The process of validating beam prediction models

Abstract

To verify the accuracy of the beam prediction model in the second device, the first device receives information from the second device indicating the N strongest predicted beams output by the beam prediction model. The first device also obtains information about the M strongest beams. Then, the first device determines whether at least K of the M strongest beams are among the N strongest predicted beams; and when at least K strongest beams are among the N strongest predicted beams, it verifies the accuracy of the beam prediction model. N, M, and K are positive integers, where M is less than or equal to N, and K is less than or equal to M.

Description

Technical Field

[0001] Various example embodiments relate to communication systems. Background Technology

[0002] Wireless communication systems are constantly evolving. New applications, use cases, and industry verticals are envisioned. Beamforming and beam management are expected to become more complex due to factors such as user mobility, a greater number of antennas, and the use of higher frequencies. Artificial intelligence, specifically machine learning, can provide efficient tools for beam management, such as beam prediction models inferred or trained in user equipment or corresponding devices. Solutions are expected to determine whether such beam prediction models are accurate enough. Summary of the Invention

[0003] The independent claims define the scope, and different embodiments are defined in the dependent claims.

[0004] According to one aspect, a first apparatus is provided, comprising at least one processor and at least one memory storing instructions, the instructions, when executed by the at least one processor, causing the first apparatus to at least: receive from a second apparatus information indicating N strongest predicted beams output by a beam prediction model; obtain information at the second apparatus regarding M strongest beams; determine whether at least K of the M strongest beams are among the N strongest predicted beams; and verify the accuracy of the beam prediction model when at least K strongest beams are among the N strongest predicted beams; wherein N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M.

[0005] In embodiments that can be combined with this aspect and other embodiments, at least one processor and at least one memory storing instructions, when executed by at least one processor, further enable the first device to obtain information about the M strongest beams at the second device by at least: receiving additional information indicating the M strongest measurement beams from the second device in the information, or by deriving information about the M strongest beams from the test setup configuration.

[0006] In embodiments that can be combined with this aspect and other embodiments, when executed by at least one processor and at least one memory storing instructions, the first device further causes the first device to at least: receive, prior to information, an indication from the second device as to whether the second device is performing inference or training of a beam prediction model; and, when the second device is performing inference or training of a beam prediction model, transmit to the second device configuration information for obtaining measurement results about a first beam set, and configuration information for obtaining predictions using measurement results about a second beam set as input.

[0007] In embodiments that can be combined with this aspect and other embodiments, the second beam set is a subset of the first beam set.

[0008] In embodiments that can be combined with this aspect and other embodiments, the second beam set includes beams different from the first beam set.

[0009] In embodiments that can be combined with this aspect and other embodiments, the first beam set includes a first number of beams, and the second beam set includes a second number of beams, the second number being at least one-quarter of the first number.

[0010] In embodiments that can be combined with this aspect and other embodiments, at least one processor and at least one memory storing instructions, when executed by at least one processor, also cause the first device to at least: transmit a command to the second device to switch to a function based on artificial intelligence / machine learning mode before receiving information, and / or a request to report at least N strongest predicted beams.

[0011] In embodiments that can be combined with this aspect and other embodiments, the information indicating the beam is a beam identifier.

[0012] In embodiments that can be combined with this aspect and other embodiments, at least one processor and at least one memory storing instructions, when executed by at least one processor, also cause the first device to at least: transmit information to the second device indicating whether the accuracy of the beam prediction model has been verified.

[0013] According to one aspect, a second apparatus is provided, comprising at least one processor and at least one memory storing instructions, which, when executed by the at least one processor, cause the second apparatus to at least: perform inference or training on a beam prediction model that outputs X strongest predicted beams; obtain measurement results about the beams; input the measurement results into the beam prediction model; and transmit to a first apparatus information indicating N strongest predicted beams output by the beam prediction model, wherein X and N are positive integers, and N is less than or equal to X.

[0014] In embodiments that can be combined with this aspect and other embodiments, at least one processor and at least one memory storing instructions, when executed by at least one processor, further cause the second device to at least: receive configuration information from the first device for obtaining measurement results with respect to a first beam set, and configuration information for obtaining predictions using measurement results with respect to a second beam set as input; input measurement results with respect to the second beam set into a beam prediction model; and transmit additional information to the first device indicating the M strongest measurement beams at the second device, where M is a positive integer less than or equal to N.

[0015] In embodiments that can be combined with this aspect and other embodiments, at least one processor and at least one memory storing instructions, when executed by at least one processor, also cause the second device to at least: transmit to the first device an indication of whether the second device is performing inference or training of a beam prediction model.

[0016] In embodiments that can be combined with this aspect and other embodiments, when executed by at least one processor and at least one memory storing instructions, the second device also causes the second device to at least: receive from the first device a command to switch to a function based on artificial intelligence / machine learning mode; and, after receiving the command and while in the function based on artificial intelligence / machine learning mode, transmit information to the first device confirming that the second device is in the function based on artificial intelligence / machine learning mode.

[0017] According to one aspect, a method is provided, the method comprising at least: receiving from a second device information indicating N strongest predicted beams output by a beam prediction model; obtaining information about M strongest beams at the second device; determining whether at least K of the M strongest beams are among the N strongest predicted beams; and verifying the accuracy of the beam prediction model when at least K strongest beams are among the N strongest predicted beams; wherein N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M.

[0018] According to one aspect, a method is provided, the method comprising at least: performing inference or training on a beam prediction model that outputs X strongest predicted beams; obtaining measurement results about the beams; inputting the measurement results into the beam prediction model; and transmitting information indicating N strongest predicted beams output by the beam prediction model to a first device, wherein X and N are positive integers, and N is less than or equal to X.

[0019] According to one aspect, a computer-readable medium including instructions, when executed by a device, causes the device to perform at least one of a first process or a second process, wherein the first process includes at least: receiving information from a second device indicating N strongest predicted beams output by a beam prediction model; obtaining information at the second device regarding M strongest beams; determining whether at least K of the M strongest beams are among the N strongest predicted beams; verifying the accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams; wherein N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M, and wherein the second process includes at least: performing inference or training on a beam prediction model that outputs X strongest predicted beams; obtaining measurement results regarding the beams; inputting the measurement results into the beam prediction model; and transmitting information indicating N strongest predicted beams output by the beam prediction model to a first device, wherein X and N are positive integers, and N is less than or equal to X.

[0020] According to one aspect, a non-transitory computer-readable medium including instructions, which, when executed by a device, cause the device to perform at least one of a first process or a second process, wherein the first process includes at least: receiving from a second device information indicating N strongest predicted beams output by a beam prediction model; obtaining information at the second device regarding M strongest beams; determining whether at least K of the M strongest beams are among the N strongest predicted beams; and verifying the accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams; wherein N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M, and wherein the second process includes at least: performing inference or training on a beam prediction model that outputs X strongest predicted beams; obtaining measurement results regarding the beams; inputting the measurement results into the beam prediction model; and transmitting to a first device information indicating N strongest predicted beams output by the beam prediction model, wherein X and N are positive integers, and N is less than or equal to X.

[0021] According to one aspect, a computer program including instructions, when executed by a device, causes the device to perform at least one of a first process or a second process, wherein the first process includes at least: receiving information from a second device indicating N strongest predicted beams output by a beam prediction model; obtaining information at the second device regarding M strongest beams; determining whether at least K of the M strongest beams are among the N strongest predicted beams; verifying the accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams, wherein N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M, and wherein the second process includes at least: performing inference or training on a beam prediction model that outputs X strongest predicted beams; obtaining measurement results regarding the beams; inputting the measurement results into the beam prediction model; and transmitting information indicating N strongest predicted beams output by the beam prediction model to a first device, wherein X and N are positive integers, and N is less than or equal to X. Attached Figure Description

[0022] The embodiments are described below by way of example only, with reference to the accompanying drawings, in which: Figure 1 An exemplary high-level system architecture is shown; Figure 2 Example functions of the device are shown; Figure 3 Example functions of the device are shown; Figure 4 An example of information exchange is shown; Figure 5 It is a schematic block diagram; Figure 6 It is a schematic block diagram; and Figure 7 It is a schematic block diagram. Detailed Implementation

[0023] The following embodiments are presented by way of example only. Although the specification may refer to "a," "an," or "some" embodiments and / or examples in several places, this does not necessarily mean that each such reference refers to the same embodiment or example, or that a particular feature applies only to a single embodiment and / or a single example. Individual features of different embodiments and / or examples may also be combined to provide other embodiments and / or examples. Furthermore, the words "comprising" and "including" should be understood not to limit the described embodiments to consisting only of those features already mentioned, and such embodiments may also include features / structures not specifically mentioned. Additionally, although ordinal terms (such as "first," "second," etc.) may be used to describe various elements, these elements are not limited by these terms. These terms are used only for the purpose of distinguishing an element from other elements. For example, without departing from the scope of this disclosure, a first device may be referred to as a device or a second device, and correspondingly, a second device may be referred to as a device or a first device.

[0024] Beyond 5G, advanced and future wireless networks are designed to support a wide range of services, use cases, and industrial verticals, such as driverless mobility with fully autonomous connected vehicles, vehicle-to-everything (V2X) services, or smart environments like smart industries, smart grids, or smart cities, to name just a few. To provide a variety of services with varying requirements, such as enhanced mobile broadband, ultra-reliable low-latency communication, and massive machine-type communication, wireless networks are envisioned to employ network slicing, flexible decentralized and / or distributed computing systems, and ubiquitous computing, where local spectrum licensing, spectrum sharing, infrastructure sharing, and intelligent automated management are supported by mobile edge computing, artificial intelligence (e.g., machine learning), tool-based technologies, cloudification, and blockchain. For example, in network slicing, multiple independent and dedicated network slice instances can be created within the same infrastructure to run services with varying requirements for latency, reliability, throughput, and mobility. In addition to the features listed above, 6G (sixth generation) networks are expected to utilize short packet communication, for example. Key features of 6G will include intelligent connectivity management and control, programmability, integrated sensing and communication, reduced energy footprint, trusted infrastructure, scalability, and affordability. In addition, 6G also targets new use cases that include integrating positioning and sensing capabilities into the system definition to unify the user experience across the physical and digital worlds.

[0025] The various techniques described in this article can also be applied to cyber-physical systems (CPS) (systems that control collaborative computing elements of physical entities). CPS can realize and utilize a large number of interconnected ICT devices (sensors, actuators, processors, microcontrollers, etc.) embedded in physical objects in different locations. Mobile cyber-physical systems, which are inherently mobile physical systems, are a subcategory of cyber-physical systems. Examples of mobile physical systems include mobile robots and electronic devices transported by humans or animals.

[0026] It is conceivable that models based on artificial intelligence (AI) / machine learning (ML) will be used to improve the performance of at least some air interface functions, thereby providing radio interfaces that support AI / ML. For example, to improve the performance of beam selection correlation processes, different beam prediction models that output information (e.g., identification information) on the strongest beam (optimal beam) can be used. This paper discusses different examples of test mechanisms used to test beam prediction models. This paper uses the principles and terminology of 5G (fifth generation) to describe the examples, but does not limit the examples or the terminology used for 5G. Those skilled in the art can apply the solutions and examples to other communication systems, such as those beyond 5G, such as 6G and 7G, providing the necessary properties.

[0027] Figure 1 An exemplary high-level network architecture is shown, illustrating only some functional entities. All functional entities are logical units, and their implementations may differ from those shown. Figure 1 The connections shown are logical connections; the actual physical connections may differ.

[0028] refer to Figure 1 The wireless network 100 or a system including a wireless network includes a device component 101 for device functions in the device domain, an access network component 102 for access network functions in the access network domain, a core network component 103 for core network functions in the core network domain, and a data network component 104 for data network functions in the data network domain.

[0029] Device component 101 may be any electrical device or apparatus 110 that can be connected to an access network and configured to wirelessly connect to an access network component 102 (e.g., access network device 120) providing, for example, a cell 121, over one or more communication channels 122 including one or more control channels. The physical link from device component 101 to access network component 102 to core network component 103 is referred to as an uplink or reverse link, while the physical link to the device component is referred to as a downlink or forward link. By way of example and not limitation, device component 101 may be referred to as a served device, downlink device, mobile device, terminal device, communication device, user equipment (UE), subscriber station (SS), portable subscriber station, mobile station (MS), or access terminal (AT). A non-limiting list of examples that may be included in or comprise device component 101 includes mobile phones, cellular phones, smartphones, Voice over Internet Protocol (VoIP) phones, wireless local loop phones, devices using wireless modems, portable computers, desktop computers, laptop embedded devices (LEEs), laptop mounted devices (LMEs), smart devices, multimedia devices, image capture terminal devices (such as digital cameras, gaming terminals, music storage and recycle bins), drones, vehicles, autonomous guided vehicles, autonomous connected vehicles, in-vehicle wireless terminal devices, wireless endpoints, Internet of Things (IoT) devices, industrial IoT devices, devices operating in the context of industrial and / or automated processing chains, consumer electronics devices, consumer IoT devices, mobile robots, mobile robotic arms, sensors, surveillance cameras, electronic health-related devices, medical monitoring devices, such as medical devices for remote surgery, wearable devices (such as smartwatches, smart rings, head-mounted displays (HMDs), personal devices, etc.). A device component may also be part of a group of device components that are considered a device component (i.e., a mobile device) by the wireless network.

[0030] The access network domain can be based on any type of access network, such as cellular access networks (e.g., 5G networks, 5G Advanced networks, 6G networks, etc.), non-terrestrial networks, traditional cellular radio access networks (e.g., 4G or older networks), or non-cellular access networks (e.g., wireless LANs), or any combination thereof. To provide wireless access, the access network includes access network components 102, such as access network apparatus 120 or access devices. Access device component 102 can provide one or more cells 121, each potentially having different cell accessibility, but the cells are provided by a single access device. However, overlapping cells may exist, such as macrocells provided by access devices operating in cooperation with access nodes providing smaller cells, or microcells, femtocells, or picocells that overlap at least partially within macrocells. A wide variety of access network components 102 exist. The access network component 102, or any example thereof, may include or be included in the access network component 102, including different types of base stations such as eNB, gNB, split gNB, transmit-receive points, network-controlled repeaters, nodes operatively coupled to one or more remote radio heads, satellites, donor nodes in integrated access and backhaul (IAB), fixed IAB nodes, mobile IAB nodes mounted on vehicles, etc. At least some of the devices in the access network may provide an abstraction platform to decouple the abstraction of network functions from the processing hardware.

[0031] Furthermore, it should be noted that some components can be multi-domain components. For example, device component 101 can also provide services to other device components, i.e., it also operates as access network component 102, such as a relay node, a mobile IAB node, or a mobile terminal portion of an IAB node. Therefore, the term mobile device is used herein for a device component or device component function in a multi-domain component, and the term access network device is used for an access network component or access network component function in a multi-domain component.

[0032] Core network component 103 forms one or more core networks. The core networks may be based on a non-standalone core network (e.g., an LTE-based network) or a standalone access network (e.g., a 5G core network). However, it should be understood that the core networks and core network component 103 may use any technology that enables the delivery of network services between devices and data networks.

[0033] The data network can be any network, such as the Internet, intranet, wide area network, etc. Different remote monitoring and / or data collection services for different use cases can be achieved through the data network and data network component 104.

[0034] exist Figure 1In the example shown, details of the test environment 105 or test setup or test architecture include a first device 120 and a second device 110. The first device 120 may be an access network component, and the second device 110 may be a device network component, or both devices may be access network components, or both devices may be device network components.

[0035] The second device 110 can be configured to transmit and receive data and different reference signals on multiple beams 122 (only a few beams are shown). A beam represents a resource, such as a Channel State Information Reference Signal (CSI-RS) resource or a Synchronization Signal Block (SSB) resource. A beam can be a narrow beam or a wide beam, and it can overlap with one or more narrow beams. Beams can be separated from each other by information indicating the beam (e.g., by identification information). For example, a beam can have an identifier that can be used in reporting to identify the beam whose value is being reported. The identifier can be an index value, such as a Channel State Information Reference Signal Resource Index (CRI) value or a Synchronization Signal Block Resource Index (SSBRI) value. It should be understood that any kind of identifier can be used. Furthermore, other kinds of information besides identification information can be used as information indicating the beam.

[0036] exist Figure 1 In the illustrated example, the second device 110 includes at least one beam prediction model (PM) 111 and is configured to perform training or inference of at least one model. The beam prediction model is an AI / ML model, such as a deep learning-based model. The beam prediction model 111 can be a spatial beam prediction model, a temporal beam prediction model (time beam prediction model), or a combination of spatial and temporal beam prediction models. Spatial beam prediction aims to predict the strongest transmit / receive beam in different spatial locations. Temporal beam prediction aims to predict the beam most likely to be used in the next time step. The input to the beam prediction model can be channel quality-related measurements, such as reference signal received power measurements. However, for testing the accuracy of the beam prediction model, the detailed structure of the beam prediction model and details about how to perform training or inference of the model are not meaningful and will not be discussed in detail here.

[0037] The first device 120 (e.g., a test device (TE)) or an access network component configured to include test device functionality may be pre-configured with a test configuration, or the test configuration may be transmitted from a service and orchestration management platform to, for example, the device, or a portion of the test configuration may be pre-configured and partially received. The test configuration may include test settings defining, for example, beams, beam transmission power, accuracy criteria, etc. During the testing phase, the first device will determine whether to verify the accuracy of the beam prediction model, as will be described in more detail below. When the first device 120 is an access network component or any network node configured with test device functionality in a real network environment, even if the second device 110 (or more precisely, the beam prediction model in the second device) performs poorly in the field, this can be avoided. However, a separate test environment can be used. In another implementation, a separate test environment is initially used to initially verify the model's accuracy, and then the network environment is used to determine whether to verify the accuracy of the beam prediction model for subsequent training / inference.

[0038] Figure 2 The device is shown (e.g., Figure 1 The example function of the first device 120 shown is that the device is configured to test or determine whether a beam prediction model, or simply a model, in another device (the second device) can predict the N strongest beams with sufficient accuracy.

[0039] refer to Figure 2 The device obtains (box 201) information about the M strongest beams at the second device. In other words, the device obtains information about the M strongest receiving beams. In one implementation, the device obtains this information by receiving information from the second device indicating the M strongest measurement beams. The device and / or another access network component may transmit reference signals on multiple beams, allowing the second device to obtain measurement results to determine the M strongest measurement beams. In another implementation, the device obtains information about the M strongest beams by deriving information from a test setup.

[0040] Furthermore, the device receives (box 202) information from the second device indicating the N strongest predicted beams output by the beam prediction model. In the example shown, the accuracy criterion is that the Top-K beams (K strongest beams, where K can be one or more) are included in the Top-N predicted beams (N strongest predicted beams). N, M, and K are positive integers, where M is less than or equal to N, and K is less than or equal to M. Non-limiting examples of the values ​​of M, N, and K are given below. Figure 4 Provided.

[0041] The information indicating the N strongest predicted beams can include the information indicating the M strongest measured beams as additional information.

[0042] The device then determines (box 203) whether at least K of the M strongest beams are among the N strongest predicted beams. For example, if the M strongest beams include the following beams B9, B1, B5, B4, and B7 starting from the strongest beam, and the N strongest predicted beams include the following beams B7, B2, B3, B4, B6, and B9 starting from the strongest predicted beam, and K=1, then the device determines that B9 is among the N strongest predicted beams. However, if K=3, then the device determines that not all of the K strongest beams B9, B1, and B5 are among the N strongest predicted beams, because the N strongest predicted beams do not include B1 and B5. In another example, the device may determine that accuracy is verified if the accuracy meets or exceeds a preset percentage limit. For example, the accuracy of the above example with a value of K=3 could be 33% (finding one of the two sets).

[0043] The device verifies the accuracy of the beam prediction model when (box 203: yes) at least K strongest beams are in N strongest predicted beams.

[0044] The device does not verify the accuracy of the beam prediction model (box 205) when at least K strongest beams are not among the N strongest predicted beams (box 203: No).

[0045] In one implementation, the device can be configured to transmit a verification result (verified / unverified) to a second device. The second device can use the verification result to determine whether to continue inference or training.

[0046] In one example implementation, when the device receives an indication from the second device as to whether the second device is performing inference or training of a beam prediction model, the device can be configured to perform exploitation. Figure 1 The described function.

[0047] In one example implementation that can be combined with other implementations, the device can be configured to: utilize Figure 2 At the beginning of the described function, configuration information for obtaining measurement results about the first beam set is transmitted to the second device, along with configuration information for obtaining predictions using measurement results about the second beam set as input. Different examples of the first and second beam sets are described below. Figure 4 Provided.

[0048] Figure 3 A device configured to predict the strongest beam using a beam prediction model, or simply a model, is shown (e.g., Figure 1 Example function of the second device 110 shown.

[0049] refer to Figure 3The device can perform (301), or can perform inference or training on the beam prediction model that outputs the X strongest predicted beams earlier. In other words, the beam prediction model can be tested while performing inference or training of the model, or the model trained / inferred earlier can be tested. Furthermore, another device may have already performed training or inference, and the model has been downloaded to this device and tested to verify its suitability for this device.

[0050] The device obtains (box 302) measurements about the beam. For example, even when Figure 3 Not shown, the device can also determine one or more measurements of the beam based on previously received configurations, and the device can accordingly measure the beam and / or enable one or more other entities to measure the beam. For example, the device and / or multiple entities can measure downlink beam reference signals, such as synchronization signal blocks, tracking reference signals, or channel state information reference signals. The measurement result may be a reference signal received power value. Furthermore, for clarity of description herein, it is assumed that the measurement results are sufficiently accurate for high prediction accuracy, i.e., their accuracy has virtually no impact on the accuracy of the predicted beam.

[0051] The device inputs (box 303) the measurement results or at least a portion thereof to a beam prediction model, which outputs X strongest predicted beams. The device then transmits (box 304) information indicating the N strongest predicted beams output by the beam prediction model to a first device. X and N are positive integers, with N less than or equal to X. When N is less than X, overhead is reduced compared to indicating X beams.

[0052] In one implementation, the previously received configuration may be a configuration for obtaining measurement results about a first beam set and inputting measurement results about a second beam set into a beam prediction model, such as referring to... Figure 4 A more detailed description.

[0053] Figure 4 A high-level information exchange is illustrated. A non-limiting example of information exchange will be described below. Optional features, i.e., information exchange or functionality, are shown using dashed lines, either of which can be omitted or implemented in the described non-limiting example. Furthermore, it should be understood that details in the example can be added to another example.

[0054] The second device may be configured, for example, by the first device or by some other network entity (Box 4-1), with an AI / ML mode for beam management. For example, the model to be trained / inferred, along with the relevant parameters used for training / inference, may have been transferred to the second device.

[0055] First Example In the first example, the first device transmits (message 4-2) a command to switch to the AI / ML-based mode. This command could be "Enable AI / ML-based beam management use case". The goal of message 4-2 is to ensure that the mode operating on the second device side is known on the first device side.

[0056] The second device can then confirm the activation of the mode requested in (message 4-3), or that it is already in the requested mode. This mode can also be indicated indirectly. For example, message 4-3 could be a functional indication of whether the second device is performing inference or training of a beam prediction model.

[0057] Then, the second device configures (message 4-9) the first device to report at least N strongest predicted beams. The second device may also configure (message 4-9) the second device to report measurement results, for example, to report M strongest measured beams simultaneously with the N strongest predicted beams.

[0058] The second device then performs the example Top-K beam identifier (ID) prediction shown in Box 4-10. In other words, the second device inputs the measurement results into the beam prediction model that outputs X strongest predicted beams, or outputs the predicted beam identifiers in the example. One or more Top-K beams can be a subset of the X strongest predicted beams starting from the strongest predicted beam.

[0059] The second device reports (Message 4-11) at least the predicted Top-K strongest beam identifiers to the first device. The second device can report M strongest measurement beams either individually (Message 4-12) or together with the Top-K predicted beam identifiers to the first device. The second device can report M strongest measurement beams even if it is not configured to report them.

[0060] Then, the first device determines (box 4-13) the accuracy of the beam prediction model. In other words, as described above, the first device determines whether to verify the accuracy. This determination may include verifying whether the strongest beam identifier (known to the first device or received from the second device) is one of the predicted Top-K strongest beam identifiers.

[0061] Second example In the second example applicable to both spatial and temporal beam prediction, two beam sets are used: a first beam set (referred to as set A in the example) and a second beam set (referred to as set B in the example). In the second example, set B is a subset of set A, and the beam prediction model is configured to predict the Top-K identifiers of the beams in set A.

[0062] In the second example, the first device transmits (message 4-2) a command to the second device to switch to the function based on AI / ML mode.

[0063] The second device transmits (message 4-3) confirmation (of its operation in AI / ML BM mode) to the first device. (Message 4-3 is optional).

[0064] The second device transmits (message 4-4) a function instruction to the first device, such as an instruction on whether the second device performs (runs) inference or training in the spatial domain Top-K downlink beam management or the temporal domain Top-K downlink beam management.

[0065] The first device checks (boxes 4-5) whether the second device has executed the instruction for inference for Top-K beam ID prediction. In the example shown, it is assumed that the second device executes the inference for Top-K beam ID prediction, and information exchange can continue.

[0066] The first device transmits (messages 4-6) a configuration to the second device for measuring the entire set A, i.e., all beams in set A. For example, message 4-6 may include the configuration and the Channel State Information Reference Signal (CSI-RS) resources for measuring the entire set A. The second device then measures the total Reference Signal Received Power (RSRP) values ​​for all beams in set A.

[0067] The first device transmits (message 4-7) a configuration to the second device for measuring the entire set A. Message 4-7 may include this configuration and Channel State Information Reference Signal (CSI-RS) resources for fixing the set B beams. For example, the second device may be configured to use 16 set B beams to predict the Top-1, Top-4, or Top-8 of the 64 set A beams. It should be understood that set A may have 32, 64, or 256 beams, while set B beams will be at least one-quarter (1 / 4) of set A; for example, if set A has 64 beams, then set B should be configured with 16 beams.

[0068] The first device also transmits (Message 4-8) to the second device a request to prepare the RSRP value for the Top-K strongest beam. The predicted Top-K strongest beam may be within set B or include beams in set A that are not in set B (excluding set B). The purpose of the request (Message 4-8) is to instruct the second device to sort the beams and obtain the Top-K beams from the measurements. (Message 4-8 is optional.) Then, the first device transmits (message 4-9) to the second device a request to report the Top-K beam identifier prediction results. Message 4-9 may include a request to report the Top-K strongest beam identifier in set A beam measurements.

[0069] Then, the second device performs (Box 4-10) prediction of the (ID) of set A beams using the RSRP values ​​(primary measurement) of set B beams as input to the beam prediction model in the example Top-K beam identifiers shown. This beam prediction model outputs the predicted Top-K strongest beam identifiers for set A. In other words, only a subset of the RSRP values ​​of set A are input to the beam prediction model.

[0070] Then, the second device reports (message 4-11) the predicted Top-K strongest beam identifier to the first device. In other words, message 4-11 transmits the output of the beam prediction model or a portion thereof.

[0071] The second device can report (message 4-12) the Top-K strongest beam identifier of the beams in set A to the first device based on the measurement results of the entire set A. As described above, the first device can have information available at the first device.

[0072] The first device then determines (Box 4-13) the accuracy of the beam prediction model. For example, the first device can use the Top-K strongest beam identifiers from set A to verify the predicted Top-K strongest beam identifiers. If the predicted Top-K strongest beam identifiers include the strongest beam identifier, the first device can compare the predicted Top-K strongest beam identifiers with the measured Top-K strongest beam identifiers. According to one example condition, if they are the same, the test passes and the accuracy is verified; otherwise, the model predicts an incorrect beam, and the test fails, resulting in unverified accuracy. According to another condition, if the predicted Top-K strongest beam identifiers include the identifier of the strongest measured beam, the test passes; otherwise, the test fails. In another option, the first device can know the conventional beam with the strongest RSRP and can compare the predicted Top-K strongest beam identifiers with the conventional beam identifiers. According to one example condition, if they are the same, the test passes. According to another condition, if the predicted Top-K strongest beam identifiers include the identifier of the strongest beam, the test passes; otherwise, the test fails. It should be noted that using a conventional beam with the strongest RSRP can produce a more limited number of beam identifiers than using measurement results. Therefore, the latter may be better for testing purposes.

[0073] The first device can report (message 4-15) the test results (e.g., pass / fail or verified / unverified) to the second device, and the second device can then use at least the reported information to decide (box 4-15) whether to continue inferring / training the beam prediction model.

[0074] Third Example In the third example applicable to both spatial and temporal beam prediction, two beam sets are used: a first beam set (referred to as set A in the example) and a second beam set (referred to as set B in the example). In this third example, set B differs from set A; set B includes one or more wide beams, while set A includes narrow beams. The beam prediction model is configured to predict the Top-K identifiers of the beams in set A. In other words, wide beams are used to predict narrow beams, enabling the selection of the best narrow beam from the best wide beams. For a wide beam in set B, there will be one or more narrow beams in set A that can overlap with the wide beam.

[0075] In the third example, the first device transmits (message 4-2) a command to the second device to switch to the AI / ML-based mode.

[0076] The second device transmits (message 4-3) confirmation (of its operation in AI / ML BM mode) to the first device. (Message 4-3 is optional).

[0077] The second device transmits (message 4-4) a function instruction to the first device, such as an instruction on whether the second device performs (runs) inference or training in the spatial domain Top-K downlink beam management or the temporal domain Top-K downlink beam management.

[0078] The first device checks (boxes 4-5) whether the second device has executed the instruction for inference for Top-K beam ID prediction. In the example shown, it is assumed that the second device executes the inference for Top-K beam ID prediction, and information exchange can continue.

[0079] The first device transmits (messages 4-6) for measuring the configuration and synchronization signal block (SSB) or CSI-RS resources of fixed set B. For example, the second device can be configured to use 16 set B beams (wide beams) to predict the Top-1, Top-4, or Top-8 of 64 set A beams (narrow beams).

[0080] The first device transmits (messages 4-7) a pointer (indicator) to set B configured as a CSI-RS or SSB. In other words, the first device transmits the pointer to identify to the second device that set B is a wide-beam set.

[0081] The first device can transmit (Message 4-8) to the second device to request the RSRP value of the Top-K strongest beam. The purpose of the request (Message 4-8) is to instruct the second device to sort the beams and obtain the Top-K beams from the measurements. (Message 4-8 is optional.) In one implementation, the first device may transmit the configuration for measuring set A to the second device in a separate message, or in messages 4-7 or 4-8.

[0082] Then, the first device transmits (message 4-9) to the second device a request to report the Top-K beam identifier prediction results. Message 4-9 may include a request to report the Top-K strongest beam identifier in set A beam measurements.

[0083] Then, the second device uses the RSRP values ​​of set B beams (wide beams) as input to the beam prediction model to perform the example Top-K beam identifier (ID) prediction of set A beams shown in Box 4-10, which outputs the predicted Top-K strongest beam identifiers of set A (narrow beams).

[0084] Then, the second device reports (message 4-11) the predicted Top-K strongest beam identifier to the first device. In other words, message 4-11 transmits the output of the beam prediction model or a portion thereof.

[0085] The second device can report (message 4-12) the Top-K strongest beam identifier of beam set A to the first device based on the measurement results of set A. As described above, the first device can have information available at the first device.

[0086] The first device then determines (Box 4-13) the accuracy of the beam prediction model. For example, the first device can use the Top-K strongest beam identifiers from set A to verify the predicted Top-K strongest beam identifiers. If the predicted Top-K strongest beam identifiers include the strongest beam identifier, the first device can compare the predicted Top-K strongest beam identifiers with the measured Top-K strongest beam identifiers. According to one example condition, if they are the same, the test passes and the accuracy is verified; otherwise, the model predicts an incorrect beam, and the test fails, resulting in unverified accuracy. According to another condition, if the predicted Top-K strongest beam identifiers include the identifier of the strongest measured beam, the test passes; otherwise, the test fails. In another option, the first device can know the conventional beam with the strongest RSRP and can compare the predicted Top-K strongest beam identifiers with the conventional beam identifiers. According to one example condition, if they are the same, the test passes. According to another condition, if the predicted Top-K strongest beam identifiers include the identifier of the strongest beam, the test passes; otherwise, the test fails. It should be noted that using a conventional beam with the strongest RSRP can produce a more limited number of beam identifiers than using measurement results. Therefore, the latter may be better for testing purposes.

[0087] The first device can report (message 4-15) the test results (e.g., pass / fail or verified / unverified) to the second device, and the second device can then use at least the reported information to decide (box 4-15) whether to continue inferring / training the beam prediction model.

[0088] As can be seen from the above example, an AI / ML-based beam prediction model can be tested for a real-time monitoring mechanism. Furthermore, since the second device is configured to report the Top-K beams in set A (not the entire set A, and excluding set B), the overhead is significantly reduced.

[0089] The above passed Figures 1 to 4 The described blocks, related functions, and information exchanges (messages, signaling) do not have an absolute temporal order, and some of them may be executed simultaneously or in a different order than given. Other functions may also be executed between or within them, and additional information and / or rules may be transferred. For example, more than one test cycle may be used to verify accuracy, during which reference signals are transmitted to obtain unused inputs to the beam prediction model. Some blocks or portions of blocks, or one or more pieces of information, may also be omitted or replaced by corresponding blocks or portions of blocks, or one or more pieces of information.

[0090] Figure 5 A device 501, such as a first device, is shown, which can be configured to perform tests or accuracy verification of AI / ML-based beam prediction models. Figure 6 It shows that it can be achieved Figure 5 The device shown is a distributed functional device. Figure 7 A device (e.g., a second device) is shown that can be configured to at least input measurement results into an AI / ML-based beam prediction model and report its output.

[0091] According to one embodiment, an apparatus is provided, comprising: means for receiving from a second means information indicating N strongest predicted beams output by a beam prediction model; means for obtaining information about M strongest beams at the second means; means for determining whether at least K of the M strongest beams are among the N strongest predicted beams; and means for verifying the accuracy of the beam prediction model when at least K strongest beams are among the N strongest predicted beams; wherein N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M.

[0092] According to one embodiment, an apparatus is provided, comprising: components for performing inference or training on a beam prediction model that outputs X strongest predicted beams; components for obtaining measurement results about the beams; components for inputting the measurement results into the beam prediction model; and components for transmitting information indicating N strongest predicted beams output by the beam prediction model to a first apparatus, wherein X and N are positive integers, and N is less than or equal to X.

[0093] Devices 501, 701 may include one or more communication control circuitry systems 520, 720 (such as at least one processor) and at least one memory 530, the at least one memory 530 including one or more algorithms 531, 731 (such as computer program code (software)), wherein the at least one memory and the computer program code (software) are configured, together with the at least one processor, to cause the device to perform any of the exemplary functions of the corresponding device, as referenced above. Figures 1 to 4 Any of the descriptions. The at least one memory 530, 730 may also include at least one database 532, 732.

[0094] refer to Figure 5 One or more communication control circuits 520 of device 501 include at least an AI / ML test circuit 521, configured to determine and transmit test-related configurations, commands, or requests, and to determine whether to verify the accuracy of the AI / ML-based beam prediction model. For this purpose, the AI / ML test circuit 521 of device 501 is configured, for example, by means of... Figure 1 , Figure 2 and Figure 4 Use one or more separate circuits to perform at least some of the functions described above.

[0095] refer to Figure 5 The memory 530 can be implemented using any suitable data storage technology, such as semiconductor-based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and removable memory.

[0096] refer to Figure 5The device 501 may also include different interfaces 510, such as one or more communication interfaces (TX / RX), which include hardware and / or software for establishing communication connections according to one or more communication protocols. The one or more communication interfaces 510 may enable connection to the core network and / or radio access network of the Internet and / or wireless communication networks and / or other devices within the device range. The one or more communication interfaces 510 may provide the device with communication capabilities to communicate in a cellular communication system and enable communication with second devices (such as different network nodes or components or equipment parts, such as mobile devices, such as terminal equipment or user equipment). The one or more communication interfaces 510 may include standard, well-known components controlled by a corresponding control unit, such as amplifiers, filters, frequency converters, (de)modulators, and encoder / decoder circuitry, and possibly one or more antennas.

[0097] In one embodiment, such as Figure 6 As shown, Figure 5 At least some functions of the device can be shared between two physically separate devices, thereby forming an operational entity. Therefore, the device can be seen as an operational entity comprising one or more physically separate devices for performing at least some of the processes described. Thus, utilizing this shared architecture... Figure 6 The apparatus may include a control unit CU 620 or a remote control unit, such as a host computer or server computer, operatively coupled (e.g., via a wireless or wired connection) to a remotely distributed unit DU 622 located in or in a remote head of the first apparatus. In embodiments, at least some of the described processes may be performed by CU 620. In embodiments, the execution of at least some of the processes described may be shared between DU 622 and CU 620.

[0098] Similar to Figure 5 , Figure 6 The device may include one or more communication control circuits (CNTL) 520, such as at least one processor, and at least one memory (MEM) 530, including one or more algorithms (PROG) 531, such as computer program code (software), wherein the at least one memory and the computer program code (software) are configured, together with the at least one processor, to cause the device to perform any of the exemplary functions of the first device described above, such as via Figure 1 , Figure 2 and Figure 4 .

[0099] refer to Figure 7 One or more communication control circuits 720 of device 701 include at least a beam prediction circuit 721, which is configured to predict and report beams, which will combine Figures 1 to 4 To this end, the beam prediction circuit 721 of device 701 is configured, for example, by means of... Figure 1 , Figure 3 and Figure 4 For example, one or more separate circuits may be used to perform at least some of the functions of the second device described above.

[0100] refer to Figure 7 The memory 730 can be implemented using any suitable data storage technology, such as semiconductor-based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and removable memory.

[0101] refer to Figure 7 The device 701 may also include different interfaces 710, such as one or more communication interfaces (TX / RX) including hardware and / or software for implementing communication connections according to one or more communication protocols. One or more communication interfaces 710 can enable connection to a wireless access network and / or other devices within the device's range and / or the Internet and / or the core network of a wireless communication network. One or more communication interfaces 710 can provide the device with communication capabilities to communicate in a cellular communication system and to communicate with different network nodes or components. One or more communication interfaces 710 may include standard, well-known components controlled by a corresponding control unit, such as amplifiers, filters, frequency converters, (de)modulators, and encoder / decoder circuitry, as well as one or more antennas.

[0102] In an embodiment, CU 620 can generate a virtual network through which CU 620 communicates with DU 622. Typically, virtual networking can involve the process of combining hardware and software network resources and network functions into a single software-based management entity (virtual network). Network virtualization can involve platform virtualization, often combined with resource virtualization. Network virtualization can be categorized into external virtual networking, which combines many networks or portions of networks into a server computer or host computer (e.g., to a CU). The goal of external network virtualization is optimized network sharing. Another type is internal virtual networking, which provides network-like functionality to software containers on a single system. Virtual networking can also be used to test terminal devices.

[0103] In this embodiment, the virtual network can provide a flexible operational distribution between the DU and CU. In fact, any digital signal processing task can be performed in either the DU or CU, and the boundaries of responsibility transferred between the DU and CU can be selected based on the implementation.

[0104] As used herein, the term "circuit" may refer to one or more of the following: (a) a hardware circuit implementation only, such as an implementation in analog and / or digital circuits only; and (b) a combination of hardware circuits and software (and / or firmware), such as (if applicable): (i) a combination of analog and / or digital hardware circuits with software / firmware; and (ii) a hardware processor with any part of the software, including digital signal processors, software, and memory that work together to enable devices such as terminal devices or access nodes to perform various functions; and (c) hardware circuits and processors, such as microprocessors or portions thereof, that require software (e.g., firmware) to operate, but which may be absent when operation is not required. This definition of "circuit" applies to all uses of the term in this application (including any claims). As another example, as used herein, the term "circuit" also covers an implementation of hardware circuits or processors (or processors) or portions thereof and their accompanying software and / or firmware. For example, and if applicable to a particular claim element, the term "circuit" also covers a baseband integrated circuit for an access node or terminal device or other computing or networking device.

[0105] In one embodiment, combined Figures 1 to 4 At least some of the processes described can be performed by means including corresponding components for performing at least some of the processes described. Some example components for performing the processes may include at least one of the following: a detector, a processor (including dual-core and multi-core processors), a digital signal processor, a controller, a receiver, a transmitter, an encoder, a decoder, a memory, RAM, ROM, software, firmware, a display, a user interface, display circuitry, user interface circuitry, user interface software, display software, a circuit, an antenna, an antenna circuit, and a circuit. In one embodiment, at least one processor, memory, and computer program code form a processing means or include one or more portions of computer program code for performing the processes according to... Figures 1 to 4 One or more operations or operations of any of the embodiments thereof.

[0106] The embodiments and examples described can also be performed in the form of a computer process defined by a computer program or parts thereof. Figures 1 to 4Embodiments of the described functionality can be performed by executing at least a portion of a computer program including the corresponding instructions. The computer program can be provided as a computer-readable medium including program instructions stored thereon, or as a non-transitory computer-readable medium including program instructions stored thereon. The computer program can be in source code form, object code form, or some intermediate form, and it can be stored in some kind of carrier, which can be any entity or device capable of carrying the program. For example, the computer program can be stored on a computer or processor-readable computer program distribution medium. The computer program medium can be, for example, but not limited to, recording media, computer memory, read-only memory, electrical carrier signals, telecommunication signals, and software distribution packages. The computer program medium can be a non-transitory medium. As used herein, the term "non-transitory" is a limitation on the medium itself (i.e., tangible, not signaling), not a limitation on the persistence of data storage (e.g., RAM and ROM). The coding of the software used to perform the embodiments shown and described is entirely within the scope of those skilled in the art.

[0107] Although embodiments have been described above with reference to examples in conjunction with the accompanying drawings, it is clear that the embodiments are not limited thereto, but can be modified in several ways within the scope of the appended claims. Therefore, all words and expressions should be interpreted broadly, and they are intended to illustrate rather than limit the embodiments. It will be apparent to those skilled in the art that the inventive concept can be implemented in various ways as technology advances. Furthermore, it will be clear to those skilled in the art that the described embodiments can, but are not required to, be combined with other embodiments in various ways.

Claims

1. A first apparatus comprising at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the first apparatus at least to: receive, from a second apparatus, information indicating N strongest predicted beams output by a beam prediction model; obtain, at the second apparatus, information about M strongest beams; determine whether at least K strongest beams of the M strongest beams are among the N strongest predicted beams; and verify accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams, wherein N, M and K are positive integers, M is less than or equal to N, and K is less than or equal to M.

2. The first apparatus of claim 1, wherein the at least one processor and the at least one memory store instructions that, when executed by the at least one processor, further cause the first apparatus at least to obtain the information about M strongest beams at the second apparatus by receiving, from the second apparatus, further information indicating M strongest measured beams in the information or by deriving the information about M strongest beams from a test setup configuration.

3. The first apparatus of claim 2, wherein the at least one processor and the at least one memory store instructions that, when executed by the at least one processor, further cause the first apparatus at least to: receive, from the second apparatus, prior to the information, an indication of whether the second apparatus is performing inference or training of the beam prediction model; transmit, to the second apparatus, when the second apparatus is performing inference or training of the beam prediction model, configuration information for obtaining measurements with respect to a first set of beams and configuration information for obtaining predictions using measurements with respect to a second set of beams as input.

4. The first apparatus of claim 3, wherein the second set of beams is a subset of the first set of beams.

5. The first apparatus of claim 3, wherein the second set of beams comprises different beams than the first set of beams.

6. The first apparatus of claim 3, 4 or 5, wherein the first set of beams comprises a first number of beams and the second set of beams comprises a second number of beams, the second number being at least one quarter of the first number.

7. The first apparatus of any one of the preceding claims, wherein the at least one processor and the at least one memory store instructions that, when executed by the at least one processor, further cause the first apparatus at least to: transmit, to the second apparatus, prior to receiving the information, a command to switch to a function based on artificial intelligence / machine learning mode and / or a request for reporting at least N strongest predicted beams.

8. The first apparatus of any one of the preceding claims, wherein the information indicating beams are beam identifiers. ​ ​ ​ ​ ​ ​ ​ ​ 9. The first apparatus according to any one of the preceding claims, wherein the at least one processor and the at least one memory store instructions, the instructions, when executed by the at least one processor, further cause the first apparatus to at least: The second device is transmitted information indicating whether the accuracy of the beam prediction model has been verified.

10. A second means comprising at least one processor and at least one memory storing instructions, the instructions, when executed by said at least one processor, causing the second means to at least: Perform inference or training on the beam prediction model that outputs the X strongest predicted beams; Obtain measurement results regarding the beam; The measurement results are input into the beam prediction model; The information indicating the N strongest predicted beams output by the beam prediction model is transmitted to the first device. Where X and N are positive integers, and N is less than or equal to X.

11. The second apparatus of claim 10, wherein the at least one processor and the at least one memory store instructions, the instructions, when executed by the at least one processor, further cause the second apparatus to at least: The device receives configuration information for obtaining measurement results about a first beam set and configuration information for obtaining a prediction using measurement results about a second beam set as input. The measurement results for the second beam set are input into the beam prediction model; as well as Further information indicating the M strongest measurement beams at the second device is transmitted to the first device, where M is a positive integer less than or equal to N.

12. The second apparatus according to claim 10 or 11, wherein the at least one processor and the at least one memory store instructions, the instructions, when executed by the at least one processor, further cause the second apparatus to at least: The first device is at least informed whether the second device is performing inference or training of the beam prediction model.

13. The second apparatus according to claim 10, 11, or 12, wherein the at least one processor and the at least one memory storing instructions, when executed by the at least one processor, further cause the second apparatus to at least: Receive a command from the first device to switch to a function based on artificial intelligence / machine learning mode; Upon receiving the command and while in the AI / machine learning-based mode, information confirming that the second device is in the AI / machine learning-based mode is transmitted to the first device.

14. A method comprising at least: Receive information from the second device indicating the N strongest predicted beams output by the beam prediction model; Information about the M strongest beams is obtained at the second device; Determine whether at least K of the M strongest beams are among the N strongest predicted beams; and Verify the accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams. Where N, M, and K are positive integers, M is less than or equal to N, and K is less than or equal to M.

15. A method comprising at least: Perform inference or training on the beam prediction model that outputs the X strongest predicted beams; Obtain measurement results regarding the beam; The measurement results are input into the beam prediction model; The information indicating the N strongest predicted beams output by the beam prediction model is transmitted to the first device. Where X and N are positive integers, and N is less than or equal to X.

16. A computer-readable medium comprising instructions that, when executed by a device, cause the device to perform at least one of a first process and a second process. The first process includes at least: Receive information from the second device indicating the N strongest predicted beams output by the beam prediction model; Information about the M strongest beams is obtained at the second device; Determine whether at least K of the M strongest beams are among the N strongest predicted beams; and Verify the accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams. Where N, M, and K are positive integers, M is less than or equal to N and K is less than or equal to M, and The second process includes at least: Perform inference or training on the beam prediction model that outputs the X strongest predicted beams; Obtain measurement results regarding the beam; The measurement results are input into the beam prediction model; The information indicating the N strongest predicted beams output by the beam prediction model is transmitted to the first device. Where X and N are positive integers, and N is less than or equal to X.

17. The computer-readable medium of claim 16, wherein the computer-readable medium is a non-transitory computer-readable medium.

18. A computer program comprising instructions, said instructions, when executed by a device, causing the device to perform at least one of a first process and a second process. The first process includes at least: Receive information from the second device indicating the N strongest predicted beams output by the beam prediction model; Information about the M strongest beams is obtained at the second device; Determine whether at least K of the M strongest beams are among the N strongest predicted beams; and Verify the accuracy of the beam prediction model when at least the K strongest beams are among the N strongest predicted beams. Where N, M, and K are positive integers, M is less than or equal to N and K is less than or equal to M, and The second process includes at least: Perform inference or training on the beam prediction model that outputs the X strongest predicted beams; Obtain measurement results regarding the beam; The measurement results are input into the beam prediction model; The information indicating the N strongest predicted beams output by the beam prediction model is transmitted to the first device. Where X and N are positive integers, and N is less than or equal to X.

Images