Devices, methods and computer readable medium for communication
By configuring and applying prediction windows for AI or ML model life cycle management, the terminal device enhances CSI prediction in communication systems, addressing the lack of clear procedures in existing technologies.
Patent Information
- Application Number
- PCT/CN2023/139661
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-12-18
- Publication Date
- 2025-06-26
AI Technical Summary
Existing technologies lack clear signaling and procedures for managing prediction windows for the life cycle management of artificial intelligence (AI) or machine learning (ML) models in time domain channel state information (CSI) prediction.
A terminal device receives a configuration for at least one prediction window from a network device, which specifies the duration, starting time, or ending time of the prediction window with respect to a reference time, and applies this configuration to manage the life cycle of the AI or ML model.
This solution enables effective CSI prediction in the time domain by providing a structured framework for managing prediction windows, thereby enhancing the performance and efficiency of AI or ML models in communication systems.
Smart Images

Figure CN2023139661_26062025_PF_FP_ABST
Abstract
Description
DEVICES, METHODS AND COMPUTER READABLE MEDIUM FOR COMMUNICATIONTECHNICAL FIELD
[0001] Embodiments of the present disclosure generally relate to the field of telecommunication, and in particular, to devices, methods and computer readable medium for communication.BACKGROUND
[0002] In order to achieve channel state information (CSI) feedback enhancement based on artificial intelligence (AI) or machine learning (ML) technology, a terminal device may apply an AI or ML model for CSI prediction in time domain. A prediction window would be necessary for life cycle management (LCM) of the AI or ML model.SUMMARY
[0003] In general, example embodiments of the present disclosure provide devices, methods and computer readable medium for communication.
[0004] In a first aspect, there is provided a terminal device. The terminal device comprises a processor. The processor is configured to cause the terminal device to: receive, from a network device, a first configuration for at least one prediction window for life cycle management (LCM) of an artificial intelligence (AI) or machine learning (ML) model, wherein the first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to reference time, or first ending time of each of the at least one prediction window with respect to the reference time; and apply one of the at least one prediction window.
[0005] In a second aspect, there is provided a network device. The network device comprises a processor. The processor is configured to cause the network device to: determine a first configuration for at least one prediction window for LCM of an AI or ML model, wherein the first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to reference time, or first ending time of each of the at least one prediction window with respect to the reference time; and transmit the first configuration to a terminal device.
[0006] In a third aspect, there is provided a method for communication. The method comprises: receiving, from a network device, a first configuration for at least one prediction window for LCM of an AI or ML model, wherein the first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to reference time, or first ending time of each of the at least one prediction window with respect to the reference time; and applying one of the at least one prediction window.
[0007] In a fourth aspect, there is provided a method for communication. The method comprises: determining a first configuration for at least one prediction window for LCM of an AI or ML model, wherein the first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to reference time, or first ending time of each of the at least one prediction window with respect to the reference time; and transmitting the first configuration to a terminal device.
[0008] In a fifth aspect, there is provided a computer readable medium having instructions stored thereon. The instructions, when executed on at least one processor of a device, cause the device to perform the method according to the third aspect or the fourth aspect.
[0009] It is to be understood that the summary section is not intended to identify key or essential features of embodiments of the present disclosure, nor is it intended to be used to limit the scope of the present disclosure. Other features of the present disclosure will become easily comprehensible through the following description.BRIEF DESCRIPTION OF THE DRAWINGS
[0010] Through the more detailed description of some embodiments of the present disclosure in the accompanying drawings, the above and other objects, features and advantages of the present disclosure will become more apparent, wherein:
[0011] Fig. 1 illustrate an example communication network in which embodiments of the present disclosure can be implemented;
[0012] Fig. 2 illustrates a signaling chart illustrating an example process for communication in accordance with some embodiments of the present disclosure;
[0013] Figs. 3A, 3B, 3C and 3D illustrate an example of a prediction window and an observation window in accordance with some embodiments of the present disclosure, respectively;
[0014] Figs. 4A, 4B, 4C and 4D illustrate an example of reporting of predicted CSI instances in accordance with some embodiments of the present disclosure, respectively;
[0015] Fig. 5 illustrates a flowchart of an example method in accordance with some embodiments of the present disclosure;
[0016] Fig. 6 illustrates a flowchart of an example method in accordance with some embodiments of the present disclosure; and
[0017] Fig. 7 is a simplified block diagram of a device that is suitable for implementing embodiments of the present disclosure.
[0018] Throughout the drawings, the same or similar reference numerals represent the same or similar element.DETAILED DESCRIPTION
[0019] Principle of the present disclosure will now be described with reference to some example embodiments. It is to be understood that these embodiments are described only for the purpose of illustration and help those skilled in the art to understand and implement the present disclosure, without suggesting any limitations as to the scope of the disclosure. The disclosure described herein can be implemented in various manners other than the ones described below.
[0020] In the following description and claims, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure belongs.
[0021] As used herein, the term “terminal device” refers to any device having wireless or wired communication capabilities. Examples of the terminal device include, but not limited to, user equipment (UE) , personal computers, desktops, mobile phones, cellular phones, smart phones, personal digital assistants (PDAs) , portable computers, tablets, wearable devices, internet of things (IoT) devices, Ultra-reliable and Low Latency Communications (URLLC) devices, Internet of Everything (IoE) devices, machine type communication (MTC) devices, device on vehicle for V2X communication where X means pedestrian, vehicle, or infrastructure / network, devices for Integrated Access and Backhaul (IAB) , Small Data Transmission (SDT) , mobility, Multicast and Broadcast Services (MBS) , positioning, dynamic / flexible duplex in commercial networks, reduced capability (RedCap) , Space borne vehicles or Air borne vehicles in Non-terrestrial networks (NTN) including Satellites and High Altitude Platforms (HAPs) encompassing Unmanned Aircraft Systems (UAS) , eXtended Reality (XR) devices including different types of realities such as Augmented Reality (AR) , Mixed Reality (MR) and Virtual Reality (VR) , the unmanned aerial vehicle (UAV) commonly known as a drone which is an aircraft without any human pilot, devices on high speed train (HST) , or image capture devices such as digital cameras, sensors, gaming devices, music storage and playback appliances, or Internet appliances enabling wireless or wired Internet access and browsing and the like. The ‘terminal device’ can further has ‘multicast / broadcast’ feature, to support public safety and mission critical, V2X applications, transparent IPv4 / IPv6 multicast delivery, IPTV, smart TV, radio services, software delivery over wireless, group communications and IoT applications. It may also incorporate one or multiple Subscriber Identity Module (SIM) as known as Multi-SIM. The term “terminal device” can be used interchangeably with a UE, a mobile station, a subscriber station, a mobile terminal, a user terminal or a wireless device.
[0022] The term “network device” refers to a device which is capable of providing or hosting a cell or coverage where terminal devices can communicate. Examples of a network device include, but not limited to, a Node B (NodeB or NB) , an evolved NodeB (eNodeB or eNB) , a next generation NodeB (gNB) , a transmission reception point (TRP) , a remote radio unit (RRU) , a radio head (RH) , a remote radio head (RRH) , an IAB node, a low power node such as a femto node, a pico node, a reconfigurable intelligent surface (RIS) , Network-controlled Repeaters, and the like.
[0023] The terminal device or the network device may have Artificial intelligence (AI) or Machine learning capability. It generally includes a model which has been trained from numerous collected data for a specific function, and can be used to predict some information.
[0024] The terminal or the network device may work on several frequency ranges, e.g. FR1 (410 MHz –7125 MHz) , FR2 (24.25GHz to 71GHz) , frequency band larger than 100GHz as well as Tera Hertz (THz) . It can further work on licensed / unlicensed / shared spectrum. The terminal device may have more than one connection with the network devices under Multi-Radio Dual Connectivity (MR-DC) application scenario. The terminal device or the network device can work on full duplex, flexible duplex and cross division duplex modes.
[0025] The network device may have the function of network energy saving, Self-Organizing Networks (SON) / Minimization of Drive Tests (MDT) . The terminal may have the function of power saving.
[0026] The embodiments of the present disclosure may be performed in test equipment, e.g. signal generator, signal analyzer, spectrum analyzer, network analyzer, test terminal device, test network device, channel emulator.
[0027] The embodiments of the present disclosure may be performed according to any generation communication protocols either currently known or to be developed in the future. Examples of the communication protocols include, but not limited to, the first generation (1G) , the second generation (2G) , 2.5G, 2.75G, the third generation (3G) , the fourth generation (4G) , 4.5G, the fifth generation (5G) communication protocols, 5.5G, 5G-Advanced networks, or the sixth generation (6G) networks.
[0028] As used herein, the singular forms ‘a’ , ‘an’ and ‘the’ are intended to include the plural forms as well, unless the context clearly indicates otherwise. The term ‘includes’ and its variants are to be read as open terms that mean ‘includes, but is not limited to. ’ The term ‘based on’ is to be read as ‘at least in part based on. ’ The term ‘some embodiments’ and ‘an embodiment’ are to be read as ‘at least some embodiments. ’ The term ‘another embodiment’ is to be read as ‘at least one other embodiment. ’ The terms ‘first, ’ ‘second, ’ and the like may refer to different or same objects. Other definitions, explicit and implicit, may be included below.
[0029] In some examples, values, procedures, or apparatus are referred to as ‘best, ’ ‘lowest, ’ ‘highest, ’ ‘minimum, ’ ‘maximum, ’ or the like. It will be appreciated that such descriptions are intended to indicate that a selection among many used functional alternatives can be made, and such selections need not be better, smaller, higher, or otherwise preferable to other selections.
[0030] Fig. 1 illustrates a schematic diagram of an example communication network 100 in which embodiments of the present disclosure can be implemented. As shown in Fig. 1, the communication network 100 comprises a network device 120 and terminal devices 110-1, 110-2…, 110-N served by the network device 120. The serving area of the network device 120 is called as a cell 102. Hereinafter, the terminal devices 110-1, 110-2…, 110-N may be collectively referred to as “terminal devices 110” or individually referred to as “aterminal devices 110” .
[0031] It is to be understood that the number of network devices and terminal devices is only for the purpose of illustration without suggesting any limitations. The communication network 100 may comprise any suitable number of network devices and terminal devices adapted for implementing embodiments of the present disclosure.
[0032] As described above, a prediction window would be necessary for LCM of the AI or ML model. For time domain CSI prediction based on AI or ML, signaling and procedures for aspects of the prediction window are not clear.
[0033] In view of the above, embodiments of the present disclosure provide a solution for communication. In this solution, a terminal device receives, from a network device, a first configuration for at least one prediction window for LCM of AI or ML model. The first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to a reference time, or first ending time of each of the at least one prediction window with respect to a reference time. In turn, the terminal device applies one of the at least one prediction window.
[0034] Hereinafter, principle of the present disclosure will be described with reference to Figs. 2 to 7.
[0035] Fig. 2 illustrates a signaling chart illustrating an example process 200 for communication in accordance with some embodiments of the present disclosure. For the purpose of discussion, the process 200 will be described with reference to Fig. 1. The process 200 may involve the terminal device 110 and the network device 120 in Fig. 1.
[0036] As shown in Fig. 2, the network device 120 determines 210 a first configuration for at least one prediction window for LCM of an AI or ML model. The first configuration for the at least one prediction window indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to a reference time, or first ending time of each of the at least one prediction window with respect to a reference time. Hereinafter, a prediction window is represented by W1.
[0037] In turn, the network device 120 transmits 220 the first configuration for at least one prediction window to the terminal device 110.
[0038] Then, the terminal device 110 applies 230 one of the at least one prediction window.
[0039] With the process 200, the terminal device 110 may determine a prediction window for LCM of an AI or ML model based on the first configuration received from the network device 120. In turn, CSI prediction in time domain based on the AI or ML model may be achieved.
[0040] It shall be understood that although some embodiments of the present disclosure will be described by taking CSI prediction use case as example, the prediction-related configuration of the present disclosure may be applied to use cases in which a set of past historical information is used to as input to generate a set of information to predict the future. For example, the prediction-related configuration of the present disclosure may be applied to the following beam prediction use case: spatial-domain downlink beam prediction for Set A of beams based on measurement results of Set B of beams. For another example, the prediction-related configuration of the present disclosure may be applied to the following beam prediction use case: temporal downlink beam prediction for Set A of beams based on the historic measurement results of Set B of beams.
[0041] In some embodiments, from capability perspective of the terminal device 110, a prediction window may be the number / time distance of predicted CSI / channel measurements applied in AI / ML model training or AI / ML model inference. In other words, a prediction window may be defined by the number of predicted CSI or channel measurements applied in AI / ML model training or AI / ML model inference. Alternatively, a prediction window may be the time distance of predicted CSI or channel measurements applied in AI / ML model training or AI / ML model inference.
[0042] In some embodiments, from configuration perspective of the network device 120, a prediction window may be a configured prediction window for an AI or ML model at the side of the terminal device 110.
[0043] In some embodiments, a model may be used interchangeably with AI, ML, AI / ML model, (AI / ML / auto-) encoder, CSI generation part or UE part / side model, functionality, AI-enabled feature / FG, which means a data driven algorithm that applies AI / ML techniques to generate a set of (AI / ML) outputs based on a set of (AI / ML) inputs.
[0044] In some embodiments, the AI / ML-enabled feature refers to a feature where AI / ML may be used.
[0045] In some embodiments, the LCM of the AI or ML model may comprise at least one of the following: data collection, model training, model deployment, model inference, model monitoring, model updating, functionality or model selection, functionality or model activation, functionality or model deactivation, functionality or model switching, or functionality or model fallback.
[0046] In some embodiments, the model inference refers to a process of using a trained AI or ML model to generate a set of outputs based on a set of inputs.
[0047] In some embodiments, the model monitoring refers to a procedure that monitors the inference performance of the AI or ML model.
[0048] In some embodiments, the first configuration for the at least one prediction window may indicate the duration of each of the at least one prediction window. In such embodiments, the first starting time of each of the at least one prediction window and the first ending time of each of the at least one prediction window may be predefined.
[0049] Alternatively, in some embodiments, the first configuration for the at least one prediction window may indicate the duration of each of the at least one prediction window and the first starting time of each of the at least one prediction window with respect to the reference time.
[0050] Alternatively, in some embodiments, the first configuration for the at least one prediction window may indicate the duration of each of the at least one prediction window and the first ending time of each of the at least one prediction window with respect to the reference time.
[0051] Alternatively, in some embodiments, the first configuration for the at least one prediction window may indicate the first starting time of each of the at least one prediction window with respect to the reference time, and the first ending time of each of the at least one prediction window with respect to the reference time.
[0052] In some embodiments, the terminal device 110 may receive, from the network device 120, a second configuration for at least one observation window for the LCM of the AI or ML model. The second configuration may indicate at least one of the following: a duration of each of the at least one observation window, second starting time of each of the at least one observation window with respect to the reference time, or second ending time of each of the at least one observation window with respect to the reference time. In turn, the terminal device 110 may apply one of the at least one observation window. Hereinafter, an observation window is represented by W2.
[0053] In some embodiments, from capability perspective of the terminal device 110, an observation window may be the number / time distance of historic CSI / channel measurements applied in AI / ML model training or AI / ML model inference.
[0054] In some embodiments, from configuration perspective of the network device 120, an observation window may be a configured observation window for an AI or ML model at the side of the terminal device 110.
[0055] In some embodiments, the second configuration for the at least one observation window may indicate the duration of each of the at least one observation window. In such embodiments, the second starting time of each of the at least one observation window and the second ending time of each of the at least one observation window may be predefined.
[0056] Alternatively, in some embodiments, the second configuration for the at least one observation window may indicate the duration of each of the at least one observation window and the second starting time of each of the at least one observation window with respect to the reference time.
[0057] Alternatively, in some embodiments, the second configuration for the at least one observation window may indicate the duration of each of the at least one observation window and the second ending time of each of the at least one observation window with respect to the reference time.
[0058] Alternatively, in some embodiments, the second configuration for the at least one observation window may indicate the second starting time of each of the at least one observation window with respect to the reference time, and the second ending time of each of the at least one observation window with respect to the reference time.
[0059] In some embodiments, the duration of each of the at least one prediction window comprises a first number of time units.
[0060] In some embodiments, the duration of each of the at least one observation window comprises a second number of time units.
[0061] In some embodiments, the time units may comprise one of the following: milliseconds, slots, sequential aperiodic occasions, or periodic occasions with a fixed interval.
[0062] In some embodiments, each of the sequential aperiodic occasions or periodic occasions may be an occasion where a predicted CSI instance is to be applied. Hereinafter, the term “predicted CSI instance” may be used interchangeably with the term “predicted CSI” or “apiece of predicted CSI” . Similarly, the term “predicted CSI instances” may be used interchangeably with the term “predicted CSI” or “multiple pieces of predicted CSI” .
[0063] In some embodiments, the reference time may comprise first time when the terminal device 110 starts model inference. As described above, the LCM of the AI or ML model may comprise the model inference.
[0064] Alternatively, in some embodiments, the reference time may comprise second time when the terminal device 110 receives an indication from the network device 120. The indication indicates that at least one of the following is to be performed by the terminal device 110: reporting of predicted CSI instances, CSI measurements, or the model inference. This will be described with reference to Fig. 3A.
[0065] Fig. 3A illustrates an example of a prediction window and an observation window in accordance with some embodiments of the present disclosure.
[0066] In the example of Fig. 3A, a first configuration for a prediction window W1 indicates first starting time (represented by t1) of the prediction window W1 with respect to reference time (represented by t0) and first ending time (represented by t2) of the prediction window W1 with respect to the reference time t0. A second configuration for an observation window W2 indicates second starting time (represented by t3) of the observation window W2 with respect to reference time t0 and second ending time (represented by t4) of the observation window W2 with respect to the reference time t0. At the reference time t0, the terminal device 110 receives an indication from the network device 120. The indication indicates that at least one of the following is to be performed by the terminal device 110: reporting of predicted CSI instances, CSI measurements, or the model inference.
[0067] In addition, in the example of Fig. 3A, the duration of the prediction window comprise a first number of milliseconds or slots, and the duration of the observation window comprise a second number of milliseconds or slots.
[0068] Alternatively, in some embodiments, the reference time may comprise third time which is determined based on the second time and a time offset. This will be described with reference to Figs. 3B and 3C.
[0069] Fig. 3B illustrates another example of a prediction window and an observation window in accordance with some embodiments of the present disclosure. The first configuration for the prediction window W1 and the second configuration for the observation window W2 in the example of Fig. 3B are similar to those in the example of Fig. 3A.
[0070] The example of Fig. 3B is different from the example of Fig. 3A in that the terminal device 110 receives, at time t5, an indication from the network device 120. The indication indicates that at least one of the following is to be performed by the terminal device 110: reporting of predicted CSI instances, CSI measurements, or the model inference. The reference time t0 is determined based on t5 and a time offset. For example, the reference time t0 may be determined as a sum of t5 and the time offset. In other words, the reference time t0 may be behind t5 by the time offset. The time offset may be predefined or configured by the network device 120.
[0071] Fig. 3C illustrates a further example of a prediction window and an observation window in accordance with some embodiments of the present disclosure. The first configuration for the prediction window W1 and the second configuration for the observation window W2 in the example of Fig. 3C are similar to those in the example of Fig. 3A. The reference time t0 in the example of Fig. 3C is similar to the example of Fig. 3B.
[0072] The example of Fig. 3C is different from the examples of Figs. 3A and 3B in that the duration of the prediction window comprise a first number of sequential aperiodic occasions, and the duration of the observation window comprise a second number of periodic occasions with a fixed interval. For example, the duration of the prediction window comprise four sequential aperiodic occasions, and the duration of the observation window may comprise eighth sequential aperiodic occasions.
[0073] Alternatively, in some embodiments, the reference time may comprise fourth time when ending historic CSI within one of at least one observation window for the LCM was applied, an ending slot or an ending occasion within one of at least one observation window. This will be described with reference to Fig. 3D.
[0074] Fig. 3D illustrates a still further example of a prediction window and an observation window in accordance with some embodiments of the present disclosure. The first configuration for the prediction window W1 and the second configuration for the observation window W2 in the example of Fig. 3D are similar to those in the example of Fig. 3C.
[0075] The example of Fig. 3D is different from the examples of Fig. 3C in that the reference time t0 in the example of Fig. 3D is time 310 when ending historic CSI within the observation window W2 was applied. Alternatively, the reference time t0 in the example of Fig. 3D is an ending slot or an ending occasion 310 within the observation window W2. In other words, the reference time t0 in the example of Fig. 3D is the same as the ending time t4 of the observation window W2.
[0076] In some embodiments, the terminal device 110 may perform at least one of the following based on the first configuration for at least one prediction window: data collection, model training, model inference, model selection, model activation, or model switching.
[0077] In some embodiments, the terminal device 110 may perform at least one of the following based on the second configuration for at least one observation window: data collection, model training, model inference, model selection, model activation, or model switching.
[0078] In some embodiments, the terminal device 110 may receive the first configuration for at least one prediction window via a first radio resource control (RRC) signalling. Alternatively, the terminal device 110 may receive the first configuration for at least one prediction window via the first RRC signalling and DCI.
[0079] In some embodiments, the terminal device 110 may receive the second configuration for at least one observation window via a second radio resource control (RRC) signalling. Alternatively, the terminal device 110 may receive the second configuration for at least one observation window via the second RRC signalling and DCI.
[0080] In some embodiments, the terminal device 110 may receive the first configuration for at least one prediction window and the second configuration for at least one observation window via a single RRC signalling. Alternatively, the terminal device 110 may receive the first configuration for at least one prediction window and the second configuration for at least one observation window via the RRC signalling and DCI.
[0081] In some embodiments, the terminal device 110 may transmit, to the network device 120, first capability information associated with the at least one prediction window. For example, the first capability information may indicate the number of predicted CSI or channel measurements applied in AI / ML model training or AI / ML model inference, or the time distance of predicted CSI or channel measurements applied in AI / ML model training or AI / ML model inference.
[0082] In some embodiments, the terminal device 110 may transmit, to the network device 120, second capability information associated with the at least one observation window. For example, the second capability information may indicate the number of historic CSI or channel measurements applied in AI / ML model training or AI / ML model inference, or the time distance of historic CSI or channel measurements applied in AI / ML model training or AI / ML model inference.
[0083] In some embodiments, the terminal device 110 may transmit, to the network device 120, the first capability information which can be applied in multi-user (MU) scheduling.
[0084] In some embodiments, the at least one prediction window may comprise a single prediction window. In such embodiments, the at least one observation window may comprise a single observation window.
[0085] In some embodiments, the terminal device 110 may apply the single prediction window and / or the single observation window to fulfil CSI report resource assignment from the network device 120.
[0086] In some embodiments, the terminal device 110 may determine a second number of predicted CSI instances by performing model inference once.
[0087] In such embodiments, the second number is below a threshold number of predicted CSI instances indicated or supported by the first configuration for the single prediction window.
[0088] Alternatively or additionally, in such embodiments, a first time interval between a starting predicted CSI instance and an ending predicted CSI instance among the second number of predicted CSI instances is below a first threshold time interval indicated or supported by the first configuration for the single prediction window.
[0089] Alternatively or additionally, in such embodiments, a second time interval between two adjacent predicted CSI instances among the second number of predicted CSI instances is above a second threshold time interval indicated or supported by the first configuration for the single prediction window.
[0090] In some embodiments, if the second number is above the threshold number, the terminal device 110 may transmit, to the network device 120, a CSI report comprising the threshold number of predicted CSI instances. For example, the CSI report may comprise the earliest predicted CSI instances.
[0091] In some embodiments, if the second number is less than the threshold number, the terminal device 110 may transmit, to the network device 120, an indication indicating exceptional event. Alternatively, if the second number is less than the threshold number, the terminal device 110 may transmit predefined information to the network device 120. For example, the terminal device 110 may transmit the indication or the predefined information in an additional CSI report occasion.
[0092] In some embodiments, the at least one prediction window may comprise multiple prediction windows. In such embodiments, the at least one observation window may comprise multiple observation windows. In such embodiments, the terminal device 110 may apply a prediction window satisfying the first configuration and / or an observation window satisfying the second configuration to fulfil CSI report resource assignment from the network device 120. In such embodiments, the terminal device 110 may expect the network device 120 will allocate largest or enough CSI report resources.
[0093] In some embodiments, the multiple prediction windows may comprise a first prediction window for single user (SU) scheduling and a second prediction window for MU scheduling.
[0094] In some embodiments, the multiple observation windows may comprise a first observation window for SU scheduling and a second observation window for MU scheduling.
[0095] In some embodiments, if at least one of the first configuration for at least one prediction window or the second configuration for at least one observation window is used for data collection, the terminal device 110 may expect that the configured prediction window and observation window can satisfy the transmitted channel state information reference signal (CSI-RS) from the network device 120.
[0096] In some embodiments, the terminal device 110 may transmit, to the network device 120, time information associated with each of at least one predicted CSI instance.
[0097] In some embodiments, the time information may indicate time when each of at least one predicted CSI instance is to be applied.
[0098] In some embodiments, the time information may indicate one of the following: an index of a first slot to which each of the at least one predicted CSI instance is to be applied, or an offset between the first slot and a second slot in which the at least one predicted CSI instance is reported.
[0099] In some embodiments, the terminal device 110 may transmit the time information together with a CSI report comprising the at least one predicted CSI instance.
[0100] Alternatively, in some embodiments, the terminal device 110 may not transmit the time information explicitly. In such embodiments, the network device 120 may determine the time information based on a configured offset between a first slot and a second slot. In the first slot, a predicted CSI instance is to be applied. In the second slot, the predicted CSI instance is reported. This will be described with reference to Fig. 4A.
[0101] Fig. 4A illustrates an example of reporting of predicted CSI instances in accordance with some embodiments of the present disclosure. As shown in Fig. 4A, the terminal device 110 performs model inference of an AI / ML model to generate a set of outputs (i.e., predicted CSI instances) in a prediction window W1 based on a set of inputs in an observation window W2. The predicted CSI instances are to be applied in slots a, b and c.
[0102] The terminal device 110 transmits, in slot n, a CSI report comprising the predicted CSI instances. A first offset between slot n and slot a, a second offset between slot n and slot b and a third offset between slot n and slot c are configured by the network device 120. Upon receiving the CSI report, the network device 120 may determine, based on slot n, the first offset, the second offset and the third offset, slots a, b, c which are associated with the predicted CSI instances.
[0103] In some embodiments, the terminal device 110 may transmit a joint CSI report in one slot for multiple predicted CSI instances. Hereinafter, some embodiments of the joint CSI report will be described with reference to Figs. 4A and 4B.
[0104] In some embodiments, the terminal device 110 may transmit a periodic CSI report to the network device 120. In such embodiments, the terminal device 110 may receive, from the network device 120, a configuration for periodic slots for CSI reports (i.e., periodic CSI reports) . In turn, the terminal device 110 may transmit one of the CSI reports to the network device 120 in each of the periodic slots.
[0105] In some embodiments, each of the periodic CSI reports may comprise multiple predicted CSI instances. For example, as shown in Fig. 4A, the terminal device 110 may transmit a periodic CSI report in slot n. Slot n is one of periodic slots. The periodic CSI report comprises the predicted CSI instances which are to be applied in slots a, b and c, respectively.
[0106] Alternatively, in some embodiments, the terminal device 110 may transmit an aperiodic CSI report or a semi-persistent CSI report to the network device 120. In such embodiments, the terminal device 110 may receive, in a third slot from the network device 120, an indication indicating a CSI report is to be reported. In turn, the terminal device 110 may transmit, to the network device 120, a CSI report in a fourth slot subsequent to the third slot. A time interval between the third slot and the fourth slot is predefined or configured by the network device 120. This will be described with reference to Fig. 4B.
[0107] Fig. 4B illustrates an example of reporting of predicted CSI instances in accordance with some embodiments of the present disclosure. As shown in Fig. 4B, the terminal device 110 performs model inference of an AI / ML model to generate a set of outputs (i.e., predicted CSI instances) in a prediction window W1 based on a set of inputs in an observation window W2. The predicted CSI instances are to be applied in slots a, b and c.
[0108] The terminal device 110 receives, in slot m, from the network device 120, an indication indicating a CSI report is to be reported. The indication may trigger model inference at the terminal device 110.
[0109] In turn, the terminal device 110 may transmit, to the network device 120, an aperiodic CSI report or a semi-persistent CSI report CSI report in slot n subsequent to slot m. A time interval between slot m and slot n is predefined or configured by the network device 120. The aperiodic CSI report or the semi-persistent CSI report comprises predicted CSI instances which are to be applied in slots a, b and c.
[0110] In some embodiments, the terminal device 110 may start the model inference at least processing time before slot n. For example, as shown in Fig. 4B, the processing time is represented by t_proc. The terminal device 110 may start the model inference at least t_proc ms before slot n.
[0111] In some embodiments, if the time interval between slot m and slot n is less than the processing time (t_proc) , the terminal device 110 may use legacy CSI report.
[0112] In some embodiments, the time information as described above may not be determined based on the slot configured as full UL symbols. In other words, i.e., only the slots with DL / Flexible symbols are considered.
[0113] In some embodiments, the terminal device 110 may transmit separate reports in a set of slots for multiple predicted CSI instances. Hereinafter, some embodiments of the separate reports will be described with reference to Figs. 4C and 4D.
[0114] In some embodiments, the terminal device 110 may transmit periodic CSI reports to the network device 120 in the set of slots. In such embodiments, the set of slots comprises periodic slots. The terminal device 110 may receive, from the network device 120, a configuration for the periodic slots for CSI reports (i.e., periodic CSI reports) . In turn, the terminal device 110 may transmit one of the CSI reports to the network device 120 in each of the periodic slots. Each of the periodic CSI reports may comprise a single predicted CSI instances. This will be described with reference to Fig. 4C.
[0115] Fig. 4C illustrates an example of reporting of predicted CSI instances in accordance with some embodiments of the present disclosure. As shown in Fig. 4C, the terminal device 110 performs model inference of an AI / ML model to generate a set of outputs (i.e., predicted CSI instances) in a prediction window W1 based on a set of inputs in an observation window W2. The predicted CSI instances are to be applied in slots a, b and c.
[0116] The terminal device 110 transmits periodic CSI reports to the network device 120 in a set of slots. The set of slots comprises periodic slots n, n+p and n+q. The terminal device 110 receives, from the network device 120, a configuration for the periodic slots for periodic CSI reports. In turn, the terminal device 110 a periodic CSI report in each of slots n, n+p and n+q. The periodic CSI report in slot n comprises a predicted CSI instance which is to be applied in slot a. The periodic CSI report in slot n+p comprises a predicted CSI instance which is to be applied in slot b. The periodic CSI report in slot n+q comprises a predicted CSI instance which is to be applied in slot c.
[0117] Alternatively, in some embodiments, the terminal device 110 may transmit an aperiodic CSI report or a semi-persistent CSI report to the network device 120. In such embodiments, the terminal device 110 may receive, in a third slot from the network device 120, an indication indicating a CSI report is to be reported. In turn, the terminal device 110 may transmit, to the network device 120, a CSI report in a fourth slot subsequent to the third slot. A time interval between the third slot and the fourth slot is predefined or configured by the network device 120. This will be described with reference to Fig. 4D.
[0118] Fig. 4D illustrates an example of reporting of predicted CSI instances in accordance with some embodiments of the present disclosure. As shown in Fig. 4D, the terminal device 110 performs model inference of an AI / ML model to generate a set of outputs (i.e., predicted CSI instances) in a prediction window W1 based on a set of inputs in an observation window W2. The predicted CSI instances are to be applied in slots a, b and c.
[0119] In slot m, the terminal device 110 receives, from the network device 120, an indication indicating a CSI report is to be reported. The indication may trigger model inference at the terminal device 110. In addition, in slot m, the terminal device 110 receives, from the network device 120, a configuration for a first time interval (i.e., n-m) between slot m and a first (i.e., initial) report slot n.
[0120] In slot n, the terminal device 110 receives, from the network device 120, an indication indicating a CSI report is to be reported. In addition, in slot n, the terminal device 110 receives, from the network device 120, a configuration for a second time interval (i.e., p) between slot n and a second subsequent report slot n+p.
[0121] In slot n+p, the terminal device 110 receives, from the network device 120, an indication indicating a CSI report is to be reported. In addition, in slot n+p, the terminal device 110 receives, from the network device 120, a configuration for a third time interval (i.e., q-p) between slot n+p and a third subsequent report slot n+q.
[0122] Alternatively, in slot m, the terminal device 110 receives, from the network device 120, an indication indicating a CSI report is to be reported. The indication may trigger model inference at the terminal device 110. In addition, in slot m, the terminal device 110 receives, from the network device 120, a configuration for a single time interval between every two subsequent reports slots and a total number of the report slots. For example, a total number of the report slots is equal to three. In such embodiments, n-m is equal to any of p and q-p.
[0123] Alternatively, in slot m, the terminal device 110 receives, from the network device 120, a configuration for multiple time intervals between every two subsequent reports slots without the total number of the report slots. In such embodiments, n-m is not equal to any of p and q-p.
[0124] In turn, the terminal device 110 may transmit, to the network device 120, an aperiodic CSI report or a semi-persistent CSI report CSI report in slot n subsequent to slot m. The aperiodic CSI report or the semi-persistent CSI report in slot n comprises a predicted CSI instance which is to be applied in slot a.
[0125] The terminal device 110 may transmit, to the network device 120, an aperiodic CSI report or a semi-persistent CSI report CSI report in slot n+p. The aperiodic CSI report or the semi-persistent CSI report in slot n+p comprises a predicted CSI instance which is to be applied in slot b.
[0126] The terminal device 110 may transmit, to the network device 120, an aperiodic CSI report or a semi-persistent CSI report CSI report in slot n+q. The aperiodic CSI report or the semi-persistent CSI report in slot n+q comprises a predicted CSI instance which is to be applied in slot c.
[0127] In some embodiments, the terminal device 110 may start the model inference at least processing time before slot n. For example, as shown in Fig. 4D, the processing time is represented by t_proc. The terminal device 110 may start the model inference at least t_proc ms before slot n.
[0128] In some embodiments, the terminal device 110 may determine at least one performance metric for the AI or ML model by performing model monitoring (also referred to as model performance monitoring) .
[0129] In some embodiments, the terminal device 110 may determine the at least one performance metric for the AI or ML model based on one of the following: a staring predicted CSI instance within a first prediction window among the at least one prediction window, or an ending predicted CSI instance within the first prediction window.
[0130] Alternatively, in some embodiments, the terminal device 110 may determine the at least one performance metric for the AI or ML model based on all predicted CSI instances within the first prediction window. For example, the terminal device 110 may determine a mean key performance indicator (KPI) across all predicted CSI instances.
[0131] In some embodiments, the terminal device 110 may receive, from the network device 120, a counter value indicating times for determining the at least one performance metric. In turn, the terminal device 110 may determine the at least one performance metric based on the counter value.
[0132] In some embodiments, the terminal device 110 may determine the AI or ML model is not applicable if the at least one performance metric is below a threshold for the times indicated by the counter value.
[0133] In some embodiments, the at least one performance metric comprises an intermediate key performance indicator (KPI) . For example, the intermediate KPI may comprise normalized mean square error (NMSE) or squared generalized cosine similarity (SGCS) .
[0134] In some embodiments, if the intermediate KPI is below a threshold, the terminal device 110 may switch from the AI or ML model to a further AI or ML model. In other words, if the AI or ML model is not applicable, the terminal device 110 may switch from the AI or ML model to a further AI or ML model.
[0135] In some embodiments, the at least one performance metric comprises an intermediate KPI and an eventual KPI.
[0136] In some embodiments, the eventual KPI may comprise user perceived throughput (UPT) , hypothetical block error ratio (BLER) , or hybrid automatic repeat request acknowledgement (HARQ-ACK) .
[0137] In some embodiments, if the intermediate KPI is above a threshold and the eventual KPI is below the threshold, the terminal device 110 may switch from the AI or ML model to a further AI or ML model.
[0138] In some embodiments, a first duration duration of a first observation window associated with the AI or ML model is shorter than a second duration duration of a second observation window associated with the further AI or ML model. Alternatively, a third duration duration of a first prediction window associated with the AI or ML model is longer than a fourth duration duration of a second prediction window associated with the further AI or ML model.
[0139] In some embodiments, if the further AI or ML model is unavailable, the terminal device 110 may transmit results of model performance monitoring to the network device 120. In such embodiments, the terminal device 110 may perform fallback to legacy CSI report based on a further indication received from the network device 120.
[0140] In some embodiments, the terminal device 110 may transmit, to the network device 120, information about switching from the AI or ML model to the further AI or ML model.
[0141] For example, the information about switching may comprise information about at least one of the prediction window and the observation window. Alternatively or additionally, the information about switching may comprise a request for a configuration for an updated prediction window. Alternatively or additionally, the information about switching may comprise a request for a configuration for an updated observation window.
[0142] In some embodiments, the terminal device 110 may receive, from the network device 120, a fourth configuration for data collection for model training or model performance monitoring. The fourth configuration indicates that only CSI measurement is performed at the terminal device 110. For example, the terminal device 110 may receive the third configuration via an RRC signalling.
[0143] In some embodiments, the fourth configuration may comprise one of the following:
[0144] ● an invalid CSI reporting resource configuration,
[0145] ● an CSI reporting quantity configuration indicating none, or
[0146] ● a predefined CSI reporting resource configuration,
[0147] Alternatively, in some embodiments, the fourth configuration may comprise at least one dedicated parameter indicating that only CSI measurement is performed at the terminal device 110.
[0148] Alternatively, in some embodiments, the fourth configuration may comprise dedicated downlink control information (DCI) indicating that only CSI measurement is performed at the terminal device 110.
[0149] Fig. 5 illustrates a flowchart of an example method in accordance with some embodiments of the present disclosure. In some embodiments, the method 500 can be implemented at a terminal device, such as the terminal device 110 as shown in Fig. 1. For the purpose of discussion, the method 500 will be described with reference to Fig. 1.
[0150] At block 510, the terminal device 110 receives, from the network device 120, a first configuration for at least one prediction window for LCM of an AI or ML model. The first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to reference time, or first ending time of each of the at least one prediction window with respect to the reference time.
[0151] At block 520, the terminal device 110 applies one of the at least one prediction window.
[0152] In some embodiments, the method 500 further comprises: receiving, from the network device, a second configuration for at least one observation window for the LCM of the AI or ML model; and applying one of the at least one observation window. The second configuration indicates at least one of the following: a duration of each of the at least one observation window, second starting time of each of the at least one observation window with respect to the reference time, or second ending time of each of the at least one observation window with respect to the reference time.
[0153] In some embodiments, the reference time comprises one of the following: first time when the terminal device starts model inference, second time when the terminal device receives an indication from the the network device, the indication indicating that at least one of the following is to be performed by the terminal device: reporting of predicted channel state information (CSI) instances, CSI measurements, or the model inference, third time which is determined based on the second time and a time offset, fourth time when ending historic CSI within one of at least one observation window for the LCM was applied, or an ending slot within one of at least one observation window, or an ending occasion within one of at least one observation window.
[0154] In some embodiments, the duration of each of the at least one prediction window comprises a first number of time units.
[0155] In some embodiments, the time units comprise one of the following: milliseconds, slots, sequential aperiodic occasions, or periodic occasions with a fixed interval.
[0156] In some embodiments, the method 500 further comprises: transmitting, to the network device, first capability information associated with the at least one prediction window.
[0157] In some embodiments, the at least one prediction window comprises a single prediction window. In such embodiments, the method 500 further comprises determining a second number of predicted channel state information (CSI) instances by performing model inference once. In such embodiments, the second number is below a threshold number of predicted CSI instances indicated or supported by the first configuration for the single prediction window; and / or a first time interval between a starting predicted CSI instance and an ending predicted CSI instance among the second number of predicted CSI instances is below a first threshold time interval indicated or supported by the first configuration for the single prediction window; and / or a second time interval between two adjacent predicted CSI instances among the second number of predicted CSI instances is above a second threshold time interval indicated or supported by the first configuration for the single prediction window.
[0158] In some embodiments, the method 500 further comprises: based on determining that the second number is above the threshold number, transmitting, to the network device, a CSI report comprising the threshold number of predicted CSI instances.
[0159] In some embodiments, the method 500 further comprises: based on determining that the second number is less than the threshold number, transmitting, to the network device, an indication indicating exceptional event.
[0160] In some embodiments, the method 500 further comprises: transmitting, to the network device, time information associated with each of at least one predicted channel state information (CSI) instance.
[0161] In some embodiments, the time information indicates one of the following: an index of a first slot to which each of the at least one predicted CSI instance is to be applied, or an offset between the first slot and a second slot in which the at least one predicted CSI instance is reported.
[0162] In some embodiments, the method 500 further comprises: receiving, from the network device, a third configuration for periodic slots for CSI reports; and transmitting, to the network device, one of the CSI reports in each of the periodic slots.
[0163] In some embodiments, each of the CSI reports comprises at least one predicted CSI instance.
[0164] In some embodiments, the method 500 further comprises: receiving, in a third slot from the network device, an indication indicating a CSI report is to be reported; and transmitting, to the network device, the CSI report in a fourth slot subsequent to the third slot. In such embodiments, a time interval between the third slot and the fourth slot is predefined or configured by the network device.
[0165] In some embodiments, the method 500 further comprises: starting model inference at least processing time before the fourth slot.
[0166] In some embodiments, the method 500 further comprises: determining at least one performance metric for the AI or ML model based on one of the following: a staring predicted CSI instance within a first prediction window among the at least one prediction window, an ending predicted CSI instance within the first prediction window, or all predicted CSI instances within the first prediction window.
[0167] In some embodiments, the at least one performance metric comprises an intermediate key performance indicator (KPI) . In such embodiments, the method 500 further comprises switching from the AI or ML model to a further AI or ML model based on determining that the intermediate KPI is below a threshold.
[0168] In some embodiments, the at least one performance metric comprises an intermediate key performance indicator (KPI) and an eventual KPI. In such embodiments, the method 500 further comprises switching from the AI or ML model to a further AI or ML model based on determining the following: the intermediate KPI is above a threshold; and the eventual KPI is below the threshold.
[0169] In some embodiments, a first duration duration of a first observation window associated with the AI or ML model is shorter than a second duration duration of a second observation window associated with the further AI or ML model; or a third duration duration of a first prediction window associated with the AI or ML model is longer than a fourth duration duration of a second prediction window associated with the further AI or ML model.
[0170] In some embodiments, the method 500 further comprises: based on determining that the further AI or ML model is unavailable, transmitting results of model performance monitoring to the network device.
[0171] In some embodiments, the method 500 further comprises: transmitting, to the network device, information about switching from the AI or ML model to the further AI or ML model.
[0172] In some embodiments, the method 500 further comprises: receiving, from the network device, a counter value indicating times for determining the at least one performance metric. In such embodiments, determining the at least one performance metric comprises determining the at least one performance metric based on the counter value.
[0173] In some embodiments, the method 500 further comprises: determining the AI or ML model is not applicable based on determining that the at least one performance metric is below a threshold for the times indicated by the counter value.
[0174] In some embodiments, the method 500 further comprises: receiving, from the network device, a fourth configuration for data collection for model training or model performance monitoring. In such embodiments, the third configuration indicates that only channel state information (CSI) measurement is performed at the terminal device.
[0175] In some embodiments, the third configuration comprises one of the following: an invalid CSI reporting resource configuration, an CSI reporting quantity configuration indicating none, a predefined CSI reporting resource configuration, at least one dedicated parameter indicating that only CSI measurement is performed at the terminal device, or dedicated downlink control information indicating that only CSI measurement is performed at the terminal device.
[0176] Fig. 6 illustrates a flowchart of an example method in accordance with some embodiments of the present disclosure. In some embodiments, the method 600 can be implemented at a network device, such as the network device 120 as shown in Fig. 1. For the purpose of discussion, the method 600 will be described with reference to Fig. 1.
[0177] At block 610, the network device 120 determines a first configuration for at least one prediction window for LCM of an AI or ML model. The first configuration indicates at least one of the following: a duration of each of the at least one prediction window, first starting time of each of the at least one prediction window with respect to reference time, or first ending time of each of the at least one prediction window with respect to the reference time; and transmit the first configuration to a terminal device.
[0178] In some embodiments, the method 600 further comprises: transmitting, to the terminal device, a second configuration for at least one observation window for the LCM of the AI or ML model. In such embodiments, the second configuration indicates at least one of the following: a duration of each of the at least one observation window, second starting time of each of the at least one observation window with respect to the reference time, or second ending time of each of the at least one observation window with respect to the reference time.
[0179] In some embodiments, the reference time comprises one of the following: first time when the terminal device starts model inference, second time when the terminal device receives an indication from the the network device, the indication indicating that at least one of the following is to be performed by the terminal device: reporting of predicted channel state information (CSI) instances, CSI measurements, or the model inference, third time which is determined based on the second time and a time offset, fourth time when ending historic CSI within one of at least one observation window for the LCM was applied, or an ending slot within one of at least one observation window, or an ending occasion within one of at least one observation window.
[0180] In some embodiments, the duration of each of the at least one prediction window comprises a first number of time units.
[0181] In some embodiments, the time units comprise one of the following: milliseconds, slots, sequential aperiodic occasions, or periodic occasions with a fixed interval.
[0182] In some embodiments, the method 600 further comprises: receiving, from the terminal device, first capability information associated with the at least one prediction window.
[0183] In some embodiments, the method 600 further comprises: receiving, from the terminal device, time information associated with each of at least one predicted channel state information (CSI) instance.
[0184] In some embodiments, the time information indicates one of the following: an index of a first slot to which each of the at least one predicted CSI instance is to be applied, or an offset between the first slot and a second slot in which the at least one predicted CSI instance is reported.
[0185] In some embodiments, the method 600 further comprises: transmitting, to the terminal device, a third configuration for periodic slots for CSI reports; and receiving, from the terminal device, one of the CSI reports in each of the periodic slots.
[0186] In some embodiments, each of the CSI reports comprises at least one predicted CSI instance.
[0187] In some embodiments, the method 600 further comprises: transmitting, in a third slot to the terminal device, an indication indicating a CSI report is to be reported; and receiving, from the terminal device, the CSI report in a fourth slot subsequent to the third slot. In such embodiments, a time interval between the third slot and the fourth slot is predefined or configured by the network device.
[0188] In some embodiments, the method 600 further comprises: receiving results of model performance monitoring from the terminal device.
[0189] In some embodiments, the method 600 further comprises: receiving, from the terminal device, information about switching from the AI or ML model to a further AI or ML model.
[0190] In some embodiments, the method 600 further comprises: transmitting, to the terminal device, a counter value indicating times for determining at least one performance metric for the AI or ML model.
[0191] In some embodiments, the method 600 further comprises: transmitting, to the terminal device, a fourth configuration for data collection for model training or model performance monitoring. The third configuration indicates that only channel state information (CSI) measurement is performed at the terminal device.
[0192] In some embodiments, the fourth configuration comprises one of the following: an invalid CSI reporting resource configuration, an CSI reporting quantity configuration indicating none, a predefined CSI reporting resource configuration, at least one dedicated parameter indicating that only CSI measurement is performed at the terminal device, or dedicated downlink control information indicating that only CSI measurement is performed at the terminal device.
[0193] Fig. 7 is a simplified block diagram of a device 700 that is suitable for implementing embodiments of the present disclosure. The device 700 can be considered as a further example embodiment of the terminal device 110 or the network device 120 as shown in Fig. 1. Accordingly, the device 700 can be implemented at or as at least a part of the terminal device 110 or or the network device 120.
[0194] As shown, the device 700 includes a processor 710, a memory 720 coupled to the processor 710, a suitable transceiver 740 coupled to the processor 710, and a communication interface coupled to the transceiver 740. The memory 710 stores at least a part of a program 730. The transceiver 740 may be for bidirectional communications or a unidirectional communication based on requirements. The transceiver 740 may include at least one of a transmitter 742 and a receiver 744. The transmitter 742 and the receiver 744 may be functional modules or physical entities. The transceiver 740 has at least one antenna to facilitate communication, though in practice an Access Node mentioned in this application may have several ones. The communication interface may represent any interface that is necessary for communication with other network elements, such as X2 / Xn interface for bidirectional communications between eNBs / gNBs, S1 / NG interface for communication between a Mobility Management Entity (MME) / Access and Mobility Management Function (AMF) / SGW / UPF and the eNB / gNB, Un interface for communication between the eNB / gNB and a relay node (RN) , or Uu interface for communication between the eNB / gNB and a terminal device.
[0195] The components included in the apparatuses and / or devices of the present disclosure may be implemented in various manners, including software, hardware, firmware, or any combination thereof. In one embodiment, one or more units may be implemented using software and / or firmware, for example, machine-executable instructions stored on the storage medium. In addition to or instead of machine-executable instructions, parts or all of the units in the apparatuses and / or devices may be implemented, at least in part, by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs) , Application-specific Integrated Circuits (ASICs) , Application-specific Standard Products (ASSPs) , System-on-a-chip systems (SOCs) , Complex Programmable Logic Devices (CPLDs) , and the like.
Claims
1.A terminal device, comprising:a processor configured to cause the terminal device to:receive, from a network device, a first configuration for at least one prediction window for life cycle management (LCM) of an artificial intelligence (AI) or machine learning (ML) model, wherein the first configuration indicates at least one of the following:a duration of each of the at least one prediction window,first starting time of each of the at least one prediction window with respect to reference time, orfirst ending time of each of the at least one prediction window with respect to the reference time; andapply one of the at least one prediction window.2.The terminal device of claim 1, wherein the terminal device is further caused to:receive, from the network device, a second configuration for at least one observation window for the LCM of the AI or ML model, wherein the second configuration indicates at least one of the following:a duration of each of the at least one observation window,second starting time of each of the at least one observation window with respect to the reference time, orsecond ending time of each of the at least one observation window with respect to the reference time; andapply one of the at least one observation window.3.The terminal device of claim 1 or 2, wherein the reference time comprises one of the following:first time when the terminal device starts model inference,second time when the terminal device receives an indication from the the network device, the indication indicating that at least one of the following is to be performed by the terminal device: reporting of predicted channel state information (CSI) instances, CSI measurements, or the model inference,third time which is determined based on the second time and a time offset,fourth time when ending historic CSI within one of at least one observation window for the LCM was applied, oran ending slot within one of at least one observation window, oran ending occasion within one of at least one observation window.4.The terminal device of claim 1, wherein the terminal device is further caused to:transmit, to the network device, first capability information associated with the at least one prediction window.5.The terminal device of claim 1, wherein the at least one prediction window comprises a single prediction window; andwherein the terminal device is further caused to determine a second number of predicted channel state information (CSI) instances by performing model inference once, wherein:the second number is below a threshold number of predicted CSI instances indicated or supported by the first configuration for the single prediction window; and / ora first time interval between a starting predicted CSI instance and an ending predicted CSI instance among the second number of predicted CSI instances is below a first threshold time interval indicated or supported by the first configuration for the single prediction window; and / ora second time interval between two adjacent predicted CSI instances among the second number of predicted CSI instances is above a second threshold time interval indicated or supported by the first configuration for the single prediction window.6.The terminal device of claim 5, wherein the terminal device is further caused to:based on determining that the second number is above the threshold number, transmit, to the network device, a CSI report comprising the threshold number of predicted CSI instances.7.The terminal device of claim 5, wherein the terminal device is further caused to:based on determining that the second number is less than the threshold number, transmit, to the network device, an indication indicating exceptional event.8.The terminal device of claim 1, wherein the terminal device is further caused to:transmit, to the network device, time information associated with each of at least one predicted channel state information (CSI) instance.9.The terminal device of claim 1, wherein the terminal device is further caused to:receive, from the network device, a third configuration for periodic slots for CSI reports; andtransmit, to the network device, one of the CSI reports in each of the periodic slots.10.The terminal device of claim 1, wherein the terminal device is further caused to:receive, in a third slot from the network device, an indication indicating a CSI report is to be reported; andtransmit, to the network device, the CSI report in a fourth slot subsequent to the third slot, wherein a time interval between the third slot and the fourth slot is predefined or configured by the network device.11.The terminal device of claim 10, wherein the terminal device is further caused to:start model inference at least processing time before the fourth slot.12.The terminal device of claim 1, wherein the terminal device is further caused to:determine at least one performance metric for the AI or ML model based on one of the following:a staring predicted CSI instance within a first prediction window among the at least one prediction window,an ending predicted CSI instance within the first prediction window, orall predicted CSI instances within the first prediction window.13.The terminal device of claim 12, wherein the at least one performance metric comprises an intermediate key performance indicator (KPI) ; andwherein the terminal device is further caused to switch from the AI or ML model to a further AI or ML model based on determining that the intermediate KPI is below a threshold.14.The terminal device of claim 12, wherein the at least one performance metric comprises an intermediate key performance indicator (KPI) and an eventual KPI;wherein the terminal device is further caused to switch from the AI or ML model to a further AI or ML model based on determining the following:the intermediate KPI is above a threshold; andthe eventual KPI is below the threshold.15.The terminal device of claim 13 or 14, wherein:a first duration duration of a first observation window associated with the AI or ML model is shorter than a second duration duration of a second observation window associated with the further AI or ML model; ora third duration duration of a first prediction window associated with the AI or ML model is longer than a fourth duration duration of a second prediction window associated with the further AI or ML model.16.The terminal device of claim 12, wherein:the terminal device is further caused to:receive, from the network device, a counter value indicating times for determining the at least one performance metric; andthe terminal device is caused to determine the at least one performance metric based on the counter value.17.The terminal device of claim 16, wherein the terminal device is further caused to:determine the AI or ML model is not applicable based on determining that the at least one performance metric is below a threshold for the times indicated by the counter value.18.The terminal device of claim 1, wherein the terminal device is further caused to:receive, from the network device, a fourth configuration for data collection for model training or model performance monitoring, wherein the third configuration indicates that only channel state information (CSI) measurement is performed at the terminal device.19.A network device, comprising:a processor configured to cause the network device to:determine a first configuration for at least one prediction window for life cycle management (LCM) of an artificial intelligence (AI) or machine learning (ML) model, wherein the first configuration indicates at least one of the following:a duration of each of the at least one prediction window,first starting time of each of the at least one prediction window with respect to reference time, orfirst ending time of each of the at least one prediction window with respect to the reference time; andtransmit the first configuration to a terminal device.20.The network device of claim 19, wherein the network device is further caused to:transmit, to the terminal device, a second configuration for at least one observation window for the LCM of the AI or ML model, wherein the second configuration indicates at least one of the following:a duration of each of the at least one observation window,second starting time of each of the at least one observation window with respect to the reference time, orsecond ending time of each of the at least one observation window with respect to the reference time.
Citation Information
Patent Citations
Model construction method and device and communication equipment
CN116415476A
Performance monitoring method and device of artificial intelligence AI model, communication equipment, communication system and storage medium
CN117136529A
Air interface test method and system based on AI / ML time domain CSI prediction
CN117221938A
Communication processing method, terminal, equipment, communication system and storage medium
CN117223375A
Method and apparatus for CSI report configuration for CSI predictions in one or more domains
US20230337036A1