability may also indicate the index of AI / ML models supported by terminal device 110-1. In some other embodiments, capability may indicate a first time delay, which is the minimum time required for terminal device 110-1 to detect a fault in the AI / ML model. Thus, network device 120 can set the corresponding AI / ML model and associated activation parameters for terminal device 110-1.

[0042] In other embodiments, capability may represent a second time delay, which is the minimum time required for online training of the AI / ML model on the terminal device 110-1 side. This facilitates integrated analysis of the AI / ML model.

[0043] The network device 120 transmits one or more settings associated with the AI / ML model to the terminal device 110-1 (2020). The settings may also be transmitted via upper-layer signaling. For example, the settings may be transmitted via RRC signaling.

[0044] In some embodiments, the one or more settings may include a first setting. In this case, the first setting may indicate an index of the AI / ML model. Additionally, the first setting may indicate a first type of parameter of the AI / ML model. In this case, the first type of parameter may include structural parameters of the AI / ML model. For example, the structural parameter may indicate a deep neural network (DNN) of the AI / ML model. Alternatively, the structural parameter may indicate a convolutional neural network (CNN) of the AI / ML model. The structural parameter may also indicate the number of layers of the AI / ML model. The structural parameter may indicate the type of layers of the AI / ML model. Additionally, the structural parameter may indicate the number of neurons of the AI / ML model. In some other embodiments, the first type of parameter may include a factor of the AI / ML model. For example, the factor may be a weight factor. Alternatively or additionally, the factor may be a bias factor. Additionally or alternatively, the first setting may indicate a first type of parameter. In other words, terminal device 110-1 may be configured to have an AI / ML model and corresponding parameters of a first type. The parameters of the first type may include the input data format of the AI / ML model. For example, the input data format may include the number of rows and columns of the input data. Additionally, the input data format may include the units of the input data. The input data format may also include the interpretation of the input data.

[0045] Additionally, the first type of parameter may include the output data format of the AI / ML model. Similarly, the output data format may include the number of rows and columns of the output data. Additionally, the output data format may include the units of the output data. The output data format may also include the interpretation of the output data.

[0046] In some embodiments, the first type of parameters may include pre-processing parameters for the AI / ML model. The first type of parameters may also include post-processing parameters for the AI / ML model. For example, the first type of parameters may include standardization coefficients. The term "standardization coefficient" may refer to flattening the input and output data to the same distribution by performing a normalization operation on the input data, thereby accelerating the training network. For example, in the case of maximum value-based normalization, all input data may be divided by the maximum value, which is the standardization coefficient.

[0047] Additionally, the first configuration may also include a first activation parameter. The first activation parameter can be used to enable the terminal device 110-1 to detect faults in the AI / ML model. For example, if the first configuration does not include the first activation parameter, the terminal device 110-1 does not have to perform fault detection in the AI / ML model. Alternatively, if the first activation parameter is set to disabled (e.g., set to "0"), the terminal device 110-1 does not have to perform fault detection in the AI / ML model. In other embodiments, if the first activation parameter is set to enabled (e.g., set to "1"), the terminal device 110-1 may perform fault detection in the AI / ML model. In some embodiments, the first activation parameter may be transmitted within the CSI reporting configuration. Alternatively, the first activation parameter may be transmitted via other higher-layer signaling. Thus, the terminal device can determine the AI / ML model and feasible behavior.

[0048] In other embodiments, the first setting may include a third activation parameter. The third activation parameter can be used to enable the terminal device 110-1 to perform online training of the AI / ML model. For example, if the first setting does not include the third activation parameter, the terminal device 110-1 does not have to perform online training of the AI / ML model. Alternatively, if the third activation parameter is set to disabled (e.g., set to "0"), the terminal device 110-1 does not have to perform online training of the AI / ML model. In other embodiments, if the third activation parameter is set to enabled (e.g., set to "1"), the terminal device 110-1 may perform online training of the AI / ML model. Thus, the terminal device can perform online training of the AI / ML model.

[0049] The network device 120 transmits an instruction to trigger a report to obtain information associated with the AI / ML model (2030). In some embodiments, the CSI report may be a periodic CSI report configurable by RRC signaling. Alternatively, the CSI report may be a semi-persistent CSI report. In this case, such a CSI report can be activated by a media access control element (MAC CE) from the network device 120. In other embodiments, the CSI report may be a non-periodic CSI report or an SP-CSI report. In this case, the CSI report may be triggered by downlink control information (DCI) from the network device 120.

[0050] In some embodiments, the network device 120 may transmit the CSI report configuration. For example, the CSI report configuration may be transmitted via a higher-layer configuration of CSI-ReportConfig. The CSI report configuration may indicate that the CSI report is associated with one or more CSI-RS or SSB resource sets. In some embodiments, the CSI-RS resource set may be periodic. Alternatively, the CSI-RS resource set may be semi-permanent. In other embodiments, the CSI-RS resource set may be aperiodic. In some embodiments, each CSI-RS resource in the CSI-RS resource set may correspond to a beam. The CSI report configuration may also indicate an index of an AI / ML model, meaning that the CSI report is associated with an AI / ML model. Thus, the terminal device can know which AI / ML model the CSI report is applied to.

[0051] The network device 120 may transmit a set of reference signals to the terminal device 110-1 (2040). The network device 120 may transmit the set of reference signals based on the CSI-RS resource set.

[0052] The terminal device 110-1 may measure the set of reference signals (2050). The terminal device 110-1 may measure the reference signal received power (RSRP) for the set of reference signals. For example, the terminal device 110-1 may measure the CSI-RS resource indicator (CRI)-RSRP based on the set of reference signals. Alternatively, the SS / PBCH resource block indicator (SSBRI)-RSRP may be measured. In another embodiment, the terminal device 110-1 may measure the signal-to-interference plus noise ratio (SINR) based on the set of reference signals. Alternatively, the terminal device 110-1 may measure the SSBRI-SINR based on the set of reference signals.

[0053] Terminal device 110-1 may send a report to network device 120 (2060). In some embodiments, the report may include information associated with the AI / ML model. For example, the report may include first information indicating whether the AI / ML model has failed. The first information may be a report quantity. For example, the first information may occupy a bit field in the report. For illustrative purposes, if the first information indicates "1", it means that an AI / ML model failure has occurred. If the first information indicates "0", it means that no AI / ML model failure has occurred. Thus, by reporting the first information, the terminal device can notify the network device whether the AI / ML model has failed.

[0054] In other embodiments, the report may indicate measurement results. For example, the report may include one of CRI-RSRP, SSBRI-RSRP, CRI-SINR, or SSBRI-SINR. In some embodiments, one or more of the above-described settings may include a second enablement parameter. The second enablement parameter may be used to indicate that the associated report is used to detect failures in the AI / ML model. In this case, if a failure is detected in the AI / ML model, the terminal device 110-1 may not report anything, in other words, the terminal device 110-1 may not send a report. Report transmission may be skipped. At the same time, if there are no other competing reports in the time domain, the network device 120 may not receive any reports. In this case, the network device 120 may determine that an AI / ML model failure has occurred. Alternatively, if no faults are detected in the AI / ML model, the terminal device 110-1 may report K (K>0) optimal beams and their corresponding beam qualities, i.e., K CRIs and their corresponding L1-RSRPs, in the report. Thus, since the report is associated with a second activation parameter, the terminal device also knows that the report is used to detect AI / ML model faults, even though the report quality is set to "CRI-RSRP".

[0055] Alternatively, if a failure is detected in the AI / ML model, the report may include a bit field indicating a predetermined value. For example, if a failure occurs in the AI / ML model, the values ​​of the bit field corresponding to the CRI and / or the L1-RSRP field in the report may all be set to "1".

[0056] In some embodiments, one or more of the above-described settings may include a second setting, which may indicate a time offset. In this case, terminal device 110-1 may transmit reports on a slot determined based on the time offset. In some embodiments, the time offset may depend on a first time delay, which is the minimum time required for terminal device 110-1 to detect a fault in the AI / ML model. Alternatively, the time offset may depend on a second time delay, which is the minimum time required for online training of the AI / ML model in terminal device 110-1. The time offset may indicate on which slot terminal device 110-1 will transfer the report amount. In some embodiments, the time offset may refer to a time offset by a higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Alternatively, the time offset may be set via an RRC setting independently of the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. In this case, if the first time offset is set within the RRC settings, the terminal device 110-1 may ignore the value provided by the higher-level settings of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. In this case, the terminal device can know when to perform the report.

[0057] Terminal device 110-1 may determine that a failure has occurred on the AI / ML model if at least one condition is met. For example, terminal device 110-1 may determine that a failure has occurred on the AI / ML model if the first target beam determined based on the AI / ML model is different from the second target beam determined based on the measurement of the set of reference signals. Alternatively, terminal device 110-1 may determine that a failure has occurred on the AI / ML model if the difference between the first amount of the first target beam and the second amount of the second target beam exceeds a first threshold. In another embodiment, terminal device 110-1 may determine that a failure has occurred on the AI / ML model if the number of times the condition has been met exceeds a second threshold. For example, the second threshold may be set via RRC signaling. In this case, the terminal device and the network device know under what conditions the AI / ML model fails.

[0058] In some embodiments, terminal device 110-1 may transmit multiple reports. In this case, each of the multiple reports may indicate a corresponding priority. For example, if it is necessary to transmit other uplink information (e.g., another report), terminal device 110-1 may determine whether the other uplink information conflicts with the report. If the other uplink information conflicts with the report, terminal device 110-1 may compare the first priority of the report with the second priority of the other uplink information. In this case, if the first priority is higher than the second priority, terminal device 110-1 may transmit the report to the network device. The value of the first priority may be determined from a set of values. The value of the second priority may also be determined from the same set of values. The set of values ​​may include 0, 1, and 2. It should be noted that the set of values ​​may also include other values. Thus, when multiple reports carrying different content conflict, the terminal device knows which CSI report to transmit first.

[0059] Alternatively, the network device 120 may determine whether or not a fault has occurred on the AI / ML model. In this case, a new reporting quantity, e.g., second information, may be introduced. The second information may include inference information and actual information. In this case, the inference information includes at least one of the following: an inference beam determined based on the AI / ML model, or the beam quality corresponding to the inference beam. Additionally, the actual information includes at least one of the following: an actual beam determined based on measurements of a set of reference signals associated with the report, or the beam quality corresponding to the actual beam. In some embodiments, the beam quality may be RSRP. Alternatively, the beam quality may be SINR. In some embodiments, the beam quality may be defined by a 7-bit value. The beam quality may be in the range of -140 dBm to -44 dBm with a step size of 1 dB. In some embodiments, the bit fields corresponding to the inference beam and the beam quality corresponding to the inference beam are before the bit fields corresponding to the actual beam and the beam quality corresponding to the actual beam. Alternatively, the bit fields corresponding to the inference beam and the beam quality corresponding to the inference beam are after the bit fields corresponding to the actual beam and the beam quality corresponding to the actual beam.

[0060] In some embodiments, the report may be associated with a third type of information. In this case, the third type of information may include a second type of parameters of the trained AI / ML model updated based on the AI / ML model. For example, the third type of information may mainly include the amount of change in the trained AI / ML model compared to the AI / ML model, e.g., the weight factor and the bias factor. This facilitates integrated analysis of the model on the network device side. Alternatively, the terminal device 110-1 does not need to report the trained AI / ML model. In this case, the terminal device 110-1 only needs to report K optimal beams.

[0061] Exemplary embodiments will be described in detail with reference to Figures 3A-35. Figure 3A shows the process by which a fault in the AI / ML model is detected by the terminal device 110-1.

[0062] Terminal device 110-1 reports one or more capabilities of terminal device 110-1 to network device 120 (3010). The one or more capabilities indicate at least that terminal device 110-1 supports AI / ML models.

[0063] In some embodiments, capability may indicate the ability to support AI / ML. In some embodiments, capability may indicate the ability to support beam management based on AI / ML. Alternatively or additionally, capability may indicate the ability to support beam prediction in the spatial domain based on AI / ML. Additionally, capability may indicate that terminal device 110-1 supports the ability to support online training. For example, fine-tuning is applicable to online training. It should be noted that other suitable methods are also applicable to online training. Capability may also indicate the index of AI / ML models supported by terminal device 110-1. In some other embodiments, capability may indicate a first time delay, which is the minimum time required for terminal device 110-1 to detect a fault in the AI / ML model. Thus, network device 120 can set the corresponding AI / ML model and associated activation parameters for terminal device 110-1.

[0064] Terminal device 110-1 may be configured to have an AI / ML model and a first activation parameter through higher-level configuration (e.g., RRC signaling). The AI / ML model may include the AI / ML model index (ID), structural parameters (e.g., DNN, CNN, number of layers, layer type, number of neurons), and factors (e.g., weight factor, bias factor). The first activation parameter is used to enable terminal device 110-1 to perform the function of detecting AI / ML model failures. If terminal device 110-1 is not configured to have the first activation parameter, or if the first activation parameter is set to disabled, terminal device 110-1 is not expected to perform AI / ML model failure detection.

[0065] In some embodiments, the terminal device 110-1 may also be configured to have a first activation parameter and a first type of parameter. The first type of parameter may include the input / output data format of the AI / ML model, and pre-processing / post-processing parameters (e.g., standardization coefficients). The AI / ML model may be associated with the first type of parameter. For example, the terminal device 110-1 may be configured to have an AI / ML model and a corresponding first type of parameter.

[0066] Network device 120 triggers a CSI report for terminal device 110-1 (3020). The CSI report may be a periodic CSI report configured by RRC, a semi-persistent CSI report activated by MAC-CE, an aperiodic CSI report triggered by DCI, or an SP-CSI report. The CSI report may also refer to the higher-level settings of CSI-ReportConfig.

[0067] The CSI report is associated with one or more CSI-RS (or SSB) resource sets 3030 with retransmission turned off. The type of CSI-RS resource set may be periodic, semi-permanent, or aperiodic. Each CSI-RS resource in the CSI-RS resource set corresponds to a beam. On the terminal equipment side, these CSI-RS resources are used to collect input / output data for detecting AI / ML model failures. For example, each CSI-RS resource may correspond to beams 131-1, 132-1, 133-1, and 134-1, respectively.

[0068] CSI reports may be associated with AI / ML models. For example, the index of an AI / ML model may be configured within the higher-level settings of the CSI report (e.g., CSI-ReportConfig).

[0069] Terminal device 110-1 reports first information (3040). First information may be a newly introduced reporting quantity, i.e., the content reported by terminal device 110-1. Specifically, first information may be used to indicate whether or not the AI / ML model associated with the CSI report has failed. The first information report occupies one bit field. "1" means "AI / ML model failure occurred," and "0" means "no AI / ML model failure occurred."

[0070] Alternatively, the reporting quantity associated with the CSI report may be set to "CRI-RSRP" (or "SSBRI-RSRP", "CRI-SINR", or "SSBRI-SINR"), and at the same time, the CSI report may be associated with a second activation parameter. In this case, the second activation parameter may be used to indicate that the associated CSI report is used to detect AI / ML model failures. In particular, if an AI / ML model failure occurs, terminal device 110-1 will not report any quantity. In other words, terminal device 110-1 will not forward the CSI report to network device 120. At this time, if there are no other conflicting CSI reports in the time domain, network device 120 will not receive any CSI reports and will therefore consider that an AI / ML model failure has occurred. Otherwise (i.e., if no AI / ML model failure occurs), terminal device 110-1 may report K (K>0) optimal beams and their corresponding beam qualities, i.e., K CRIs and their corresponding L1-RSRPs. Optionally, if an AI / ML model failure occurs, the bit field values ​​corresponding to the CRI and / or L1-RSRP reported by the UE are "1111..." (i.e., all "1"s).

[0071] In some embodiments, the CSI report may be associated with a first time offset 310. The first time offset 310 may be used to indicate on which slot the terminal device 110-1 transfers the reported amount. Specifically, the first time offset 310 may refer to a time offset specified by a higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Furthermore, the value of the first time offset 310 depends on the reported first time delay. Optionally, the first time offset 310 may be set via RRC independently of the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Furthermore, if the first time offset 310 is set (or if a field for the first time offset exists), the terminal device 110-1 should ignore the value provided by the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList.

[0072] Terminal device 110-1 may determine the availability of the AI / ML model according to the following criteria. In this case, terminal device 110-1 may report first information indicating whether or not an AI / ML model failure has occurred.

[0073] Criterion 1: The optimal beam inferred based on the AI / ML model does not match the actual optimal beam not based on the AI / ML model. Specifically, "inferred optimal beam" means "the CRI corresponding to the inferred CSI-RS resource with the largest quantity (e.g., L1-RSRP, L1-SINR)", and "actual optimal beam" means the CRI corresponding to the actual CSI-RS resource with the largest quantity. For example, the inferred optimal beam based on the AI / ML model may be beam 131-1, and the actual optimal beam not based on the AI / ML model may be beam 133-1. Criterion 2: The difference between the quantity corresponding to the estimated optimal beam (e.g., L1-RSRP, L1-SINR) and the quantity corresponding to the actual optimal beam is greater than the first threshold (set by RRC). For example, the difference between the L1-RSRP of beam 131-1 and the L1-RSRP of beam 133-1 may exceed the first threshold. Specifically, if either Criterion 1 or Criterion 2 is met, terminal device 110-1 may determine that an AI / ML model failure has occurred. Furthermore, terminal device 110-1 may be configured to have a second threshold (set by RRC). Assuming the value of the second threshold is 3, if Criterion 1 or Criterion 2 is met three or more times consecutively, the AI / ML model is unavailable. Otherwise, the AI / ML model is available.

[0074] When multiple CSI reports are triggered simultaneously by the network device 120, these CSI reports may collide. For example, two CSI reports may be considered to collide if the time occupancy of the physical channels scheduled to carry these CSI reports overlaps within at least one OFDM symbol and is transmitted on the same carrier. In this case, a priority should be defined for the CSI reports. For example, if a CSI report associated with first information collides with another CSI report for beam (e.g., CRI-RSRP) or CSI acquisition (e.g., CRI-RI-PMI-CQI) and the time domain types corresponding to these CSI reports are the same (e.g., aperiodic), the terminal device 110-1 will preferentially transmit the CSI report associated with first information (e.g., carrying). In this case, for a CSI report carrying the first information, k=0, and for CSI reports carrying L1-RSRP or L1-SINR and CSI reports not carrying L1-RSRP, L1-SINR, or the first information, k=1 and 2, respectively, where "k" is a parameter used to calculate the priority value of the CSI report. Optionally, the values ​​of k corresponding to CSI reports carrying the first information, CSI reports carrying L1-RSRP or L1-SINR, and CSI reports not carrying L1-RSRP, L1-SINR, or the first information are [1,0,2] or [2,0,1], respectively. That is, a CSI report carrying the first information has an intermediate or lowest priority.

[0075] Figure 3B shows the process by which faults in the AI / ML model are detected by the network device 120.

[0076] As shown in Figure 3B, terminal device 110-1 reports second information (3050). In this case, in some embodiments, the second information may be a new reported quantity. The second information may include the inferred optimal beam (e.g., beam 131-1), the actual optimal beam (e.g., beam 133-1), the corresponding beam quality, i.e., the CRI corresponding to the estimated CSI-RS resource having the largest quantity (e.g., L1-RSRP), and the corresponding quantity, the CRI corresponding to the actual CSI-RS resource having the largest quantity. In some embodiments, the reported L1-RSRP values ​​corresponding to the estimated optimal beam and the actual optimal beam may be defined by a 7-bit value in the range of [-140,-44] dBm with a step size of 1 dB. Furthermore, the estimated (or actual) optimal beam and the corresponding L1-RSRP may occupy the forward bit field, while the other may occupy the remaining bit field. Optionally, the reporting quantity may be set to "CRI-RSRP" (or "SSBRI-RSRP", "CRI-SINR", or "SSBRI-SINR"), and simultaneously, if the CSI report is associated with a second activation parameter, terminal device 110-1 should not be expected to report K optimal beams (reusing R15 / 16), but should report the inferred optimal beam, the actual optimal beam, and the corresponding beam quality.

[0077] Figure 4 shows the process by which online training of the AI / ML model is performed by terminal device 110-1.

[0078] The terminal device 110-1 reports one or more capabilities of the terminal device 110-1 to the network device 120 (4010). These one or more capabilities indicate at least that the terminal device 110-1 supports AI / ML. In some embodiments, the capabilities may indicate a second time delay, which is the minimum time required for online training on the terminal device 110-1 side. This facilitates the integrated analysis of the AI / ML model.

[0079] Terminal device 110-1 may be configured to have an AI / ML model and a third activation parameter. The third activation parameter can be used to enable terminal device 110-1 to perform online training of the AI / ML model. In this case, if terminal device 110-1 is not configured to have the third activation parameter, or if the third activation parameter is set to disabled, terminal device 110-1 will not be expected to perform online training on the AI / ML model.

[0080] Network device 120 triggers a CSI report for terminal device 110-1 (4020). The CSI report may be a periodic CSI report configured by RRC, a semi-persistent CSI report activated by MAC-CE, an aperiodic CSI report triggered by DCI, or an SP-CSI report. The CSI report may also refer to the higher-level settings of CSI-ReportConfig.

[0081] In some embodiments, CSI reports may be associated with CSI-RS resource set 4030. Compared to fault detection, CSI reports may be associated with more CSI-RS resource sets because more data is required to train the AI / ML model.

[0082] Terminal device 110-1 reports the trained AI / ML model (4040). For example, the CSI report may be associated with a third piece of information. In this case, the third piece of information may be a newly introduced reporting quantity. Specifically, the third piece of information may include parameters related to the trained AI / ML model, for example, mainly the amount of change of the trained AI / ML model compared to the AI / ML model, such as the weight factor and the bias factor. Optionally, terminal device 110-1 does not have to report the trained AI / ML model. For example, the reporting quantity may be set to "CRI-RSRP", and terminal device 110-1 only needs to report the K optimal beams.

[0083] In some embodiments, the CSI report may be associated with a second time offset 410. The second time offset 410 may be used to indicate on which slot the terminal device 110-1 transfers the reported amount. Specifically, the second time offset 410 may refer to a time offset by a higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Furthermore, the value of the second time offset 410 depends on the reported second time delay. Optionally, the second time offset 410 may be set via RRC independently of the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Furthermore, if the second time offset 410 is set (or if a field for the second time offset exists), the terminal device 110-1 should ignore the value provided by the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList.

[0084] In some embodiments, priority should be defined for CSI reports. In some embodiments, the first information and the third information may coexist. The values ​​of k corresponding to a CSI report carrying the first information, a CSI report carrying the third information, a CSI report carrying L1-RSRP or L1-SINR, and a CSI report that does not carry L1-RSRP, L1-SINR, the first information, or the third information are [0,1,2,3], [2,3,0,1], [0,2,1,3], or [0,3,1,2].

[0085] Figure 5 is a flowchart of an exemplary method 500 according to an embodiment of the present disclosure. Method 500 can be implemented in any suitable apparatus. For illustrative purposes only, Method 500 can be implemented in a terminal device 110-1 as shown in Figure 1.

[0086] In block 510, terminal device 110-1 reports one or more capabilities of terminal device 110-1 to network device 120. These one or more capabilities indicate at least that terminal device 110-1 supports AI / ML models.

[0087] In some embodiments, capability may indicate the ability to support AI / ML. In some embodiments, capability may indicate the ability to support beam management based on AI / ML. Alternatively or additionally, capability may indicate the ability to support beam prediction in the spatial domain based on AI / ML. Additionally, capability may indicate that terminal device 110-1 supports the ability to support online training. For example, fine-tuning is applicable to online training. It should be noted that other suitable methods are also applicable to online training. Capability may also indicate the index of AI / ML models supported by terminal device 110-1. In some other embodiments, capability may indicate a first time delay, which is the minimum time required for terminal device 110-1 to detect a fault in the AI / ML model. Thus, network device 120 can set the corresponding AI / ML model and associated activation parameters for terminal device 110-1.

[0088] In other embodiments, capability may represent a second time delay, which is the minimum time required for online training of the AI / ML model on the terminal device 110-1 side. This facilitates integrated analysis of the AI / ML model.

[0089] In block 520, terminal device 110-1 receives one or more settings of the data processing model from network device 120. These one or more settings may be transmitted via upper-layer signaling. For example, these one or more settings may be transmitted via RRC signaling.

[0090] In some embodiments, the one or more settings may include a first setting. In this case, the first setting may indicate an index of the AI / ML model. Additionally, the first setting may indicate a first type of parameter of the AI / ML model. In this case, the first type of parameter may include structural parameters of the AI / ML model. For example, the structural parameter may indicate a deep neural network (DNN) of the AI / ML model. Alternatively, the structural parameter may indicate a convolutional neural network (CNN) of the AI / ML model. The structural parameter may also indicate the number of layers of the AI / ML model. The structural parameter may indicate the type of layers of the AI / ML model. Additionally, the structural parameter may indicate the number of neurons of the AI / ML model. In some other embodiments, the first setting may indicate a factor of the AI / ML model. For example, the factor may be a weight factor. Alternatively or additionally, the factor may be a bias factor.

[0091] As an addition or alternative, the first setting may indicate a first type of parameter. In other words, the terminal device 110-1 may be configured to have an AI / ML model and a corresponding first type of parameter. The first type of parameter may include the input data format of the AI / ML model. For example, the input data format may include the number of rows and columns of the input data. Additionally, the input data format may include the units of the input data. The input data format may also include the interpretation of the input data.

[0092] Additionally, the first type of parameter may include the output data format of the AI / ML model. Similarly, the output data format may include the number of rows and columns of the output data. Additionally, the output data format may include the units of the output data. The output data format may also include the interpretation of the output data.

[0093] In some embodiments, the first type of parameters may include pre-processing parameters for the AI / ML model. The first type of parameters may also include post-processing parameters for the AI / ML model. For example, the first type of parameters may include standardization coefficients. The term "standardization coefficient" may refer to flattening the input and output data to the same distribution by performing a normalization operation on the input data, thereby accelerating the training network. For example, in the case of maximum value-based normalization, all input data may be divided by the maximum value, which is the standardization coefficient.

[0094] Additionally, the first setting may also include a first activation parameter. The first activation parameter can be used to enable the terminal device 110-1 to detect faults in the AI / ML model. For example, if the first setting does not include the first activation parameter, the terminal device 110-1 does not have to perform fault detection in the AI / ML model. Alternatively, if the first activation parameter is set to disabled (e.g., set to "0"), the terminal device 110-1 does not have to perform fault detection in the AI / ML model. In other embodiments, if the first activation parameter is set to enabled (e.g., set to "1"), the terminal device 110-1 may perform fault detection in the AI / ML model. In this way, the terminal device can determine the AI / ML model and its executable behavior.

[0095] In other embodiments, the first setting may include a third activation parameter. The third activation parameter can be used to enable the terminal device 110-1 to perform online training of the AI / ML model. For example, if the first setting does not include the third activation parameter, the terminal device 110-1 does not have to perform online training of the AI / ML model. Alternatively, if the third activation parameter is set to disabled (e.g., set to "0"), the terminal device 110-1 does not have to perform online training of the AI / ML model. In other embodiments, if the third activation parameter is set to enabled (e.g., set to "1"), the terminal device 110-1 may perform online training of the AI / ML model. Thus, the terminal device can perform online training of the AI / ML model.

[0096] In block 530, terminal device 110-1 receives an instruction to trigger a report to obtain information associated with the data processing model. In some embodiments, the CSI report may be a periodic CSI report configurable by RRC signaling. Alternatively, the CSI report may be a semi-persistent CSI report. In this case, such a CSI report can be activated by a media access control element (MAC CE) from network device 120. In other embodiments, the CSI report may be an aperiodic CSI report or an SP-CSI report. In this case, the CSI report may be triggered by downlink control information (DCI) from network device 120.

[0097] In some embodiments, terminal device 110-1 may receive CSI report settings. For example, CSI report settings may be transmitted via higher-level settings of CSI-ReportConfig. CSI report settings may indicate that the CSI report is associated with one or more CSI-RS or SSB resource sets. In some embodiments, the CSI-RS resource sets may be periodic. Alternatively, the CSI-RS resource sets may be semi-permanent. In other embodiments, the CSI-RS resource sets may be aperiodic. In some embodiments, each CSI-RS resource in the CSI-RS resource set may correspond to a beam. CSI report settings may also indicate an index of an AI / ML model, meaning that the CSI report is associated with an AI / ML model. Thus, the terminal device can know which AI / ML model the CSI report is applied to.

[0098] In some embodiments, terminal device 110-1 may receive a set of reference signals. Network device 120 may transmit the set of reference signals based on a CSI-RS resource set.

[0099] In some embodiments, the terminal device 110-1 may measure the set of reference signals. The terminal device 110-1 may measure the RSRP for the set of reference signals. For example, the terminal device 110-1 may measure the CRI-RSRP based on the set of reference signals. Alternatively, the SSBRI-RSRP may be measured. In some embodiments, the terminal device 110-1 may measure the CRI SINR based on the set of reference signals. Alternatively, the terminal device 110-1 may measure the SSBRI-SINR based on the set of reference signals.

[0100] Terminal device 110-1 may transmit a report to network device 120. In some embodiments, the report may include information associated with the AI / ML model. For example, the report may include first information indicating whether the AI / ML model has failed. The first information may be a report quantity. For example, the first information may occupy a bit field in the report. For illustrative purposes, if the first information indicates "1", it means that an AI / ML model failure has occurred. If the first information indicates "0", it means that no AI / ML model failure has occurred. Thus, by reporting the first information, the terminal device can notify the network device whether the AI / ML model has failed.

[0101] In other embodiments, the report may indicate measurement results. For example, the report may include one of CRI-RSRP, SSBRI-RSRP, CRI-SINR, or SSBRI-SINR. In some embodiments, one or more of the above-described settings may include a second enablement parameter. The second enablement parameter may be used to indicate that the associated report is used to detect failures in the AI / ML model. In this case, if a failure is detected in the AI / ML model, the terminal device 110-1 may not report anything, in other words, the terminal device 110-1 may not send a report. Report transmission may be skipped. At the same time, if there are no other competing reports in the time domain, the network device 120 may not receive the CSI report. In this case, the network device 120 may determine that an AI / ML model failure has occurred. Alternatively, if no faults are detected in the AI / ML model, the terminal device 110-1 may report K (K>0) optimal beams and their corresponding beam qualities, i.e., K CRIs and their corresponding L1-RSRPs, in the report. Thus, since the report is associated with a second activation parameter, the terminal device still knows that the CSI report is used to detect AI / ML model faults, even though the report quality is set to "CRI-RSRP".

[0102] Alternatively, if a failure is detected in the AI / ML model, the report may include a bit field indicating a predetermined value. For example, if a failure occurs in the AI / ML model, the values ​​of the bit field corresponding to the CRI and / or the L1-RSRP field in the report may all be set to "1".

[0103] In some embodiments, one or more of the above-described settings may include a second setting, which may indicate a time offset. In this case, terminal device 110-1 may transmit reports on a slot determined based on the time offset. In some embodiments, the time offset may depend on a first time delay, which is the minimum time required for terminal device 110-1 to detect a fault in the AI / ML model. Alternatively, the time offset may depend on a second time delay, which is the minimum time required for online training of the AI / ML model in terminal device 110-1. The time offset may indicate on which slot terminal device 110-1 will transfer the report amount. In some embodiments, the time offset may refer to a time offset by a higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Alternatively, the time offset may be set via an RRC setting independently of the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. In this case, if the first time offset is set within the RRC settings, the terminal device 110-1 may ignore the value provided by the higher-level settings of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. In this case, the terminal device can know when to perform the report.

[0104] Terminal device 110-1 may determine that a failure has occurred on the AI / ML model if at least one condition is met. For example, terminal device 110-1 may determine that a failure has occurred on the AI / ML model if the first target beam determined based on the AI / ML model is different from the second target beam determined based on the measurement of the set of reference signals. Alternatively, terminal device 110-1 may determine that a failure has occurred on the AI / ML model if the difference between the first amount of the first target beam and the second amount of the second target beam exceeds a first threshold. In another embodiment, terminal device 110-1 may determine that a failure has occurred on the AI / ML model if the number of times the condition has been met exceeds a second threshold. For example, the second threshold may be set via RRC signaling. In this case, the terminal device and the network device know under what conditions the AI / ML model fails.

[0105] In some embodiments, terminal device 110-1 may transmit multiple reports. In this case, each of the multiple reports may indicate a corresponding priority. For example, if it is necessary to transmit other uplink information (e.g., another report), terminal device 110-1 may determine whether the other uplink information conflicts with the report. If the other uplink information conflicts with the report, terminal device 110-1 may compare the first priority of the report with the second priority of the other uplink information. In this case, if the first priority is higher than the second priority, terminal device 110-1 may transmit the report to the network device. The value of the first priority may be determined from a set of values. The value of the second priority may also be determined from the same set of values. The set of values ​​may include 0, 1, and 2. It should be noted that the set of values ​​may also include other values. Thus, when multiple reports carrying different content conflict, the terminal device knows which CSI report to transmit first.

[0106] Alternatively, the network device 120 may determine whether or not a fault has occurred on the AI / ML model. The second information may include inference information and actual information. In this case, the inference information includes at least one of the following: an inference beam determined based on the AI / ML model, or the beam quality corresponding to the inference beam. In addition, the actual information includes at least one of the following: an actual beam determined based on measurements of a set of reference signals associated with the report, or the beam quality corresponding to the actual beam. In some embodiments, the beam quality may be RSRP. Alternatively, the beam quality may be SINR. In some embodiments, the beam quality may be defined by a 7-bit value. The beam quality may be -140 dBm to -44 dBm with a step size of 1 dB. In some embodiments, the bit fields corresponding to the inference beam and the beam quality corresponding to the inference beam are before the bit fields corresponding to the actual beam and the beam quality corresponding to the actual beam. Alternatively, the bit fields corresponding to the inference beam and the beam quality corresponding to the inference beam are after the bit fields corresponding to the actual beam and the beam quality corresponding to the actual beam.

[0107] In some embodiments, the report may be associated with a third type of information. In this case, the third type of information may include a second type of parameters of the trained AI / ML model updated based on the AI / ML model. For example, the third type of information may mainly include the amount of change in the trained AI / ML model compared to the AI / ML model, e.g., the weight factor and the bias factor. This facilitates integrated analysis of the model on the network device side. Alternatively, the terminal device 110-1 does not need to report the trained AI / ML model. In this case, the terminal device 110-1 only needs to report K optimal beams.

[0108] Figure 6 is a flowchart of an exemplary method 600 according to an embodiment of the present disclosure. Method 600 can be implemented in any suitable device. For illustrative purposes only, Method 600 can be implemented in a network device 120-1 as shown in Figure 1.

[0109] In block 610, the network device 120 receives a capability report from the terminal device 110-1. The capability report includes one or more capabilities, each of which indicates that the terminal device 110-1 supports an AI / ML model.

[0110] In some embodiments, capability may indicate the ability to support AI / ML. In some embodiments, capability may indicate the ability to support beam management based on AI / ML. Alternatively or additionally, capability may indicate the ability to support beam prediction in the spatial domain based on AI / ML. Additionally, capability may indicate that terminal device 110-1 supports the ability to support online training. For example, fine-tuning is applicable to online training. It should be noted that other suitable methods are also applicable to online training. Capability may also indicate the index of AI / ML models supported by terminal device 110-1. In some other embodiments, capability may indicate a first time delay, which is the minimum time required for terminal device 110-1 to detect a fault in the AI / ML model. Thus, network device 120 can set the corresponding AI / ML model and associated activation parameters for terminal device 110-1.

[0111] In other embodiments, capability may represent a second time delay, which is the minimum time required for online training of the AI / ML model on the terminal device 110-1 side. This facilitates integrated analysis of the AI / ML model.

[0112] In block 620, the network device 120 transmits one or more settings of the data processing model to the terminal device 110-1. The settings may be transmitted via upper-layer signaling. For example, the settings may be transmitted via RRC signaling.

[0113] In some embodiments, the one or more settings may include a first setting. In this case, the first setting may indicate an index of the AI / ML model. Additionally, the first setting may indicate a first type of parameter of the AI / ML model. In this case, the first type of parameter may include structural parameters of the AI / ML model. For example, the structural parameter may indicate a deep neural network (DNN) of the AI / ML model. Alternatively, the structural parameter may indicate a convolutional neural network (CNN) of the AI / ML model. The structural parameter may also indicate the number of layers of the AI / ML model. The structural parameter may indicate the type of layers of the AI / ML model. Additionally, the structural parameter may indicate the number of neurons of the AI / ML model. In some other embodiments, the first type of parameter may include a factor of the AI / ML model. For example, the factor may be a weight factor. Alternatively or additionally, the factor may be a bias factor. Additionally or alternatively, the first setting may indicate a first type of parameter. In other words, terminal device 110-1 may be configured to have an AI / ML model and corresponding parameters of a first type. The parameters of the first type may include the input data format of the AI / ML model. For example, the input data format may include the number of rows and columns of the input data. Additionally, the input data format may include the units of the input data. The input data format may also include the interpretation of the input data.

[0114] Additionally, the first type of parameter may include the output data format of the AI / ML model. Similarly, the output data format may include the number of rows and columns of the output data. Additionally, the output data format may include the units of the output data. The output data format may also include the interpretation of the output data.

[0115] In some embodiments, the first type of parameters may include pre-processing parameters for the AI / ML model. The first type of parameters may also include post-processing parameters for the AI / ML model. For example, the first type of parameters may include standardization coefficients. The term "standardization coefficient" may refer to flattening the input and output data to the same distribution by performing a normalization operation on the input data, thereby accelerating the training network. For example, in the case of maximum value-based normalization, all input data may be divided by the maximum value, which is the standardization coefficient.

[0116] Additionally, the first setting may also include a first activation parameter. The first activation parameter can be used to enable the terminal device 110-1 to detect faults in the AI / ML model. For example, if the first setting does not include the first activation parameter, the terminal device 110-1 does not have to perform fault detection in the AI / ML model. Alternatively, if the first activation parameter is set to disabled (e.g., set to "0"), the terminal device 110-1 does not have to perform fault detection in the AI / ML model. In other embodiments, if the first activation parameter is set to enabled (e.g., set to "1"), the terminal device 110-1 may perform fault detection in the AI / ML model. In this way, the terminal device can determine the AI / ML model and its executable behavior.

[0117] In other embodiments, the first setting may include a third activation parameter. The third activation parameter can be used to enable the terminal device 110-1 to perform online training of the AI / ML model. For example, if the first setting does not include the third activation parameter, the terminal device 110-1 does not have to perform online training of the AI / ML model. Alternatively, if the third activation parameter is set to disabled (e.g., set to "0"), the terminal device 110-1 does not have to perform online training of the AI / ML model. In other embodiments, if the third activation parameter is set to enabled (e.g., set to "1"), the terminal device 110-1 may perform online training of the AI / ML model. Thus, the terminal device can perform online training of the AI / ML model.

[0118] In block 630, the network device 120 transmits an instruction to trigger a report for obtaining information associated with the data processing model. In some embodiments, the CSI report may be a periodic CSI report configurable by RRC signaling. Alternatively, the CSI report may be a semi-persistent CSI report. In this case, such a CSI report can be activated by a media access control element (MAC CE) from the network device 120. In other embodiments, the CSI report may be an aperiodic CSI report or an SP-CSI report. In this case, the CSI report may be triggered by downlink control information (DCI) from the network device 120.

[0119] In some embodiments, the network device 120 may transmit the CSI report configuration. For example, the CSI report configuration may be transmitted via a higher-layer configuration of CSI-ReportConfig. The CSI report configuration may indicate that the CSI report is associated with one or more CSI-RS or SSB resource sets. In some embodiments, the CSI-RS resource set may be periodic. Alternatively, the CSI-RS resource set may be semi-permanent. In other embodiments, the CSI-RS resource set may be aperiodic. In some embodiments, each CSI-RS resource in the CSI-RS resource set may correspond to a beam. The CSI report configuration may also indicate an index of an AI / ML model, meaning that the CSI report is associated with an AI / ML model. Thus, the terminal device can know which AI / ML model the CSI report is applied to.

[0120] The network device 120 may transmit a set of reference signals to the terminal device 110-1. The network device 120 may transmit the set of reference signals based on the CSI-RS resource set.

[0121] Network device 120 may receive CSI reports from terminal device 110-1. In some embodiments, the CSI report may include first information indicating whether or not the AI / ML model has failed. The first information may be a report quantity. For example, the first information may occupy a bit field in the CSI report. For illustrative purposes, if the first information indicates "1", it means that an AI / ML model failure has occurred. If the first information indicates "0", it means that no AI / ML model failure has occurred. Thus, by reporting the first information, the terminal device can notify the network device whether or not the AI / ML model has failed.

[0122] In other embodiments, the CSI report may indicate measurement results. For example, the CSI report may include one of CRI-RSRP, SSBRI-RSRP, CRI-SINR, or SSBRI-SINR. The CSI report may be associated with a second activation parameter, which may be used to indicate that the associated CSI report is used to detect failures in the AI / ML model. In this case, if a failure is detected in the AI / ML model, the terminal device 110-1 may not report anything, in other words, the terminal device 110-1 may not transmit a CSI report. Transmission of the CSI report may be skipped. At the same time, if there are no other competing CSI reports in the time domain, the network device 120 may not receive any CSI reports. In this case, the network device 120 may determine that an AI / ML model failure has occurred. Alternatively, if no faults are detected in the AI / ML model, the terminal device 110-1 may report K (K>0) optimal beams and their corresponding beam qualities, i.e., K CRIs and their corresponding L1-RSRPs, in the CSI report. Thus, since the CSI report is associated with a second activation parameter, the terminal device also knows that the CSI report is used to detect AI / ML model faults, even though the report quality is set to "CRI-RSRP".

[0123] Alternatively, if a failure is detected in the AI / ML model, the CSI report may include a bit field indicating a predetermined value. For example, if a failure occurs in the AI / ML model, the values ​​of the bit field corresponding to CRI and / or the L1-RSRP field in the CSI report may all be set to "1".

[0124] In some embodiments, the network device 120 may transmit a second setting to the terminal device 110-1. The second setting may indicate a time offset. In this case, the terminal device 110-1 may transmit the CSI report on a slot determined based on the time offset. In some embodiments, the time offset may depend on a first time delay, which is the minimum time required for the terminal device 110-1 to detect a fault in the AI / ML model. Alternatively, the time offset may depend on a second time delay, which is the minimum time required for the terminal device 110-1 to train the AI / ML model online. The time offset may indicate on which slot the terminal device 110-1 will transfer the report amount. In some embodiments, the time offset may refer to a time offset by a higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. Alternatively, the time offset may be set via an RRC setting independently of the higher-level setting of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. In this case, if the first time offset is set within the RRC settings, the terminal device 110-1 may ignore the value provided by the higher-level settings of CSI-ReportPeriodicityAndOffset or reportSlotOffsetList. In this case, the terminal device can know when to perform the report.

[0125] In some embodiments, the network device 120 may receive multiple CSI reports. In this case, each of the multiple CSI reports may indicate a corresponding priority. The priority value may be determined from a set of values, which may include 0, 1, and 2. It should be noted that the set of values ​​may also include other values. Thus, if multiple CSI reports carrying different content collide, the terminal device knows which CSI report to send first.

[0126] Alternatively, the network device 120 may determine whether a fault has occurred on the AI / ML model. In this case, a new reporting quantity, e.g., second information, may be introduced. The second information may include inference information and actual information. In this case, the inference information includes at least one of the following: an inference beam determined based on the AI / ML model, or the beam quality corresponding to the inference beam. Additionally, the actual information includes at least one of the following: an actual beam determined based on measurements of a set of reference signals associated with the report, or the beam quality corresponding to the actual beam. In some embodiments, the beam quality may be RSRP. Alternatively, the beam quality may be SINR. In some embodiments, the beam quality may be defined by a 7-bit value. The beam quality may be -140 dBm to -44 dBm with a step size of 1 dB. In some embodiments, the bit fields corresponding to the inference beam and the beam quality corresponding to the inference beam are before the bit fields corresponding to the actual beam and the beam quality corresponding to the actual beam. Alternatively, the bit fields corresponding to the inference beam and the beam quality corresponding to the inference beam are after the bit fields corresponding to the actual beam and the beam quality corresponding to the actual beam.

[0127] In some embodiments, the report may be associated with a third type of information. In this case, the third type of information may include a second type of parameters of the trained AI / ML model updated based on the AI / ML model. For example, the third type of information may mainly include the amount of change in the trained AI / ML model compared to the AI / ML model, e.g., the weight factor and the bias factor. This facilitates integrated analysis of the model on the network device side. Alternatively, the terminal device 110-1 does not need to report the trained AI / ML model. In this case, the terminal device 110-1 only needs to report K optimal beams.

[0128] In some embodiments, the terminal device comprises a circuit which reports to a network device one or more capabilities of the terminal device, the one or more capabilities which at least indicate that the terminal device supports a data processing model, the circuit which is configured to receive one or more settings associated with the data processing model from the network device, and to receive instructions from the network device to trigger a report for obtaining information associated with the data processing model.

[0129] In some embodiments, the one or more capabilities further include at least one of the following: the ability to support artificial intelligence (AI) or machine learning (ML); the ability to support beam management based on AI or ML; the ability to support beam prediction in a spatial domain based on AI or ML; the ability to support online training; the index of the data processing model; a first time delay which is the minimum time required to detect a fault in the data processing model in the terminal device; or a second time delay which is the minimum time required to conduct online training of the data processing model in the terminal device.

[0130] In some embodiments, the one or more settings further indicate at least one of the following: an index of the data processing model, a first type of parameter of the data processing model, a first enablement parameter used to enable the terminal device to detect failures in the data processing model, a second enablement parameter used to enable the terminal device to perform online training, or the time offset used to indicate the terminal device to transmit the report on a slot determined according to the time offset.

[0131] In some embodiments, the first type of parameter includes at least one of the following: a structural parameter of the data processing model, a weighting factor of the data processing model, a bias factor of the data processing model, an input data format of the data processing model, an output data format of the data processing model, a preprocessing parameter of the data processing model, or a postprocessing parameter of the data processing model.

[0132] In some embodiments, the time offset is determined according to a first time delay or a second time delay, the first time delay being the minimum time required to detect a failure in the data processing model in the terminal device, and the second time delay being the minimum time required to train the data processing model online in the terminal device.

[0133] In some embodiments, the terminal device comprises a circuit configured to ignore another time offset indicated in another setting, in accordance with the determination that the terminal device is configured to have the time offset.

[0134] In some embodiments, the information associated with the data processing model includes one of the following: first information indicating whether or not a failure has occurred in the data processing model; second information including inference information and actual information; or third information indicating a second type of parameter of a trained data processing model updated based on the data processing model.

[0135] In some embodiments, the inference information includes at least one of an inference beam determined based on the data processing model, or a beam quality corresponding to the inference beam, and the actual information includes at least one of an actual beam determined based on measurements of a set of reference signals associated with the report, or a beam quality corresponding to the actual beam.

[0136] In some embodiments, the beam quality includes at least one of the following: reference signal received power (RSRP) or signal-to-interference noise ratio (SINR).

[0137] In some embodiments, the beam quality corresponding to the inference beam is defined by a 7-bit value in the range of -140 to -44 dBm with a step size of 1 dB, and the beam quality corresponding to the actual beam is defined by a 7-bit value in the range of -140 to -44 dBm with a step size of 1 dB.

[0138] In some embodiments, the inference beam and the bit field corresponding to the beam quality corresponding to the inference beam are located before or after the actual beam and the bit field corresponding to the beam quality corresponding to the actual beam.

[0139] In some embodiments, the second type of parameter includes at least one of the structural parameters of the trained data processing model, the weight factors of the trained data processing model, or the bias factors of the trained data processing model.

[0140] In some embodiments, the terminal device comprises a circuit configured to transmit the report to the network device for obtaining the information associated with the data processing model.

[0141] In some embodiments, the terminal device comprises a circuit configured to transmit the report to the network device in accordance with a determination that a fault has been detected in the data processing model, and the bit field corresponding to the report is set to a predetermined number.

[0142] In some embodiments, the terminal device includes a circuit configured to skip the transmission of the report in accordance with a decision in the data processing model that a fault has been detected.

[0143] In some embodiments, the terminal device includes a circuit which, upon determination that it is necessary to transmit a second uplink information, determines whether the second uplink information conflicts with the report, and upon determination that the second uplink information conflicts with the report, compares the first priority of the report with the second priority of the second uplink information, and upon determination that the first priority is higher than the second priority, transmits the report to the network device.

[0144] In some embodiments, the first priority and the second priority are determined from a set of values ​​including 0, 1, or 2.

[0145] In some embodiments, the terminal device comprises a circuit configured to determine that a failure has occurred in the data processing model, according to the determination that at least one condition has been met.

[0146] In some embodiments, the at least one condition includes a first condition in which a first target beam determined based on the data processing model is different from a second target beam determined based on measurements of a set of reference signals associated with the report; a second condition in which the difference between the beam quantity of the first target beam and the beam quantity of the second target beam exceeds a first threshold; or a third condition in which the number of times the first or second condition is met exceeds a second threshold.

[0147] In some embodiments, the report includes at least one of periodic channel status information (CSI) reports, semi-persistent CSI reports, or aperiodic CSI reports.

[0148] In some embodiments, the network device comprises a circuit that receives capability reports from terminal devices, the reports including one or more capabilities of the terminal devices, the one or more capabilities of which at least indicate that the terminal devices support a data processing model, the circuit is configured to transmit one or more settings associated with the data processing model to the terminal devices, and to transmit instructions to the terminal devices to trigger a report for obtaining information associated with the data processing model.

[0149] In some embodiments, the one or more capabilities further include at least one of the following: the ability to support artificial intelligence (AI) or machine learning (ML); the ability to support beam management based on AI or ML; the ability to support beam prediction in a spatial domain based on AI or ML; the ability to support online training; the index of the data processing model; a first time delay which is the minimum time required to detect a fault in the data processing model in the terminal device; or a second time delay which is the minimum time required to conduct online training of the data processing model in the terminal device.

[0150] In some embodiments, the one or more settings further indicate at least one of the following: an index of the data processing model, a first type of parameter of the data processing model, a first enablement parameter used to enable the terminal device to detect failures in the data processing model, a second enablement parameter used to enable the terminal device to perform online training, or the time offset used to indicate the terminal device to transmit the report on a slot determined according to the time offset.

[0151] In some embodiments, the first type of parameter includes at least one of the following: a structural parameter of the data processing model, a weighting factor of the data processing model, a bias factor of the data processing model, an input data format of the data processing model, an output data format of the data processing model, a preprocessing parameter of the data processing model, or a postprocessing parameter of the data processing model.

[0152] In some embodiments, the time offset is determined according to a first time delay or a second time delay, the first time delay being the minimum time required to detect a failure in the data processing model in the terminal device, and the second time delay being the minimum time required to train the data processing model online in the terminal device.

[0153] In some embodiments, the report is associated with one or more of the settings associated with the data processing model.

[0154] In some embodiments, the information associated with the data processing model includes one of the following: first information indicating whether or not a failure has occurred in the data processing model; second information including inference information and actual information; or third information indicating a second type of parameter of a trained data processing model updated based on the data processing model.

[0155] In some embodiments, the inference information includes at least one of an inference beam determined based on the data processing model, or a beam quality corresponding to the inference beam, and the actual information includes at least one of an actual beam determined based on measurements of a set of reference signals associated with the report, or a beam quality corresponding to the actual beam.

[0156] In some embodiments, the beam quality includes at least one of the following: reference signal received power (RSRP) or signal-to-interference noise ratio (SINR).

[0157] In some embodiments, the beam quality corresponding to the inference beam is defined by a 7-bit value in the range of -140 to -44 dBm with a step size of 1 dB, and the beam quality corresponding to the actual beam is defined by a 7-bit value in the range of -140 to -44 dBm with a step size of 1 dB.

[0158] In some embodiments, the inference beam and the bit field corresponding to the beam quality corresponding to the inference beam are located before or after the actual beam and the bit field corresponding to the beam quality corresponding to the actual beam.

[0159] In some embodiments, the second type of parameter includes at least one of the structural parameters of the trained data processing model, the weight factors of the trained data processing model, or the bias factors of the trained data processing model.

[0160] In some embodiments, the network device comprises a circuit configured to receive reports from the terminal device for obtaining the information associated with the data processing model.

[0161] In some embodiments, the network device comprises a circuit configured to transmit the report to the network device in accordance with a determination that a fault has been detected in the data processing model, wherein a bit field corresponding to the report is set to a predetermined number.

[0162] Figure 7 is a schematic block diagram of a device 700 suitable for implementing an embodiment of the present disclosure. The device 700 may be considered as another exemplary embodiment of the terminal device 110 as shown in Figure 1. Thus, the device 700 can be implemented in or as at least part of the terminal device 110. Alternatively, the device 700 may be considered as another exemplary embodiment of the network device 120 as shown in Figure 1. Thus, the device 700 may be implemented in or as at least part of the network device 120.

[0163] As illustrated, the device 700 comprises a processor 710, a memory 720 coupled to the processor 710, appropriate transmitters (TX) and receivers (RX) 740 coupled to the processor 710, and a communication interface coupled to the TX / RX 740. The memory 720 stores at least a portion of the program 730. The TX / RX 740 is used for bidirectional communication. The TX / RX 740 has at least one antenna to facilitate communication, although the access node referred to herein may actually have multiple antennas. The communication interface may represent any interface necessary for communication with other network elements, such as an X2 interface for bidirectional communication between eNBs, an S1 interface for communication between a mobility management entity (MME) / serving gateway (S-GW) and an eNB, an Un interface for communication between an eNB and a relay node (RN), or a Uu interface for communication between an eNB and a terminal device.

[0164] It is assumed that program 730 includes program instructions that, when executed by the associated processor 710, enable the device 700 to operate according to embodiments of the present disclosure, as described herein with reference to Figures 2 to 6. Embodiments of the present disclosure may be implemented by computer software executable by the processor 710 of the device 700, by hardware, or by a combination of software and hardware. The processor 710 may be configured to implement various embodiments of the present disclosure. Furthermore, a combination of the processor 710 and memory 720 may form processing means 750 suitable for implementing various embodiments of the present disclosure.

[0165] Memory 720 may be of any type suitable for a local technology network and may be implemented using any suitable data storage technology, such as non-temporary computer-readable storage media, semiconductor-based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory and removable memory, as non-limiting examples. Although only one memory 720 is shown in device 700, several physically different memory modules may be present in device 700. Processor 710 may be of any type suitable for a local technology network and may include, as non-limiting examples, one or more of general-purpose computers, dedicated computers, microprocessors, digital signal processors (DSPs), and processors based on multicore processor architectures. Device 700 may have multiple processors, for example, application-specific integrated circuit chips that are time-dependent to a clock that synchronizes the main processor.

[0166] Overall, various embodiments of the Disclosure may be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some embodiments may be implemented in hardware, while others may be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. Although various embodiments of the Disclosure are illustrated and described using block diagrams, flowcharts, or any other pictorial representation, it should be understood that any blocks, devices, systems, techniques, or methods described herein may be implemented, in non-limiting examples, in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or any combination thereof.

[0167] This disclosure also provides at least one computer program product tangibly stored on a non-temporary computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions contained in a program module, which are executed within a device on a target real or virtual processor to perform the processes or methods described above with reference to Figures 2-6. Generally, a program module includes routines, programs, libraries, objects, classes, components, data structures, etc., that perform a specific task or realize a specific abstract data type. In various embodiments, the functions of program modules may be combined or separated among program modules as needed. The machine-executable instructions of a program module may be executed within a local or distributed device. In a distributed device, the program module may reside in both local and remote storage media.

[0168] Program code for performing the methods of this disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a dedicated computer, or other programmable data processing device, and when executed by the processor or controller, the program code may implement the functions / operations specified in the flowcharts and / or block diagrams. The program code may run entirely on a machine, partially on a machine, as an independent software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0169] The program code described above may be implemented on a machine-readable medium, which may be any tangible medium that can contain or store programs used by or associated with an instruction execution system, device, or apparatus. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or apparatus, or any suitable combination of the aforementioned mediums. More specific examples of machine-readable storage media may include electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0170] While the operations have been described in a specific order, it should not be understood that, in order to obtain the desired results, these operations must be performed in the specific order shown, or in a sequential order, or that all of the described operations must be performed. In some cases, multitasking and parallel processing may be advantageous. Similarly, while some specific implementation details are included in the above discussion, these should not be interpreted as limitations on the scope of this disclosure, but rather as descriptions of features that may be specific to a particular embodiment. Some features described in the context of individual embodiments may be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may be implemented separately or in any suitable subcombination in multiple embodiments.

[0171] While this disclosure has been described in language specific to structural features and / or methodological behavior, it should be understood that the disclosure as defined in the attached claims is not necessarily limited to the specific features or behaviors described above. Rather, the specific features and behaviors described above are disclosed as exemplary forms of implementing the claims.

[0172] As used herein, the term “terminal device” refers to any device having wireless or wired communication capabilities. Examples of terminal devices include user equipment (UE), personal computers, desktops, mobile phones, cellular phones, smartphones, personal digital assistants (PDAs), portable computers, tablets, wearable devices, Internet of Things (IoT) devices, ultra-reliable low-latency communication (URLLC) devices, any Internet of Things (IoT) devices, machine-type communication (MTC) devices, in-vehicle devices for V2X communication where X represents pedestrians, vehicles, or infrastructure / networks, devices for integrated access and integrated access and backhaul (IAB), satellite-borne or aircraft-borne vehicles within non-terrestrial networks (NTN) including high-altitude platforms (HAP) encompassing satellites and unmanned aircraft systems (UAS), extended reality (XR) devices including different types of reality such as augmented reality (AR), mixed reality (MR), and virtual reality (VR), unmanned aerial vehicles (UAVs), which are aircraft without human pilots and are commonly referred to as drones, and high-speed trains (HSTs). The “Terminal device” includes, but is not limited to, devices on a train, or image acquisition devices such as digital cameras, sensor game devices, music storage and playback devices, or internet-connected home appliances that enable wireless or wired internet access and browsing. The “Terminal device” may further have “multicast / broadcast” capabilities to support V2X applications, transparent IPv4 / IPv6 multicast distribution, IPTV, smart TV, wireless services, wireless software distribution, group communications, and Iota applications, where public safety and mission are of paramount importance. It may also incorporate one or more Subscriber Identity Modules (SIMs), known as multi-SIMs.The term "terminal device" may be used interchangeably with UE, mobile station, subscriber station, mobile terminal, user terminal, or radio device.

[0173] The term "network device" refers to a device that can provide or host a cell or coverage on which terminal devices can communicate. Examples of network devices include, but are not limited to, low-power nodes such as Node B (Node or NB), Evolutionary Node (Node or eNB), Next Generation Node (gNB), Transmit / Receive Point (TRP), Remote Radio Unit (RRU), Radio Head (RH), Remote Radio Head (RRH), IAB Node, Femtonode, Piconode, and Reconfigurable Intelligent Surface (RIS).

[0174] Terminal devices or network devices may possess artificial intelligence (AI) or machine learning capabilities. Generally, this includes a trained model derived from a large amount of data collected for a specific function, which can be used to predict certain information.

[0175] Terminal or network devices may operate on several frequency ranges, such as FR1 (410 MHz to 7125 MHz), FR2 (24.25 GHz to 71 GHz), frequency bands greater than 100 GHz, and terahertz (THz). Furthermore, they can operate on permitted / unpermitted / shared spectrum. Terminal devices may have two or more connections to network devices under Multi-Radio Dual Connectivity (MR-DC) application scenarios. Terminal or network devices can operate in full-duplex, flexible-duplex, or cross-split-duplex modes.

[0176] Embodiments of this disclosure may be implemented, for example, in test equipment such as signal generators, signal analyzers, spectrum analyzers, network analyzers, test terminal devices, test network devices, and channel emulators.

[0177] Embodiments of the present disclosure may be implemented in accordance with any generation of communication protocols currently known or to be developed in the future. Examples of communication protocols include, but are not limited to, first-generation (1G), second-generation (2G), 2.5G, 2.75G, third-generation (3G), fourth-generation (4G), 4.5G, fifth-generation (5G) communication protocols, 5.5G, 5G-Advanced networks, or sixth-generation (6G) networks.

Claims

1. A method performed by a terminal device, To transmit capability information corresponding to artificial intelligence (AI) / machine learning (ML) models to network devices, The network device receives the Radio Resource Control (RRC) configuration, The network device transmits a first channel status information (CSI) report based on the RRC configuration. Includes, The aforementioned first CSI report includes first information, The first information is determined based on whether the first prediction related to the AI / ML model is considered to be an accurate prediction. The first prediction is not considered accurate if the CSI-RS resource with the best beam, based on reference signal received power (RSRP) measurements of multiple resources, is not mapped to the predicted CSI-RS resource being reported. method.

2. The first CSI report is transmitted based on a first priority value when the time occupancy of a first physical channel scheduled to carry the first CSI report and the time occupancy of a second physical channel scheduled to carry the second CSI report overlap in at least one orthogonal frequency division multiplexing (OFDM) symbol on the same carrier. The method according to claim 1.

3. The parameter k used to calculate the first priority value of the first CSI report is equal to zero, If the first CSI report does not contain the first information, then the parameter k is equal to 1. The method according to claim 2.

4. A method performed by a network device, Receiving capability information corresponding to artificial intelligence (AI) / machine learning (ML) models from terminal devices, The terminal device is to transmit a wireless resource control (RRC) configuration, The terminal device receives a first channel status information (CSI) report based on the radio resource control (RRC) configuration, Includes, The aforementioned first CSI report includes first information, The first information is determined based on whether the first prediction related to the AI / ML model is considered to be an accurate prediction. The first prediction is not considered accurate if the CSI-RS resource with the best beam, based on reference signal received power (RSRP) measurements of multiple resources, is not mapped to the predicted CSI-RS resource being reported. method.

5. The first CSI report is transmitted based on a first priority value when the time occupancy of a first physical channel scheduled to carry the first CSI report and the time occupancy of a second physical channel scheduled to carry the second CSI report overlap in at least one orthogonal frequency division multiplexing (OFDM) symbol on the same carrier. The method according to claim 4.

6. The parameter k used to calculate the first priority value of the first CSI report is equal to zero, If the first CSI report does not contain the first information, then the parameter k is equal to 1. The method according to claim 5.

7. A means for transmitting capability information corresponding to an artificial intelligence (AI) / machine learning (ML) model to a network device, Means for receiving a wireless resource control (RRC) configuration from the aforementioned network device, The network device includes means for transmitting a first channel status information (CSI) report based on the RRC configuration, Includes, The aforementioned first CSI report includes first information, The first information is determined based on whether the first prediction related to the AI / ML model is considered to be an accurate prediction. The first prediction is not considered accurate if the CSI-RS resource with the best beam, based on reference signal received power (RSRP) measurements of multiple resources, is not mapped to the predicted CSI-RS resource being reported. Terminal device.

8. The first CSI report is transmitted based on a first priority value if the time occupancy of a first physical channel scheduled to carry the first CSI report and the time occupancy of a second physical channel scheduled to carry the second CSI report overlap in at least one orthogonal frequency division multiplexing (OFDM) symbol on the same carrier. The terminal device according to claim 7.

9. The parameter k used to calculate the first priority value of the first CSI report is equal to zero, If the first CSI report does not contain the first information, then the parameter k is equal to 1. The terminal device according to claim 8.

10. A means for receiving capability information corresponding to an artificial intelligence (AI) / machine learning (ML) model from a terminal device, The terminal device includes means for transmitting a wireless resource control (RRC) configuration, The terminal device has means for receiving a first channel status information (CSI) report based on a radio resource control (RRC) configuration, Includes, The aforementioned first CSI report includes first information, The first information is determined based on whether the first prediction related to the AI / ML model is considered to be an accurate prediction. The first prediction is not considered accurate if the CSI-RS resource with the best beam, based on reference signal received power (RSRP) measurements of multiple resources, is not mapped to the predicted CSI-RS resource being reported. Network device.

11. The first CSI report is transmitted based on a first priority value when the time occupancy of a first physical channel scheduled to carry the first CSI report and the time occupancy of a second physical channel scheduled to carry the second CSI report overlap in at least one orthogonal frequency division multiplexing (OFDM) symbol on the same carrier. The network device according to claim 10.

12. The parameter k used to calculate the first priority value of the first CSI report is equal to zero, If the first CSI report does not contain the first information, then the parameter k is equal to 1. The network device according to claim 11.

Images