Method and communication device for transmitting training data for an AI model

Abstract

The present application provides a method for transmitting training data for an AI model, which is applied to a scenario in which a reporting network element of the training data transmits the training data to a training network element of the AI ​​model. After obtaining the training data, the reporting network element classifies the training data into reference data and non-reference data, and indicates the training data by transmitting first and second types of information to the training network element. The first type of information includes complete information of the reference data, and the second type of information includes incremental information of the non-reference data relative to the reference data corresponding to the non-reference data. To help reduce the overhead of transmitting the training data, thereby helping to reduce air interface overhead, reduce transmission delays, and improve model training performance in the AI ​​model training or update process, the training data is indicated using the complete information of the reference data and the incremental information corresponding to the non-reference data.

Description

[Technical Field]

[0001] This application claims priority to Chinese Patent Application No. 202211730492.0, entitled "Method and Communication Apparatus for Transmitting Training Data for AI Models," filed with the State Intellectual Property Office of China on December 30, 2022, which is incorporated herein by reference in its entirety.

[0002] TECHNICAL FIELD Embodiments of the present application relate to the field of machine learning, and more particularly to a method and communication device for transmitting training data for an AI model. [Background technology]

[0003] Currently, artificial intelligence (AI) has been introduced into wireless communication networks and has been widely applied in many application scenarios of air interface technologies, such as AI-based channel state information (CSI) prediction, AI-based CSI feedback, and AI-based positioning.

[0004] When an AI model is applied to air interface technology, both offline and online model updating / training require data collection in the actual deployed network to form the training dataset required for model updating / training. If the training network element of the AI ​​model and the reporting network element of the training data are not located in the same network element, the reporting network element needs to transmit (or feed back, provide, report, etc.) the training data to the training network element. With existing training data transmission solutions, the training data is usually transmitted periodically or continuously, resulting in high transmission overhead. Summary of the Invention

[0005] The present application provides a method and communication apparatus for transmitting training data for an AI model to help reduce the overhead of transmitting the training data. [Means for solving the problem]

[0006] According to a first aspect, there is provided a method for transmitting training data for an AI model, the method being applicable to a training network element for the training data for the AI ​​model, the method comprising: a first network element receives first information from a second network element, the first information including a first type of information and a second type of information, the first type of information and the second type of information indicating training data for an AI model provided by the second network element; The training data includes one or more reference data and non-reference data, the first type of information includes complete information of the one or more reference data, and the second type of information includes incremental information of the non-reference data relative to the reference data corresponding to the non-reference data.

[0007] According to the method for transmitting training data provided in the present application, the transmitting network element of the training data no longer directly transmits the training data, but classifies the training data into reference data and non-reference data, and all the training data is indicated to the first network element by transmitting the complete information of the reference data and the incremental information of the non-reference data relative to the reference data corresponding to the non-reference data to the first network element, so that the overhead of transmitting the training data can be reduced.

[0008] This helps reduce the overhead of transmitting training data while providing other beneficial technical effects.

[0009] For example, after the overhead of transmitting training data is reduced, when the reporting network element reports the same amount of training data to the training network element, the technical solution of the present application can use fewer bits to represent the training data, and occupy less air interface resources compared with the direct transmission of the training data, which helps to reduce the air interface overhead in the AI ​​model training process.

[0010] In another example, with the same air interface resources occupied, after the overhead of transmitting training data is reduced, the reporting network element may indicate more training data to be used for AI model training or updating, which may result in improved AI model training performance, which helps improve the performance of communication systems that use AI models.

[0011] In another example, after the overhead of transmitting training data is reduced, the technical solution of the present application can reduce the transmission delay compared with directly transmitting training data at the same transmission rate.

[0012] The above beneficial technical effects can improve the performance of communication systems in which AI models are deployed.

[0013] With reference to the first aspect, in some implementations of the first aspect, before the first network element receives the first information from the second network element, the method includes: The method further includes the first network element sending first instruction information to the second network element, and the first instruction information being used to determine the first clustering algorithm.

[0014] In this embodiment, the first network element indicates a first clustering algorithm to the second network element, so that the second network element performs a clustering process on the collected training data of the AI ​​model based on the first clustering algorithm to classify the training data into reference data and non-reference data, and the second network element reports information obtained by clustering the training data, specifically, the complete information of the reference data and the incremental information corresponding to the non-reference data, which reduces the overhead of transmitting training data.

[0015] Referring to the first aspect, in some implementation forms of the first aspect, the AI ​​model is applied to uplink positioning, the first network element includes a location management function (LMF) network element, and the second network element includes an access network device; The method comprises: the first network element receives third information from a third network element, the third information including first type information and second type information, the first type information and the second type information included in the third information indicating training data provided by the third network element; the training data includes training data provided by a second network element and training data provided by a third network element, the training data provided by the second network element includes channel measurement results, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the training data provided by the third network element includes location information of the third network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data; The first type of information included in the first information includes complete information of channel measurement results belonging to reference data, and the second type of information included in the first information includes incremental information corresponding to channel measurement results belonging to non-reference data; The first type of information included in the third information includes complete information of location information belonging to the reference data, and the second type of information included in the third information includes incremental information corresponding to location information belonging to the non-reference data; It further includes:

[0016] When the technical solution provided in the present application is applied to an uplink positioning scenario, the access network device (i.e., the second network element) and / or the PRU / UE (i.e., the third network element) performs / performs a clustering process on the provided training data to classify the training data into reference data and non-reference data, and reports / reports the complete information of the reference data and the incremental information corresponding to the non-reference data, thereby reducing the overhead of reporting the training data.

[0017] According to a second aspect, there is provided a method for transmitting training data for an AI model, which may be applied to a providing network element (also called a reporting network element) of the training data for the AI ​​model, the method comprising: the second network element sends first information to the first network element, the first information including first type information and second type information, the first type information and the second type information indicating training data for the AI ​​model provided by the second network element; the training data includes one or more reference data and non-reference data, the first type of information includes complete information of the one or more reference data, and the second type of information includes incremental information of the non-reference data relative to the reference data corresponding to the non-reference data; This includes:

[0018] Referring to the second aspect, in some implementations of the second aspect, the method includes: a second network element obtains candidate training data; The second network element filters the candidate training data to obtain training data. It further includes:

[0019] Optionally, the filtering criteria of the training data is not limited in the present application. For example, the filtering criteria may be the following: thresholds for quality indicators of the training data and criteria for determining the quality indicators; A threshold for the amount of training data that satisfies the criteria for determining the quality indicator, and a criterion for determining the amount of training data; In one example, the quality indicator of the training data may be the SINR of the training data and the amount of training data, and the criteria for determining the SINR and the criteria for determining the amount of training data may be that the SINR is greater than or equal to Q, the amount of training data is greater than or equal to N, and both Q and N are integers.

[0020] To ensure reliable AI air interface performance, the model needs to be maintained or switched to non-AI mode in a timely manner. The network element that reports the training data can filter the collected data, so that redundant and invalid data can be removed. This helps reduce the air interface overhead / backhaul overhead caused by training / updating.

[0021] Optionally, performing a clustering process on the training data by the second network element comprises: The second network element may include filtering out data that is not in any cluster, for example discarding invalid outlier information.

[0022] In some implementations of the first or second aspect, the one or more reference data are determined from the training data based on a first clustering algorithm.

[0023] In some implementations of the first or second aspect, the first instruction information further indicates to the second network element to collect training data for the AI ​​model.

[0024] According to this embodiment, the first instruction information, in addition to indicating the clustering algorithm, also indicates to the second network element to collect training data, so that instruction overhead can be reduced.

[0025] In some implementations of the first or second aspect, the method includes: The method further includes the first network element sending second instruction information to the second network element, the second instruction information indicating to the second network element to collect training data for the AI ​​model.

[0026] According to this implementation, the clustering algorithm of the AI ​​model and the training data that the first network element indicates the second network element should start collecting are indicated by using separate messages, so that the clustering algorithm can be flexibly adjusted. Furthermore, it is convenient for the first network element to indicate that the second network element should collect training data based on model training / update requirements.

[0027] In some implementations of the first or second aspect, the AI ​​model is applied to the downlink positioning, the first network element includes a location management function (LMF) network element, and the second network element includes a terminal device; The training data includes channel measurements and / or location information of a second network element, the channel measurements being obtained by the second network element based on a positioning reference signal coming from a third network element.

[0028] When the technical solution provided in this application is applied to a downlink positioning scenario, the PRU / UE (i.e., the second network element) performs a clustering process on the training data obtained by measuring the positioning reference signal of the third network element (e.g., the access network device), classifies the training data into reference data and non-reference data, and reports the complete information of the reference data and the incremental information corresponding to the non-reference data, thereby reducing the overhead of reporting the training data.

[0029] Optionally, the channel measurement result is obtained by the second network element based on a positioning reference signal coming from a third network element, The channel measurement result is obtained by the second network element based on a plurality of positioning reference signals, the plurality of positioning reference signals coming from one or more third network elements, and the plurality of positioning reference signals differing from each other in terms of one or more of a transmission time point, a transmission frequency, an occupied frequency band, and a transmission port.

[0030] According to this embodiment, the training data in the uplink positioning scenario can be obtained in multiple ways, which helps to improve the implementation flexibility of the technical solution, so that the technical solution can be implemented in different application scenarios.

[0031] In some implementations of the first or second aspect, an AI model is applied to the uplink positioning; The first network element includes a location management function (LMF) network element, the second network element includes an access network device, the training data includes channel measurement results, the channel measurement results are acquired by the second network element based on a sounding reference signal coming from a third network element, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data.

[0032] According to this embodiment, in an uplink positioning scenario, the base station performs clustering on the collected training data, and feeds back the complete information of some training data (specifically, reference data) and the incremental information of some training data (specifically, non-reference data) to the LMF network element. Then, the LMF network element restores all the training data. This can reduce the air interface feedback overhead caused by model training / updating.

[0033] Optionally, in the AI-based uplink positioning, channel measurement results are provided by one or more second network elements (e.g., base stations) for a first network element (e.g., an LMF network element). For one second network element, the channel measurement results are obtained by the second network element by measuring sounding reference signals coming from one or more third network elements (e.g., terminals). This is not limiting. Optionally, different sounding reference signals differ from each other in terms of one or more of a transmission time instant, a transmission frequency, an occupied frequency band, and a transmission port.

[0034] In some implementations of the first or second aspect, the AI ​​model is applied to the uplink positioning, the first network element includes a location management function (LMF) network element, and the second network element includes a terminal device; The training data includes location information of a second network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data, the first type of information includes complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the location information belonging to the non-reference data.

[0035] According to this implementation, in an uplink positioning scenario, the PRU / UE performs clustering on the collected training data (specifically, the labels of the output values ​​of the AI ​​model) and feeds back some complete information of the training data and some incremental information of the training data to the first network element. The LMF network element then restores the training data. This can reduce the air interface feedback overhead caused by model training / updating.

[0036] In some implementations of the first or second aspect, an AI model is applied to the CSI prediction; The first network element includes an access network device, and the second network element includes a terminal device; the training data includes CSI estimation results obtained by a second network element based on a channel state information reference signal (CSI-RS) coming from a first network element, the CSI estimation results including CSI estimation results belonging to reference data and CSI estimation results belonging to non-reference data; The first type of information includes complete information of CSI estimation results belonging to reference data, and the second type of information includes incremental information corresponding to CSI estimation results belonging to non-reference data.

[0037] When the technical solution provided in this application is applied to an AI-based CSI prediction scenario, the UE performs a clustering process on the provided training data to classify the training data into reference data and non-reference data, and reports the complete information of the reference data and the incremental information corresponding to the non-reference data, thereby reducing the overhead of reporting the training data.

[0038] In some implementations of the first or second aspect, the AI ​​model is applied to the channel state information (CSI) feedback, the first network element includes an access network device, and the second network element includes a terminal device; The training data includes a CSI estimation result obtained by the second network element based on a channel state information reference signal CSI-RS coming from the first network element, or the training data includes a CSI estimation result obtained by the second network element based on the CSI-RS and quantization compression information corresponding to the CSI estimation result.

[0039] When the technical solution provided in this application is applied to an AI-based CSI feedback scenario, the UE performs a clustering process on the provided training data to classify the training data into reference data and non-reference data, and reports the complete information of the reference data and the incremental information corresponding to the non-reference data, thereby reducing the overhead of reporting the training data.

[0040] In some implementations of the first or second aspect, the training data includes CSI estimation results, the CSI estimation results include CSI estimation results belonging to reference data and CSI estimation results belonging to non-reference data, the first type of information includes complete information of the CSI estimation results belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation results belonging to the non-reference data; or The training data includes a CSI estimation result and quantized compressed information of the CSI estimation result, the CSI estimation result includes a CSI estimation result belonging to reference data and a CSI estimation result belonging to non-reference data, the first type of information includes complete information of the CSI estimation result belonging to the reference data and complete information of the quantized compressed information of the CSI estimation result belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation result belonging to the non-reference data and incremental information corresponding to the quantized compressed information of the CSI estimation result belonging to the non-reference data.

[0041] According to this embodiment, there are two cases in which the UE performs clustering processing on the training data, depending on whether the UE-side model is trained and distributed by the base station. If the UE-side model is trained and distributed by the base station, the UE performs clustering on the channel measurement results obtained by measuring CSI-RS to classify the training data into reference data and non-reference data, and then reports the information obtained by clustering (specifically, the complete information of the reference data and the incremental information corresponding to the non-reference data). This can reduce the overhead of reporting the training data. If the UE-side model does not depend on the training and distribution by the base station, the UE separately performs clustering on the channel measurement results obtained by measuring CSI-RS and the corresponding quantization compression information, and then reports the information obtained by clustering. This reduces the overhead of reporting the training data.

[0042] According to a third aspect, the present application provides a communications device. In one design, the communications device may include modules configured to perform the methods / acts / steps / operations described in the first aspect. The modules may be hardware circuits, software, or hardware circuits in combination with software. In one design, the communications device may include a processing module and a communications module. In one example, the communications device is a positioning device or an access network device, and the positioning device may be, for example, an LMF network element.

[0043] According to a fourth aspect, the present application provides a communications device. In one design, the communications device may include modules configured to perform the methods / acts / steps / operations described in the second aspect. The modules may be hardware circuits, software, or hardware circuits in combination with software. In one design, the communications device may include a processing module and a communications module. In one example, the communications device is a terminal device, such as a PRU or a common terminal.

[0044] Optionally, the PRU may be considered a special network element and may typically be configured by a network vendor. For example, the network vendor may configure one or more of the PRU's location, transmission capabilities, reception capabilities, processing capabilities, etc. The PRU may provide its location information to an access network device and may also be referred to as a location reference device. For example, the PRU may be a reference UE or an automated guided vehicle (AGV). It should be understood that the common UE in this embodiment of the present application relates to the PRU. The common UE may obtain its location information using some positioning methods and then provide the location information to the positioning device.

[0045] According to a fifth aspect, the present application provides a communications device. The communications device includes a processor configured to perform a method according to the first aspect or any one of the embodiments of the first aspect. The processor is coupled to a memory. The memory is configured to store instructions and data. When the processor executes the instructions stored in the memory, the method according to the first aspect or any one of the embodiments of the first aspect can be performed. Optionally, the communications device may further include a memory. Optionally, the communications device may further include a communications interface. The communications interface is used by the device to communicate with another device. For example, the communications interface may be a transceiver, a hardware circuit, a bus, a module, a pin, or another type of communications interface. In one example, the communications device may be a positioning device or an access network device, or may be a device, module, chip, etc. located within the positioning device / access network device, or a device usable in cooperation with the positioning device / access network device.

[0046] According to a sixth aspect, the present application provides a communications device. The communications device includes a processor configured to perform a method according to the second aspect or any one of the embodiments of the second aspect. The processor is coupled to a memory. The memory is configured to store instructions and data. When the processor executes the instructions stored in the memory, the method according to the second aspect or any one of the embodiments of the second aspect can be performed. Optionally, the communications device may further include a memory. Optionally, the communications device may further include a communications interface. The communications interface is used by the device to communicate with another device. For example, the communications interface may be a transceiver, a hardware circuit, a bus, a module, a pin, or another type of communications interface. In one example, the communications device may be a terminal device, a device, a module, a chip, etc. located in the terminal device, or a device that can be used in cooperation with the terminal device.

[0047] According to a seventh aspect, the present application provides a communication system including a first network element and a second network element. For example, an interaction between the first network element and the second network element is as follows:

[0048] the first network element receives first information from the second network element, the first information including first type information and second type information, the first type information and the second type information indicating training data for the AI ​​model provided by the second network element; The training data includes one or more reference data and basic data, the first type of information includes complete information of the one or more reference data, and the second type of information includes incremental information of the basic data relative to the reference data corresponding to the basic data.

[0049] Specifically, the first network element may be understood with reference to an implementation of the first aspect, and the second network element may be understood with reference to an implementation of the second aspect. Details will not be repeated here. For example, the communication system may include a terminal device and an access network device in an AI-based CSI prediction or CSI feedback application scenario. Optionally, the communication system may include a terminal device, an access network device, and a positioning device in an AI-based positioning scenario.

[0050] According to an eighth aspect, the present application provides a communication system including a communication device according to the third or fifth aspect and a communication device according to the fourth or sixth aspect.

[0051] According to a ninth aspect, the present application further provides a computer program which, when run on a computer, enables the computer to perform the method provided in the first aspect, the second aspect, or any one of the embodiments of the first or second aspect.

[0052] According to a tenth aspect, the present application further provides a computer program product comprising instructions that, when executed on a computer, enable the computer to perform a method according to the first aspect, the second aspect, or any one of the implementations of the first or second aspect.

[0053] According to an eleventh aspect, the present application further provides a computer-readable storage medium, the computer-readable storage medium storing a computer program or instructions, which, when executed on a computer, enables the computer to perform a method according to the first aspect, the second aspect, or any one of the embodiments of the first or second aspect.

[0054] According to a twelfth aspect, the present application further provides a chip, the chip being configured to read a computer program stored in a memory to perform a method according to the first aspect, the second aspect, or any one of the embodiments of the first or second aspect, or The chip includes circuitry configured to perform the method according to the first aspect, the second aspect, or any one of the implementations of the first or second aspect.

[0055] According to a thirteenth aspect, the present application further provides a chip system. The chip system includes a processor configured to support an apparatus in performing a method according to the first aspect, the second aspect, or any one of the embodiments of the first or second aspects. In one possible design, the chip system further includes a memory. The memory is configured to store programs and data required by the apparatus. The chip system may include a chip, or may include a chip and another discrete device.

[0056] For the technical effects of the solutions provided in the second to thirteenth aspects or any one of the implementations of the second to thirteenth aspects, please refer to the corresponding description of the first aspect, and the details will not be described again. [Brief explanation of the drawings]

[0057] [Figure 1] 1 is a diagram of the architecture of a communication system to which an embodiment of the present application is applicable; [Figure 2] 2 is a schematic flowchart of a method 200 for transmitting training data for an AI model according to the present application. [Figure 3] A diagram of downlink positioning based on an AI model. [Figure 4] FIG. 1 illustrates an example of applying the technical solution provided in this application to AI-based downlink positioning. [Figure 5] A diagram of uplink positioning based on an AI model. [Figure 6] FIG. 1 illustrates an example of applying the technical solution provided in this application to AI-based uplink positioning. [Figure 7] 1 is a diagram of CSI prediction based on an AI model. [Figure 8] FIG. 1 illustrates an example of applying the technical solution provided in the present application to AI-based CSI prediction. [Figure 9] 1 is a diagram of an AI-based CSI feedback process. [Figure 10] FIG. 1 illustrates an example of applying the technical solution provided in the present application to AI-based CSI feedback. [Figure 11] 1 is a diagram of a communication device 1000 according to the present application. [Figure 12] 11 is another diagram of a communication device 1100 according to the present application. DETAILED DESCRIPTION OF THE INVENTION

[0058] The following describes the technical solutions of the present application with reference to the accompanying drawings.

[0059] The technical solutions provided in this application can be applied to various communication systems, such as long term evolution (LTE) systems, 5th generation (5G) communication systems, worldwide interoperability for microwave access (WiMAX), wireless local area network (WLAN) systems, satellite communication systems, future communication systems such as 6G communication systems, and systems that integrate multiple systems. 5G communication systems may also be referred to as new radio (NR) systems.

[0060] For example, a communication system may include a terminal device and a network device.

[0061] In the embodiment of the present application, the terminal device may be an entity located at a user's side and configured to receive or transmit signals, such as a mobile phone. The terminal device may include a handheld device with wireless connectivity, other processing devices connected to a wireless modem, or an in-vehicle device. The terminal device may be a portable, pocket-sized, handheld, computer-integrated, or in-vehicle mobile device. The terminal device can be widely applied to various scenarios, such as cellular communication, wireless fidelity (Wi-Fi) systems, device-to-device (D2D), vehicle-to-everything (V2X), peer-to-peer (P2P), machine-to-machine (M2M), machine-type communication (MTC), Internet of things (IoT), virtual reality (VR), augmented reality (AR), industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wear, intelligent transportation, smart city, unmanned aerial vehicle, robot, remote sensing, passive sensing, positioning, navigation and tracking, and autonomous delivery and mobility. Some examples of the terminal device 120 are as follows:The following devices are considered to be ubiquitous: user equipment (UE) of 3GPP (registered trademark) standards, stations (STA) of Wi-Fi systems, fixed devices, mobile devices, handheld devices, wearable devices, mobile phones, smartphones, session initiation protocol (SIP) phones, notebook computers, personal computers, smartbooks, vehicles, satellites, global positioning system (GPS) devices, target tracking devices, unmanned aerial vehicles, helicopters, aircraft, ships, remote control devices, smart home devices, industrial devices, personal communication service (PCS) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), wireless network cameras, tablet computers, palmtop computers, mobile internet devices (MIDs), wearable devices such as smart watches, virtual reality (VR) devices, augmented reality (AR) devices, industrial controls, etc. wireless terminals in smart city systems, terminals in vehicle internet systems, wireless terminals in self-driving, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities such as smart fuel dispensers and terminal devices on high-speed trains, and wireless terminals in smart homes such as smart speakers, smart coffee machines, and smart printers. The terminal devices may be wireless devices in the various scenarios mentioned above, or devices located within wireless devices, for example, communication modules, modems, or chips within the devices mentioned above.The terminal device may be referred to as a terminal, a UE, a mobile station (MS), a mobile terminal (MT), etc. Alternatively, the terminal device may be a terminal device in a future wireless communication system. The terminal device may also include a position reference device, such as an automated guided vehicle (AGV) or a device with a similar function. For ease of explanation, the following uses an example in which the terminal device is a UE for explanation.

[0062] In the present application, a communication device configured to perform the functions of a terminal device may be a terminal device, a terminal device having some functions of the aforementioned communication device, or a device capable of supporting the performance of the functions of a terminal device, such as a chip system. The device may be installed in the terminal device or used in cooperation with the terminal device. In the present application, a chip system may include a chip, or may include a chip and another separate device.

[0063] A network device may be a device that provides wireless communication services and communicates with terminal devices, and is usually located on the network side. A network device may also be called an access network device or a radio access network device. For example, a network device may be a base station. For example, the access network device in the embodiments of the present application includes, but is not limited to, a next generation NodeB (gNodeB, gNB) in a 5G communication system, a base station in a sixth generation (6G) mobile communication system, a base station in a future mobile communication system, an access node (AP) in a Wi-Fi system, an evolved NodeB (eNB) in an LTE system, a radio network controller (RNC), a NodeB (NodeB, NB), a base station controller (BSC), a home NodeB (e.g., home evolved NodeB or home NodeB, HNB), a baseband unit (BBU), a transmission reception point (TRP), a transmission point (TP), a base transceiver station (BTS), a satellite, and an unmanned aerial vehicle. In the network structure, the network device may include a central unit (CU) node, a distributed unit (DU) node, a RAN device including a CU node and a DU node, a RAN device including a CP control plane node, a CU user plane node, and a DU node, or a radio controller, relay station, in-vehicle device, wearable device, etc. in a cloud radio access network (CRAN) scenario.Additionally, the base station may be a macro base station, a micro base station, a relay node, a donor node, or a combination thereof. Alternatively, the base station may be a communication module, modem, or chip disposed in the aforementioned device or apparatus. Alternatively, the base station may be a mobile switching center, a device performing base station functions in device-to-device (D2D), vehicle-to-everything (V2X), or machine-to-machine (M2M) communications, a network-side device in a 6G network, a device performing base station functions in a future communication system, etc. The base station may support networks using the same or different access technologies. This is not limiting. The base station may be stationary or mobile. For example, a helicopter or unmanned aerial vehicle may be configured as a mobile base station, and one or more cells may move based on the location of the mobile base station. In another example, the helicopter or unmanned aerial vehicle may be configured as a device for communicating with another base station.

[0064] In this application, an apparatus configured to implement the functions of a network device may be an access network device, a network device having some functions of an access network, or an apparatus capable of supporting the implementation of the functions of an access network, such as a chip system, a hardware circuit, a software module, or a combination of a hardware circuit and a software module. The apparatus may be installed in an access network device or used in cooperation with an access network device. In the method of this application, an example in which a communication apparatus configured to implement the functions of an access network device is an access network device is used for explanation.

[0065] Optionally, the communication system further comprises at least one AI node.

[0066] Optionally, the AI ​​node may be deployed in one or more of the following locations within a communications system: an access network device, a terminal device, a core network device, etc. Alternatively, the AI ​​node may be independently deployed, e.g., in a location other than one of the above devices, e.g., in a host or cloud server within an over-the-top (OTT) system. The AI ​​node may communicate with other devices within the communications system, e.g., one or more of the following: a network device, a terminal device, a network element of a core network, etc.

[0067] Optionally, the AI ​​node is configured to perform AI-related operations, which may include, for example, one or more of the following: model fault testing, model performance testing, model training, data collection, etc.

[0068] For example, the network device may forward data related to the AI ​​model and reported by the terminal device to the AI ​​node, and the AI ​​node performs the AI-related operation. In another example, the network device or the terminal device may forward data related to the AI ​​model to the AI ​​node, and the AI ​​node performs the AI-related operation. In another example, the AI ​​node may transmit outputs of the AI-related operation, such as one or more of the following: a trained neural network model, model evaluation results, model testing results, etc., to the network device and / or the terminal device. For example, the AI ​​node may transmit outputs of the AI-related operation directly to the network device and the terminal device. In another example, the AI ​​node may transmit outputs of the AI-related operation to the terminal device via the network device. In another example, the AI ​​node may transmit outputs of the AI-related operation to the network device via the terminal device.

[0069] It can be understood that the number of AI nodes is not limited in the present application. For example, if there are multiple AI nodes, the multiple AI nodes may be classified based on their functions. For example, different AI nodes are responsible for different functions.

[0070] It may also be understood that an AI node may be an independent device, may be integrated into the same device to implement different functions, may be a network element within a hardware device, may be a software function running on dedicated hardware, or may be a virtualized function instantiated on a platform (e.g., a cloud platform). The specific form of the AI ​​node is not limited by this application.

[0071] For ease of understanding, the following provides a brief description of related art or concepts in this application.

[0072] An AI model is an algorithm or computer program that can realize an AI function. An AI model represents a mapping relationship between the input and output of the model, or is a function model that maps a specific dimensional input to a specific dimensional output. Model parameters of the AI ​​model are obtained through machine learning training. For example, f(x)=ax²+b is a quadratic function model and may be considered as an AI model, where a and b are parameters of the AI ​​model and may be obtained through machine learning training. For example, the AI ​​models referred to in the following embodiments of the present application are not limited to neural networks, linear regression models, decision tree models, support vector machines (SVMs), Bayesian networks, Q-learning models, or other machine learning (ML) models.

[0073] A training dataset is data used for model training, validation, and testing in machine learning. The quantity and quality of data affect the effectiveness of machine learning. The training data may include inputs for an AI model, or may include inputs and target outputs for an AI model. The target outputs are target values ​​output by an AI model and may also be called output ground truth, output ground truth, labels, or label samples.

[0074] Model training is the process of selecting an appropriate loss function and training the model parameters by using an optimization algorithm so that the value of the loss function is below a threshold or meets a target requirement.

[0075] AI model design mainly includes a data collection phase (e.g., collecting training data and / or inference data), a model training phase, and a model inference phase, and may further include an inference result application phase. In the data collection phase, a data source is used to provide a training dataset and inference data. In the model training phase, the training data provided by the data source is analyzed or trained to obtain an AI model. The AI ​​model represents a mapping relationship between the model's input and output. Obtaining an AI model by learning using a model training node is equivalent to obtaining a mapping relationship between the model's input and output by learning using training data. In the model inference phase, the AI ​​model obtained by training in the model training phase is used to perform inference based on the inference data provided by the data source to obtain an inference result. This phase can also be understood as inputting inference data into the AI ​​model to obtain an output using the AI ​​model. The output is the inference result. The inference result may indicate the configuration parameters used (executed) by the actor object and / or the operation performed by the actor object. The inference result is published in the inference result application phase. For example, the inference result may be planned by the same actor entity. For example, the actor entity may send inference results to one or more actor objects (e.g., a core network device, an access network device, or a terminal device) for execution. As another example, the actor entity may further feed back model performance to a data source to facilitate subsequent model updates and training.

[0076] The loss function is used to measure the difference between the model's predicted value and the true value.

[0077] Model application is the use of a trained model to solve a real-world problem.

[0078] It will be appreciated that the AI ​​model may be implemented using hardware circuitry, software, or a combination of software and hardware, without limitation. Non-limiting examples of software include program code, programs, subprograms, instructions, instruction sets, code, code segments, software modules, applications, software applications, etc.

[0079] In the embodiments of the present application, an "indication" may include a direct indication, an indirect indication, an explicit indication, and an implicit indication. The indication information indicating A may be understood as follows: That is, the indication information carries A and may indicate A directly or indirectly. An indirect indication may mean that the indication information directly indicates B and the correspondence between B and A to indicate A using the indication information. The correspondence between B and A may be predefined in a protocol, pre-stored, or obtained using a configuration between network elements.

[0080] The present application provides a method for transmitting training data by a reporting network element of the training data to a training network element, which helps reduce the overhead of transmitting training data by a reporting network element to a training network element.

[0081] A wireless communication system to which the present application is applicable may include one or more network side devices and one or more terminal devices. The network side devices may include the aforementioned access network devices and may optionally further include core network devices. This is not limited thereto. The term "device" may also be replaced with a network element, entity, network entity, communication device, communication module, node, communication node, etc. In this application, a device or a network element is used as an example for explanation.

[0082] FIG. 1 is a diagram of a communication system architecture to which an embodiment of the present application can be applied. (a) of FIG. 1 is a diagram of a communication system architecture to which an embodiment of the present application can be applied. For example, the communication system includes a network device 110, a terminal device 120, and a terminal device 130. The terminal devices 120 and 130 may access and communicate with the network device 110. Optionally, the network device 110 may be an access network device. In one implementation, the communication system may further include an AI entity. The network device 110 may forward data related to an AI model and reported by the terminal device to the AI ​​entity. The AI ​​entity performs AI-related operations, such as training dataset construction and model training, and provides outputs of the AI-related operations, such as a trained AI model, model evaluation results, and model testing results, to the network device 110. In another implementation, the AI ​​entity may alternatively be located within the network device 110, i.e., in a module of the network device 110. (b) of FIG. 1 is another diagram of a communication system architecture to which an embodiment of the present application can be applied. The communication system includes a network device 110, a terminal device 120, a terminal device 130, and a positioning device 140. The positioning device 140 and the network device 110 may communicate with each other using interface messages. For example, the positioning device 140 may be a location management function (LMF), and the network device 110 may be an access network device, such as a gNB or an eNB. This is not limiting. For example, if the access network device 110 is a gNB, the gNB and the LMF may exchange information using NR positioning protocol A (NRPPa) messages. If the access network device 110 is an eNB, the eNB and the LMF may exchange information using LTE positioning protocol (LPP) messages.Optionally, the terminal device may directly communicate with the positioning device 140, for example, the interaction between the terminal device 130 and the positioning device 140 shown in FIG. 1(b). In FIG. 1, the AI ​​entity may be configured inside the positioning device 140 or may be located separately from the positioning device 140. This is not a limitation.

[0083] It should be understood that FIG. 1 is merely a diagram. Devices included in the wireless communication system vary according to different application scenarios. For example, (a) in FIG. 1 is applicable to application scenarios such as AI-based CSI feedback and AI-based CSI prediction. (b) in FIG. 1 is applicable to an AI-based positioning scenario. In addition, the communication system may further include other devices, such as core network devices, wireless relay devices, wireless backhaul devices, and / or devices configured to implement artificial intelligence functions. The other devices are not shown in FIG. 1.

[0084] In practical applications, a wireless communication system may include multiple network devices (e.g., access network devices and core network devices) or multiple terminal devices. This is not limited. One access network device may provide services to one or more terminal devices. One terminal device may also access one or more access network devices. The number of terminal devices and network devices included in a wireless communication system is not limited in the embodiments of the present application.

[0085] 2 is a schematic flowchart of a method 200 for transmitting training data for an AI model according to the present application. The method 200 includes a first network element and a second network element. The first network element may be a training network element for the AI ​​model, and the second network element is a providing network element or a reporting network element for the training data for the AI ​​model. Optionally, the first network element and the second network element may be logically deployed separately. In different implementations, the first network element and the second network element may be physically deployed on the same network element or different network elements. This is not a limitation.

[0086] 210, the first network element receives first information from the second network element, the first information including first type information and second type information, the first type information and the second type information indicating training data for the AI ​​model provided by the second network element.

[0087] The training data includes one or more reference data and non-reference data, wherein the first type of information includes complete information of the one or more reference data, and the second type of information includes incremental information of the non-reference data relative to the reference data corresponding to the non-reference data.

[0088] In one embodiment, one or more reference data are determined from training data based on a first clustering algorithm.In this embodiment, reference data and non-reference data are determined using a first clustering algorithm.Specifically, by clustering training data using a first clustering algorithm, training data is classified into reference data and non-reference data.Reference data can also be understood as the center of a cluster, and non-reference data are the data points in a cluster other than the center.Depending on specific training data and specific clustering algorithm, there may be one or more reference data points.

[0089] It should be understood that clustering belongs to unsupervised learning in machine learning, and is to divide a data set into different classes or clusters according to a certain standard (for example, distance), so that data in the same cluster are as similar as possible to each other, and data that are not in the same cluster are as different as possible to each other. That is, after clustering, data of the same type are aggregated as much as possible, and data of different types are separated as much as possible.

[0090] In the embodiment of the present application, the specific clustering algorithm is not limited. For example, the first clustering algorithm may include, but is not limited to, the following clustering algorithms:

[0091] Partition-based clustering, such as K-means, determines cluster centers and calculates the distance between each data point and the cluster center to perform class assignment with fast convergence.

[0092] Density-based clustering, e.g., density-based spatial clustering of applications with noise (DBSCAN), starts from the core point to the density-reachable region to obtain the largest region containing the core point and the boundary points. The number of clusters does not need to be specified, the shape of the clusters is arbitrary, and the clusters are less susceptible to noise.

[0093] Hierarchical clustering, such as hierarchical density-based spatial clustering of applications with noise (HDBSCAN), uses reachability distance as the neighbor edge weight to form a minimum spanning tree of all nodes, and then performs hierarchical clustering and cluster compression. Hierarchical clustering is characterized by high speed, variable cluster density, and low parameter sensitivity; only a minimum number of clusters needs to be defined.

[0094] Graph-based clustering, such as spectral clustering. K-means clustering is performed after the data space is transformed, and the transformed space is restored to the original space. The idea of ​​dimensionality reduction is introduced to improve the clustering effect for high-dimensional spaces. As another example, in Chinese Whispers, only the similarity threshold needs to be specified, and the number of clusters, etc., is not set.

[0095] Network-based clustering, such as STING, divides the data space into rectangular cells and performs clustering on the rectangular cells. The clustering speed is related to the number of units obtained by the division, and the operation speed is fast. As another example, CLIQUE divides the data space into non-intersecting rectangular cells, and data density units are connected to each other to form clusters. It has high performance and can find data units where high-density data sample objects are located. As yet another example, WAVE CLUSTER divides the data space into a frequency domain space with a grid structure and performs convolution with a kernel function to obtain natural clustering properties of the data. Clustering can be performed at multiple resolutions, has good noise resistance, and is applicable to large data sets.

[0096] Fuzzy clustering, such as fuzzy C-means (FCM), has a desirable clustering effect on data that meets a normal distribution.

[0097] Alternatively, the first clustering algorithm may be the following clustering algorithm derived from the clustering algorithm described above: These include deep clustering based on K-means clustering, deep clustering based on Gaussian mixture models, deep clustering based on subspace clustering, deep clustering based on spectral clustering, deep clustering based on mutual information, deep clustering based on autoencoders, deep clustering based on deep neural networks, deep clustering based on variational autoencoders, deep clustering based on generative adversarial networks, and DBSCAN with noise.

[0098] In another embodiment, the one or more reference data are pre-negotiated, agreed upon, or configured by the first network element and the second network element. For example, a differential scheme applied to CSI prediction is used. For example, the first network element and the second network element pre-agree that the reference data is reported once at an interval of one time period, and non-reference data is reported between two times when the reference data is reported. After the reference data is reported once, non-reference data corresponding to the reference data is reported after the reference data. Specifically, the second network element uses the first type of information to indicate the complete information of the reference data and the second type of information to indicate incremental information of the non-reference data relative to the reference data corresponding to the non-reference data. The incremental information of the non-reference data relative to the reference data is determined by performing a differential operation between the non-reference data and the reference data corresponding to the non-reference data.

[0099] Optionally, step 220 is further included before step 210. For example, the first clustering algorithm may be predefined, pre-stored, or pre-configured in the protocol. In these cases, the first clustering algorithm is known to the first network element and the second network element, and therefore the first network element does not need to indicate the first clustering algorithm to the second network element. Otherwise, the first network element needs to indicate the first clustering algorithm to the second network element.

[0100] 220: The first network element sends first indication information to the second network element, where the first indication information is used to determine a first clustering algorithm.

[0101] In one example, the first instruction information explicitly indicates the first clustering algorithm, for example, the first instruction information indicates the K-means clustering algorithm.

[0102] In another example, the first network element pre-configures a correspondence between multiple indexes and multiple clustering algorithms, and the first instruction information indicates one index. For example, the correspondence between multiple indexes and multiple clustering algorithms is listed in Table 1. The first instruction information indicates indexes 0, 1, and 2 to the second network element, indicating the clustering algorithms corresponding to these indexes, respectively. In this example, pre-configuration is used merely as an example. Alternatively, the correspondence may be pre-defined in a protocol, pre-stored, or the like. This is not a limitation.

[0103] [Table 1]

[0104] Optionally, in one example, the first network element pre-configures a correspondence between a plurality of indexes and a plurality of AI model-based application scenarios. The clustering algorithm used to report the training data in each application scenario may be pre-defined, pre-configured, pre-stored, etc. in the protocol. Table 2 is used as an example. The first instruction information indicates one index. Based on the correspondence between the plurality of indexes and the plurality of application scenarios, the second network element may determine an application scenario corresponding to the index indicated by the first instruction information and determine a clustering algorithm for the application scenario.

[0105] [Table 2]

[0106] Optionally, one application scenario may correspond to multiple clustering algorithms, and the indexes of the clustering algorithms corresponding to different application scenarios may be different from each other. The clustering algorithms corresponding to different indexes may be pre-configured, pre-stored, or pre-defined in the protocol, as shown in Table 3. Alternatively, the indexes of the clustering algorithms corresponding to different application scenarios may be the same, as shown in Table 4. In the example of Table 4, the first indication information may indicate the application scenario and an index value. For example, different identifiers may be pre-configured for the application scenarios. For example, CSI feedback, uplink positioning, downlink positioning, and CSI prediction correspond to identifiers A, B, C, and D, respectively. The first indication information indicating A0 indicates the clustering algorithm corresponding to CSI feedback, and the first indication information indicating B0 indicates the clustering algorithm corresponding to index 0 in the uplink positioning scenario, and so on.

[0107] [Table 3]

[0108] [Table 4]

[0109] The second network element determines a first clustering algorithm based on the first instruction information, and performs clustering on the training data using the first clustering algorithm.

[0110] Optionally, the first instruction information further indicates to the second network element to collect training data for the AI ​​model. Since the first instruction information indicates the first clustering algorithm and implicitly indicates to the second network element to collect training data for the AI ​​model, instruction overhead may be reduced.

[0111] Alternatively, the first network element may use a separate message to indicate to the second network element to collect training data for the AI ​​model, in which case step 230 is further included before step 210.

[0112] 230: The first network element sends second instruction information to the second network element, where the second instruction information indicates to the second network element to collect training data.

[0113] Additionally, optionally, the method 200 further comprises step 240 .

[0114] 240: The first network element trains or updates the AI ​​model based on the training data.

[0115] After the first network element obtains the training data reported by the second network element, the training data may be used to train, update, etc. the AI ​​model. This is not specifically limited.

[0116] According to the method for transmitting training data provided in the present application, the transmitting network element of the training data no longer directly transmits the training data, but performs a clustering process on the training data using a clustering algorithm to classify the training data into reference data and non-reference data. Then, by transmitting the complete information of the reference data and the incremental information of the non-reference data relative to the reference data corresponding to the non-reference data to the first network element, all the training data are represented to the first network element, so that the overhead of transmitting the training data can be reduced.

[0117] The application of the technical solutions of the present application will be described below with reference to several application scenarios.

[0118] In some examples, AI model application scenarios include, but are not limited to, the following scenarios: AI-based (AI-based) downlink positioning, AI-based uplink positioning, AI-based CSI prediction, or AI-based CSI feedback.

[0119] The following describes the application of the method for transmitting training data for an AI model provided in this application to these application scenarios individually.

[0120] 1. AI-based downlink positioning FIG. 3 is a diagram of downlink positioning based on an AI model. As shown in FIG. 3, in downlink positioning, AI model training is performed by a positioning device on the network side, e.g., a location management function (LMF) network element. The AI ​​model deployed in the positioning device uses channel measurement results, such as a channel impulse response (CIR), acquired by a positioning reference unit (PRU) or a common UE by measuring a positioning reference signal transmitted by an access network device as input, and the location of the PRU or UE as output. Optionally, the positioning reference signal may come from one or more access network devices. This is not a limitation. When the positioning device performs AI model training, the positioning device needs to acquire the channel measurement results acquired by the PRU / UE from the PRU / UE side by performing channel estimation based on the PRS transmitted by the access network device and received by the PRU / UE, and the location information of the PRU / common UE to perform training.

[0121] Optionally, the channel measurement result includes, but is not limited to, a CIR, a channel time-domain quality indicator calculated based on the CIR, a channel frequency response (CFR), a channel frequency-domain quality indicator calculated based on the CFR, a power delay profile (PDP), a channel quality indicator calculated based on the PDP, etc. The CIR may be a normalized CIR, and the CFR may be a normalized CFR. This is not limited. When the channel time-domain quality indicator is calculated based on the CIR, the number of sampling points is not limited. When the channel frequency-domain quality indicator is calculated based on the CFR, the bandwidth, subbands, and number of ports corresponding to the CFR are not limited. Alternatively, the channel measurement result may be expressed using other quality indicators not listed. The description of the channel measurement result is applicable to all the following embodiments and will not be repeated below.

[0122] Figure 4 illustrates an example of applying the technical solution provided in this application to AI-based downlink positioning. In the downlink positioning example, the first network element is a positioning device, and the second network element is a PRU or UE. The access network device may be referred to as the third network element.

[0123] 401: Optionally, the positioning device determines that the AI ​​model needs to be trained or updated.

[0124] 402: The positioning device sends instruction information a to the access network device, where the instruction information a indicates the access network device to send the PRS.

[0125] 403: The access network device sends the PRS to the PRU / UE.

[0126] The PRU / UE receives the PRS coming from the access network device and obtains channel measurements.

[0127] 404: Optionally, the positioning device sends first indication information to the UE, where the first indication information is used by the UE to determine a first clustering algorithm.

[0128] Optionally, the first instruction information further implicitly indicates to the UE to collect training data for the AI ​​model.

[0129] Optionally, in another implementation, the positioning device may use separate information to indicate to the UE to collect training data for the AI ​​model. For example, the positioning device sends instruction information b to the UE, and the instruction information b indicates to the UE to collect training data for the AI ​​model.

[0130] It should be understood that the order of steps 402 and 404 is not limited.

[0131] 405: The PRU / UE performs a clustering process on the channel measurement results based on a first clustering algorithm to obtain reference data and non-reference data. Optionally, the PRU / UE performs a clustering process on the positions of the PRU / UE corresponding to the channel measurement results to obtain reference data and non-reference data.

[0132] Optionally, in the clustering process, if there is a channel measurement that is not in any cluster, the channel measurement is removed.

[0133] Optionally, the PRU / UE may further filter the channel measurement results (which may be understood as candidate channel measurement results) obtained by measuring the PRS, and then perform a clustering process on the channel measurement results obtained by filtering, without limitation.

[0134] Optionally, during the clustering process, there are one or more reference data, depending on the particular clustering algorithm and the particular training data.

[0135] In addition, for channel measurement results obtained by a PRU / UE by measuring a PRS, each channel measurement result may correspond to one location information, or only some of the channel measurement results may correspond to location information. This is not limited thereto. For example, a PRU / UE may obtain 10 channel measurement results by measuring a PRS, and each of the 10 channel measurement results may correspond to one location information. During clustering, the PRU / UE performs clustering on the 10 channel measurement results and on the 10 location information corresponding to the 10 channel measurement results to separately obtain reference data and non-reference data related to the channel measurement results and reference data and non-reference data related to the location information. In another example, a PRU / UE obtains 10 channel measurement results by measuring a PRS. The PRU / UE reports location information corresponding to six of the 10 channel measurement results and does not report location information corresponding to the remaining four channel measurement results. During clustering, the PRU / UE performs clustering on 10 channel measurement results and on six pieces of location information corresponding to six channel measurement results to separately obtain reference data and non-reference data for the channel measurement results and reference data and non-reference data for the location information. In yet another example, the PRU / UE reports only the channel measurement results and does not report the location information corresponding to the channel measurement results. In this example, during clustering, the PRU / UE performs clustering on 10 channel measurement results to obtain reference data and non-reference data.

[0136] 406: The PRU / UE sends first information to the positioning device, where the first information includes first type information and second type information, specifically, the first type information includes complete information of one or more reference data, and the second type information includes incremental information of the non-reference data relative to the reference data corresponding to the non-reference data.

[0137] It should be understood that if the training data includes channel measurement results, the reference data and non-reference data in step 406 are respectively reference data and non-reference data obtained by performing a clustering process on the channel measurement results. If the training data includes channel measurement results and location information corresponding to the channel measurement results, the reference data and non-reference data in step 406 include reference data and non-reference data obtained by performing a clustering process on the channel measurement results, and reference data and non-reference data obtained by performing a clustering process on the location information.

[0138] It should be understood that non-reference data is relative to reference data. For example, suppose there are 10 pieces of training data. After clustering process is performed, the 10 training data are classified into two clusters, and each of the two clusters has one reference data and some non-reference data. When showing the 10 pieces of training data, the first type of information shows the complete information of all reference data, for example, the complete information of the reference data of each of the two clusters, and the second type of information shows the incremental information of each non-reference data relative to the reference data corresponding to the non-reference data.

[0139] As an example, two-dimensional training data is used. One reference data is determined by clustering, and the reference data is (25,25) and the non-reference data is (26,26). In this case, the incremental information of the non-reference data relative to the reference data is (1,1).

[0140] The positioning device receives first information from the PRU / UE.

[0141] 407: The positioning device recovers one or more reference data and non-reference data based on the first clustering algorithm and the first information, i.e. recovers all training data.

[0142] 408: Optionally, the positioning device trains or updates an AI model based on the restored training data.

[0143] It should be understood that in AI-based downlink positioning, the training data recovered by the positioning device includes channel measurement results. Optionally, the channel measurement results come from one or more second network elements. For one second network element, the positioning reference signal measured by the second network element comes from one or more third network elements. This is not limited thereto. Different positioning reference signals differ from each other in one or more of transmission time instants, transmission frequencies, occupied frequency bands, and transmission ports. For example, one second network element may obtain channel measurement results by measuring positioning reference signals coming from multiple third network elements, or one second network element may obtain channel measurement results by measuring multiple positioning reference signals coming from one third network element.

[0144] It can be seen that when the technical solution provided in this application is applied to a downlink positioning scenario, the PRU / UE performs a clustering process on the training data to classify the training data into reference data and non-reference data, and reports the complete information of all the reference data and the incremental information of each non-reference data relative to the reference data corresponding to the non-reference data, thereby reducing the overhead of reporting the training data.

[0145] 2. AI-based uplink positioning Machine learning can be used to improve positioning accuracy. Some channel measurement results (CIR, CFR, PDP, etc.) are used as inputs to an AI model to obtain the final location of the UE. The AI ​​model may be deployed on the LMF side, or the UE / base station may use the AI ​​model to extract characteristics of the channel measurement results. The LMF uses the extracted characteristics as inputs to the AI ​​model to obtain the UE's location.

[0146] As an example, uplink positioning is used. Specifically, the UE transmits a sounding reference signal (SRS) to the base station, and the base station obtains channel measurement results by measuring the SRS. Figure 5 illustrates AI-based uplink positioning. Figure 5(a) shows an AI model deployed on the LMF side. The base station transmits channel measurement results to the LMF, and the LMF uses the channel measurement results of multiple base stations as inputs to the AI ​​model and outputs UE location information. Assuming the base station has 16 antennas and 4096 subcarriers, each base station needs to transmit 16 × 4096 complex values ​​to the LMF. Figure 5(b) shows an AI model deployed on both the base station and LMF sides. Each base station uses its channel measurement results as input. The AI ​​model on the base station side extracts characteristics from the channel measurement results and transmits them to the LMF. The LMF uses the received channel characteristics as input and obtains the UE's location using the AI ​​model on the LMF side. The dimensionality of the channel characteristics is determined by the output dimensionality of the AI ​​model on the base station side. For example, the base station may extract a characteristic with a dimension of

[0128] from a channel measurement result with a dimension of [16, 4096] and transmit the characteristic to the LMF.

[0147] It should be understood that the characteristics of the channel measurements referred to in this embodiment of the present application may be the channel measurements themselves or may be channel characteristics extracted from the channel measurements by an AI model.

[0148] In uplink positioning, training or updating of the AI ​​model is performed on the network side. A 5G system is used as an example. The AI ​​model may be deployed on a positioning device in the core network, such as an LMF network element. For uplink positioning, the input of the AI ​​model is one or more channel measurement results corresponding to one or more sounding reference signals, and the output of the AI ​​model is the location of the UE. There may be one or more transmitting ends, such as one or more UEs, of one or more sounding reference signals, and there may be one or more receiving ends, such as one or more access network devices.

[0149] When training the AI ​​model used for positioning, the positioning device acquires, from the access network side, multiple channel measurement results obtained by one access network device by measuring multiple sounding reference signals or by each of multiple access network devices by measuring one or more sounding reference signals, and / or location information of third network elements (PRUs / UEs). The multiple sounding reference signals may include multiple sounding reference signals from one third network element, or may include one or more sounding reference signals from each of multiple third network elements. The location information of the third network element at different times for transmitting the multiple sounding reference signals, or the location information of each of the multiple third network elements at one or more times for transmitting the one or more sounding reference signals, is used as a ground label (i.e., label) for the location information output by the AI ​​model.

[0150] For uplink positioning, AI model training is performed by a positioning device on the network side, e.g., an LMF network element. For the AI ​​model deployed in the positioning device, the channel measurement results obtained by the access network device by measuring the sounding reference signal transmitted by the PRU / UE are used as input, and the location of the PRU / UE is used as output. During AI model training, the positioning device needs to obtain the PRU / UE's location information as the AI ​​model label from the PRU / UE side and the channel measurement results obtained by the access network device by measuring the SRS transmitted by the PRU / UE from the access network device side to perform training.

[0151] Figure 6 shows an example of applying the technical solution provided in the present application to AI-based uplink positioning. This example includes the following three possible cases: In one possible case, the first network element is a positioning device, the second network element is an access network device, and the third network element is a PRU / UE. The training data reported by the second network element and received by the positioning device is specifically a channel measurement result obtained by the second network element by measuring the SRS transmitted by the PRU / UE. In another possible case, the first network element is a positioning device and the second network element is a PRU / UE. The training data reported by the second network element and received by the positioning device is specifically location information of the PRU / UE. In yet another possible case, the first network element is a positioning device, the second network element is an access network device, and the third network element is a PRU / UE. The positioning device receives training data reported by the second network element (specifically, channel measurement results obtained by the second network element by measuring the SRS transmitted by the PRU / UE) and training data reported by the third network element (specifically, location information of the PRU / UE).

[0152] 601: Optionally, the positioning device determines that the AI ​​model needs to be trained or updated.

[0153] 602: The positioning device sends instruction information c to the access network device, where the instruction information c indicates to the access network device to configure the PRU / UE to send the SRS.

[0154] 603: The access network device configures the UE to transmit the SRS.

[0155] 604: The PRU / UE transmits the SRS.

[0156] The access network device obtains channel measurements by measuring the SRS.

[0157] 605: Optionally, the positioning device sends first indication information to the access network device, where the first indication information is used to determine a first clustering algorithm.

[0158] The access network device receives first indication information from the positioning device.

[0159] Optionally, the first instruction information further indicates an access network device for collecting training data for the AI ​​model. Alternatively, separate information may indicate an access network device for collecting training data for the AI ​​model. This is not limiting.

[0160] 606: Optionally, the positioning device sends third indication information to the PRU / UE, where the third indication information is used to determine a second clustering algorithm.

[0161] The PRU / UE receives third indication information from the positioning device.

[0162] 607: The access network device performs a clustering process on the channel measurement results based on a first clustering algorithm to obtain reference data and non-reference data.

[0163] 608: The access network device sends first information to the positioning device, the first information including first type information and second type information, where the first type information and second type information included in the first information indicate training data provided by the access network device. Specifically, the training data provided by the access network device includes channel measurement results obtained by the access network device by measuring the SRS transmitted by the PRU / UE. The first type information included in the first information includes complete information of the channel measurement results belonging to the reference data, and the second type information included in the first information includes incremental information corresponding to the channel measurement results belonging to the non-reference data.

[0164] 609: The PRU / UE performs a clustering process on the location information of the PRU / UE based on a second clustering algorithm to obtain reference data and non-reference data.

[0165] 610: The PRU / UE transmits third information to the positioning device, where the third information includes first type information and second type information. The first type information and second type information included in the third information indicate training data provided by the PRU / UE. Specifically, the training data provided by the PRU / UE includes location information of the PRU / UE. The first type information included in the third information includes complete information of location information belonging to the reference data, and the second type information included in the third information includes incremental information corresponding to location information belonging to the non-reference data.

[0166] 611: The positioning device restores all training data provided by the access network device based on a first clustering algorithm and first information from the access network device, and / or the positioning device restores all training data provided by the PRU / UE based on a second clustering algorithm and third information from the PRU / UE.

[0167] Optionally, in this example, the access network device and the PRU / UE separately perform clustering operations on the training data provided by the access network device and the PRU / UE by using different clustering algorithms. In another implementation, the access network device and the PRU / UE may alternatively perform clustering operations on the training data provided by the access network device and the PRU / UE by using the same clustering algorithm, without limitation.

[0168] 612: Optionally, the positioning device trains and / or updates an AI model based on the training data.

[0169] In this example, the technical solution provided in this application is applied to an uplink positioning scenario. The access network device and / or PRU / UE performs a clustering process on the provided training data to classify the training data into reference data and non-reference data, and reports the complete information of the reference data and the incremental information of the non-reference data, thereby reducing the overhead of reporting the training data.

[0170] 3. AI-based CSI prediction FIG. 7 is a diagram illustrating CSI prediction based on an AI model. As shown in FIG. 7, the basic principle of AI-CSI prediction is to use the feature extraction and adaptation capabilities of the neural network to train a neural network by using a training set containing a large amount of offline CSI. As a result, the neural network can learn channel change modes and adapt a nonlinear mapping relationship between historical CSI and future CSI, replacing the closed-form expression of channel prediction based on a mathematical model. In the inference stage, some historical CSI (e.g., CSI fed back by the receiving end and / or CSI predicted by the transmitting end at a past time) is input into the neural network used for channel prediction, and a predicted value of CSI at a future time is output. The transmitting end can predict CSI, and the receiving end does not need to feedback estimated CSI at the prediction time. Therefore, air interface CSI feedback overhead can be reduced.

[0171] In AI-based CSI prediction, when AI model training is performed on the access network side, the access network device needs to obtain the channel estimation results of one or more CSI-RSs on the UE side and use the channel estimation results as labels to perform AI model training.

[0172] In another example, different from the above embodiment in which information obtained by clustering training data is reported, the second network element may report training data to the first network element by using a differential method in an AI-based CSI prediction scenario, as described below with reference to FIG. 8. In this scenario, the first network element is an access network device and the second network element is a UE.

[0173] FIG. 8 illustrates an example of applying the technical solution provided in this application to AI-based CSI prediction.

[0174] 801: Optionally, the access network device determines that the AI ​​model needs to be trained or updated.

[0175] 802: The access network device sends second instruction information to the UE, where the second instruction information indicates to the UE to collect training data for the AI ​​model.

[0176] 803: The access network device sends the CSI-RS to the UE.

[0177] The UE obtains channel measurements by measuring the CSI-RS coming from the access network device.

[0178] 804: The UE sends first information to the access network device, where the first information includes first type information and second type information, and the first type information and the second type information indicate training data provided by the UE.

[0179] Specifically, the training data provided by the UE includes reference data and non-reference data. The first type of information includes complete information of channel measurement results belonging to the reference data, and the second type of information includes incremental information of channel measurement results belonging to the non-reference data relative to the reference data corresponding to the channel measurement results belonging to the non-reference data. Optionally, in this example, the reference data and non-reference data are not determined using a clustering algorithm, but are negotiated or agreed upon in advance by the first network element and the second network element, pre-configured by the first network element, or pre-defined in a protocol. This is not limiting. For example, the first network element pre-configures a period T for reporting training data by the second network element, and configures the second network element to report the non-reference data twice after each reporting of the reference data. For example, the complete information of channel measurement results obtained by the UE by measuring CSI-RS is reported at time t1. In this case, the channel measurement results reported at time t1 belong to the reference data. When channel measurement results are reported at time t2 and time t3, differential values ​​of the channel measurement results obtained by the UE by measuring the CSI-RS with respect to the channel measurement result reported at time t1 are reported. Specifically, if the channel measurement results estimated at time t1 to time t3 are channel measurement result 1, channel measurement result 2, and channel measurement result 3, respectively, complete information of channel measurement result 1 is reported at time t1, differential information corresponding to channel measurement result 2 (specifically, the difference between channel measurement result 2 and channel measurement result 1) is reported at time t2, and differential information corresponding to channel measurement result 3 (specifically, the difference between channel measurement result 3 and channel measurement result 1) is reported at time t3.

[0180] The access network device receives first information from the UE.

[0181] Optionally, in step 804, the transmission of the training data is from the perspective of the entire transmission. Specifically, in a differential manner, the first type of information and the second type of information may be transmitted using different messages. For example, each time the UE acquires one channel measurement result, the UE may report the channel measurement result once. During a specific report, if the currently acquired channel measurement result is reported as reference data, the UE transmits the first type of information to indicate the complete information of the reference data, or if the currently acquired channel measurement result is reported as non-reference data, the UE transmits the second type of information to indicate the incremental information corresponding to the non-reference data. For example, if the channel measurement result corresponding to time t1 is reference data, the first type of information is transmitted at time t1 to indicate the complete information of channel measurement result 1. If the channel measurement result corresponding to time t2 or time t3 is non-reference data, the second type of information is transmitted at time t2 or time t3 to indicate the incremental information of the channel measurement result acquired at time t2 or time t3 relative to channel measurement result 1, respectively. Therefore, step 804 may alternatively be divided into multiple steps, each step corresponding to one transmission of full information of the reference data or one transmission of incremental information corresponding to the non-reference data.

[0182] Optionally, the time-domain differential method is used as an example in the above example, and training data may alternatively be reported using a frequency-domain differential method. For example, at time t1, the UE reports the full information of channel measurement result 1 obtained by measuring CSI-RS on subband 1 and the differential information, i.e., incremental information, of channel measurement results 2 to N relative to channel measurement result 1. Channel measurement results 2 to N are channel measurements obtained by the UE by measuring CSI-RS on subbands 2 to N. In the frequency-domain differential method, channel measurement result 1 obtained by measuring CSI-RS on subbands 2 to N is used as reference data, and full information is reported in this case. Channel measurement results 2 to N obtained on subbands 2 to N, respectively, are all used as non-reference data, and corresponding incremental information is reported in this case. The method of reporting training data at time t2 and time t3 is the same as that at time t1. Details will not be described again.

[0183] 805, the access network device recovers all training data provided by the UE based on the first information.

[0184] 806, optionally, the access network device trains or updates the AI ​​model based on the training data.

[0185] In this example, the technical solution provided in this application is applied to an AI-based CSI prediction scenario. The UE performs a clustering process on the provided training data to classify the training data into reference data and non-reference data, and reports the complete information of the reference data and the incremental information of the non-reference data, thereby reducing the overhead of reporting the training data.

[0186] 4. AI-based CSI feedback In communication systems such as LTE and NR, a base station needs to acquire downlink channel state information (CSI) to determine configurations such as resources, modulation and coding scheme (MCS), and precoding for scheduling a UE's downlink data channel. In a time division duplex (TDD) system, because there is reciprocity between the uplink and downlink channels, a base station may acquire uplink CSI by measuring an uplink reference signal to estimate accurate downlink CSI, e.g., to use the uplink CSI as downlink CSI. In a frequency division duplex (FDD) system, because reciprocity between the uplink and downlink cannot be guaranteed, downlink CSI is acquired by a UE by measuring a downlink reference signal. For example, a UE acquires downlink CSI by measuring a CSI-RS or a synchronization signal / physical broadcast channel block (SSB). Therefore, the UE needs to generate a CSI report in a manner predefined in the protocol or configured by the base station, and feed back the CSI to the base station so that the base station obtains the downlink CSI.

[0187] Figure 9 illustrates the AI-based CSI feedback process. In AI-CSI feedback, when an AI model is deployed on the access network device side, the access network device needs to obtain the CSI-RS estimation results from the UE side and perform training using the estimation results as labels (also known as ground truth). The AE model includes two submodels: an encoder and a decoder. AE may generally refer to a network structure including two submodels. The AE model may also be called a bilateral model, a dual-ended model, or a collaborative model. The encoder and decoder of the AE are usually trained together and may be used collaboratively. CSI feedback may be implemented based on the AI ​​model of the AE. For example, the UE side performs CSI compression and quantization through the encoder, and the access network device performs CSI reconstruction through the decoder. As shown in Figure 9, for the access network device, the input of the AI ​​model is the CSI fed back by the UE side, and the output is the reconstructed CSI. The CSI measured by the UE side needs to be used as the ground truth for the recovered CSI in model training.

[0188] FIG. 10 illustrates an example of applying the technical solution provided in this application to AI-based CSI feedback.

[0189] 901: Optionally, an access network device determines that an AI model needs to be trained or updated.

[0190] 902: Optionally, the access network device sends first indication information to the UE, where the first indication information is used by the UE to determine a first clustering algorithm.

[0191] Optionally, the first clustering algorithm may be predefined in the protocol.

[0192] 903: Optionally, the access network device sends instruction information d to the UE, where the instruction information d indicates to the UE to collect training data.

[0193] 904: The access network device sends the CSI-RS to the UE.

[0194] The UE obtains channel measurements by measuring the CSI-RS coming from the access network device.

[0195] If possible, when the AI ​​model on the UE side is distributed by the access network device (or the AI ​​model is jointly trained by the access network side and the UE side), the UE performs a clustering process on the channel measurement results based on a first clustering algorithm to obtain reference data and non-reference data. Further, the UE reports the complete information of the reference data and the incremental information of the non-reference data relative to the reference data corresponding to the non-reference data to the access network device, as shown in steps 905 and 906.

[0196] 905: The UE performs a clustering process on the channel measurement result based on a first clustering algorithm to obtain reference data and non-reference data.

[0197] 906: The UE sends first information to the access network device, where the first information includes first type information and second type information, where the first type information and the second type information indicate training data provided by the UE. The training data includes reference data and non-reference data. The first type information includes complete information of channel measurement results belonging to the reference data, and the second type information includes incremental information corresponding to the channel measurement results belonging to the non-reference data.

[0198] In another possible case, the AI ​​model on the UE side does not depend on the training and distribution by the access network side (i.e., the access network side and the UE side are trained separately). In this case, the UE separately performs clustering on the channel measurement results and the quantized compressed information corresponding to the channel measurement results, and reports the information obtained by the clustering, as shown in steps 907 and 908.

[0199] 907: The UE performs a clustering process on the channel measurement results based on a first clustering algorithm to obtain reference data and non-reference data related to the channel measurement results, and performs clustering on the quantization compression information of the channel measurement results to obtain reference data and non-reference data related to the quantization compression information.

[0200] Optionally, the same or different clustering algorithms may be used for the channel measurement results and the quantized compressed information of the channel measurement results, without any limitation thereon.

[0201] 908: The UE sends first information to the access network device, where the first information includes first type information and second type information, where the first type information and the second type information indicate training data provided by the UE. The training data includes channel measurement results and quantization compression information of the channel measurement results. Specifically, the first type information includes complete information of the channel measurement results belonging to the reference data and complete information of the quantization compression information of the reference data, and the second type information includes incremental information corresponding to the channel measurement results belonging to the non-reference data and incremental information corresponding to the quantization compression information of the non-reference data.

[0202] 909: The access network device restores all training data based on the corresponding clustering algorithm and the first information.

[0203] 910: Optionally, the access network device trains or updates the AI ​​model based on the training data.

[0204] In this example, the technical solution provided in this application is applied to an AI-based CSI feedback scenario. The UE performs a clustering process on the provided training data to classify the training data into reference data and non-reference data, and reports the complete information of the reference data and the incremental information of the non-reference data, thereby reducing the overhead of reporting the training data.

[0205] The above details the method for reporting training data of an AI model provided in the present application. A corresponding communication device will now be described. Referring to FIG. 11, the present application provides a communication device 1000.

[0206] 11 , the communication device 1000 includes a processing module 1001 and a communication module 1002. The communication device 1000 may be a terminal device, or a communication device capable of implementing a method, such as a chip, chip system, or circuit, used in a terminal device, used in cooperation with a terminal device, or executed on the terminal device side. Alternatively, the communication device 1000 may be a network device, or a communication device capable of implementing a method, such as a chip, chip system, or circuit, used in a network device, used in cooperation with a network device, or executed on the network device side. For example, the network device may be a positioning device, an access network device, etc. in the method embodiments of the present application.

[0207] The communication module may also be referred to as a transceiver module, a transceiver, a transceiver device, etc. The processing module may also be referred to as a processor, a processing board, a processing unit, a processing device, etc. Optionally, the communication module is configured to perform transmitting and receiving operations at the terminal device side or the network device side in the manner described above. A component configured to perform a receiving function in the communication module may be considered a receiving unit, and a component configured to perform a transmitting function in the communication module may be considered a transmitting unit. In other words, the communication module includes a receiving unit and a transmitting unit.

[0208] When the communication apparatus 1000 is used in a terminal device, the processing module 1001 may be configured to perform the processing functions of the terminal device in the embodiments of Figures 2 to 10, and the communication module 1002 may be configured to perform the receiving and transmitting functions of the terminal device in the embodiments of Figures 2 to 10.

[0209] When the communication device 1000 is used in a network device, the processing module 1001 may be configured to perform the processing functions of the network device (e.g., a positioning device or an access network device) in the embodiments of Figures 2 to 10, and the communication module 1002 may be configured to perform the receiving and transmitting functions of the network device in the embodiments of Figures 2 to 10.

[0210] It should be noted that the first network element or the second network element shown in Figure 2 may be different network elements in different application scenarios detailed in the above method embodiments. The specific implementation of the first network element or the second network element can be understood in the method embodiments. The details will not be described again.

[0211] In addition, it should be noted that the communication module and / or the processing module may be implemented by using virtual modules. For example, the processing module may be implemented using a software functional unit or a virtual device, and the communication module may be implemented using a software function or a virtual device. Alternatively, the processing module or the communication module may be implemented using a physical device. For example, if the device is implemented using a chip / chip circuit, the communication module may be an input / output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operations) and output operations (corresponding to the aforementioned transmitting operations). The processing module is an integrated processor, a microprocessor, or an integrated circuit.

[0212] The division into modules in this application is merely an example and is merely a logical division of functions, and other divisions may be used in actual implementation. Furthermore, the functional modules in the examples of this application may be integrated into one processor, and each module may exist physically alone, or two or more modules may be integrated into one module. The integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0213] Based on the same technical concept, referring to Fig. 12, the present application further provides a communication device 1100. Optionally, the communication device 1100 may be a chip or a chip system. Optionally, in the present application, the chip system may include a chip, or may include a chip and another individual device.

[0214] The communications device 1100 may be configured to perform the functions of any network element in the communications system described in the previous examples. The communications device 1100 may include at least one processor 1110. Optionally, the processor 1110 is coupled to a memory. The memory may be located within the device, integrated with the processor, or external to the device. For example, the communications device 1100 may further include at least one memory 1120. The memory 1120 stores computer programs, computer programs or instructions, and / or data necessary to perform any of the previous examples. The processor 1110 may execute computer programs stored in the memory 1120 to complete the method of any one of the previous examples.

[0215] The communication device 1100 may further include a communication interface 1130, and the communication device 1100 may exchange information with other devices via the communication interface 1130. For example, the communication interface 1130 may be a transceiver, a circuit, a bus, a module, a pin, or another type of communication interface. If the communication device 1100 is a chip-type device or circuit, the communication interface 1130 in the device 1100 may alternatively be an input / output circuit that inputs information (or receives information) and outputs information (or transmits information). The processor may be an integrated processor, a microprocessor, an integrated circuit, or a logic circuit. The processor may determine output information based on the input information.

[0216] The term "coupling" in this application may refer to an indirect coupling or communication connection between devices, units, or modules, and may be implemented in an electrical, mechanical, or other form, and is used to exchange information between the devices, units, or modules. The processor 1110 may operate in cooperation with the memory 1120 and the communication interface 1130. The specific connection medium between the processor 1110, the memory 1120, and the communication interface 1130 is not limited in this application.

[0217] Optionally, as shown in Figure 11, the processor 1110, the memory 1120, and the communication interface 1130 are connected to each other via a bus 1140. The bus 1140 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be categorized into an address bus, a data bus, a control bus, etc. For ease of representation, the bus is represented using only one thick line in Figure 11, but this does not imply that there is only one bus or only one type of bus.

[0218] In this application, a processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array, another programmable logic device, a discrete gate, a transistor logic device, or a discrete hardware component that may implement or perform the methods, steps, and logic block diagrams disclosed in this application. A general-purpose processor may be a microprocessor, any conventional processor, etc. The steps of the methods disclosed with reference to this application may be performed directly by a hardware processor, or may be performed by a combination of hardware and software modules within a processor.

[0219] In this application, memory may be non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or may be volatile memory, such as random-access memory (RAM). Memory is any other medium accessible to a computer that can carry or store program code, such as instructions or data structures. Memory in this application may alternatively be a circuit or any other device capable of implementing a storage function and configured to store program instructions and / or data.

[0220] In a possible implementation, the communication device 1100 may be used on a network device side, such as an access network device or a core network device (e.g., a positioning device, specifically an LMF network element) in an embodiment of the present application, or a host or cloud device in an OTT system. Specifically, the communication device 1100 may be a network device or a device capable of supporting the network device in performing corresponding functions of the network device side in any one of the aforementioned examples. The memory 1120 stores computer programs (or instructions) and / or data for performing functions of the network device side in any one of the aforementioned examples. The processor 1110 may execute the computer program stored in the memory 1120 to complete the method performed by the network device side in any one of the aforementioned examples. A communication interface in the communication device 1100 may be configured to interact with a terminal device, transmit information to the terminal device, or receive information from the terminal device. Additionally, optionally, the communication interface within the communication device 1000 may be further configured to interact with another network element (e.g., a third network element), receive a reference signal from the third network element, etc.

[0221] In another possible implementation, the communication device 1100 may be used in a terminal device. Specifically, the communication device 1100 may be a terminal device or a device capable of supporting a terminal device in performing the functions of the terminal device in any one of the foregoing examples. The memory 1120 stores computer programs (or instructions) and / or data for performing the functions of the terminal device in any one of the foregoing examples. The processor 1110 may execute the computer programs stored in the memory 1120 to complete the method performed by the terminal device in any one of the foregoing examples. A communication interface in the communication device 1100 may be configured to interact with a network device (e.g., an access network device) and to transmit information to or receive information from the network device.

[0222] The communication device 1100 provided in this example may be used in a network device (e.g., an access network device) to complete a method performed by the network device, or may be used in a terminal device to complete a method performed by the terminal device. Therefore, for technical effects that can be achieved by the communication device 1100, please refer to the aforementioned method embodiments. Details will not be repeated here.

[0223] Based on the foregoing examples, the present application provides a communication system. In one example, the communication system includes a first network element and a second network element. In another example, the communication system includes a first network element, a second network element, and a third network element. The communication system may implement the method for reporting training data of an AI model provided in the embodiments shown in Figures 2 to 10.

[0224] All or part of the technical solutions provided in this application may be implemented by using software, hardware, firmware, or any combination thereof. When software is used to implement the technical solutions, all or part of the technical solutions may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the procedures or functions according to the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a terminal device, an access network device, or another programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired connection (e.g., coaxial cable, optical fiber, or digital subscriber line (DSL)) or via a wireless connection (e.g., infrared, radio waves, or microwaves). The computer-readable storage medium may be any available medium accessible to a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, or a magnetic tape), an optical medium (e.g., a digital video disc (DVD)), a semiconductor medium, etc.

[0225] In this application, cross-references may be made between examples without logical contradiction, for example, cross-references may be made between methods and / or terms in method examples, cross-references may be made between functions and / or terms in device examples, and cross-references may be made between functions and / or terms in device examples and method examples.

[0226] The units described as separate parts may or may not be physically separate, and the parts presented as units may or may not be physical units. Specifically, the parts may be located in one location or distributed over multiple network units. Some or all of the units may be selected according to actual requirements to achieve the objectives of the solutions of the embodiments.

[0227] In addition, the functional units of the embodiments of the present application may be integrated into one processing unit, and each of the units may exist physically alone, or two or more units may be integrated into one unit.

[0228] In the embodiments of the present application, "at least one (item)" refers to one (item) or multiple (items). "Multiple (items)" means two (items) or more (items). The term "and / or" describes an association relationship for describing related objects and indicates that three relationships may exist. For example, A and / or B may represent the following three cases: when only A exists, when both A and B exist, and when only B exists. The character " / " generally indicates an "or" relationship between related objects. Furthermore, although terms such as "first," "second," etc. may be used to describe objects in the present disclosure, it should be understood that these objects are not limited to these terms. These terms are used merely to distinguish objects from each other.

[0229] In the embodiment of the present application, "instruction information a / instruction information b" etc. are merely expressions for distinguishing different information, and may be expressed as, for example, instruction information #1 / instruction information #2 etc. This should not constitute any limitation on the technical solution.

[0230] The term "comprises" and any other variations thereof referred to in the embodiments of this application are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or device may include a series of steps or units that are not limited to the enumerated steps or units, and may optionally further include other unenumerated steps or units, or may optionally further include other steps or units inherent to the process, method, product, or device. It should be noted that in this application, terms such as "example," "for example," and the like are used to provide an example, illustration, or explanation. A method or design approach described in this application as an "example" or "for example" should not be construed as being preferred or having more advantages than other methods or design approaches. Strictly speaking, the use of terms such as "example," "for example," and the like is intended to present the related concept in a particular manner.

[0231] When a function is implemented in the form of a software functional unit and sold or used as an independent product, the function may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may essentially be implemented in the form of a software product, or a portion of the technical solution may be implemented in the form of a software product. The computer software product is stored in a storage medium and includes some instructions that instruct a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes any medium that can store program code, such as a USB flash drive, a removable hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0232] Those skilled in the art can make various modifications and variations to this application without departing from the scope of this application. Thus, this application intends to cover these modifications and variations of this application as long as they fall within the scope defined by the following claims and the equivalent technology of this application. [Explanation of symbols]

[0233] 110 Network Devices 120 Terminal Devices 130 Terminal Devices 140 Positioning Devices 1000 Communication Equipment 1001 Processing Module 1002 communication module 1100 Communication equipment 1110 processor 1120 memory 1130 Communication Interface 1140 Bus

Claims

1. 1. A method for transmitting training data for an AI model, comprising: receiving, by a first network element, first information from a second network element, the first information including a first type of information and a second type of information, the first type of information and the second type of information indicating training data for the AI ​​model provided by the second network element; Including, the training data includes one or more reference data and non-reference data, the first type of information includes complete information of the one or more reference data, and the second type of information includes incremental information of the non-reference data relative to reference data corresponding to the non-reference data; A method for sending training data for an AI model.

2. The method of claim 1 , wherein the one or more reference data are determined from the training data based on a first clustering algorithm.

3. Prior to the step of receiving, by the first network element, the first information from the second network element, the method further comprises: sending, by the first network element, first indication information to the second network element, wherein the first indication information is used to determine the first clustering algorithm; 3. The method of claim 2, further comprising:

4. The AI ​​model is applied to downlink positioning, the first network element includes a Location Management Function (LMF) network element, and the second network element includes a terminal device; the training data includes channel measurement results and / or location information of the second network element, and the channel measurement results are obtained by the second network element based on a positioning reference signal coming from a third network element; The method of claim 3.

5. If the training data includes the channel measurement results, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data; or or, If the training data includes the location information of the second network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data, the first type of information includes complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the location information belonging to the non-reference data; or or, When the training data includes the channel measurement results and the location information of the second network element, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the location information includes location information belonging to the reference data and location information belonging to the non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data and complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data and incremental information corresponding to the location information belonging to the non-reference data. The method of claim 4.

6. The AI ​​model is applied to uplink positioning, the first network element includes a location management function (LMF) network element, and the second network element includes an access network device; the training data includes channel measurement results, the channel measurement results are acquired by the second network element based on a sounding reference signal coming from a third network element, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data; The method of claim 3.

7. The AI ​​model is applied to uplink positioning, the first network element includes a location management function (LMF) network element, and the second network element includes a terminal device; the training data includes location information of the second network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data, the first type of information includes complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the location information belonging to the non-reference data; The method of claim 3.

8. The AI ​​model is applied to uplink positioning, the first network element includes a location management function (LMF) network element, and the second network element includes an access network device; The method comprises: receiving, by the first network element, third information from the third network element, the third information including the first type of information and the second type of information, the first type of information and the second type of information included in the third information indicating training data provided by the third network element; further comprising the training data includes the training data provided by the second network element and the training data provided by the third network element, the training data provided by the second network element includes channel measurement results, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the training data provided by the third network element includes location information of the third network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data; the first type of information included in the first information includes complete information of the channel measurement result belonging to the reference data, and the second type of information included in the first information includes incremental information corresponding to the channel measurement result belonging to the non-reference data; the first type of information included in the third information includes complete information of the location information belonging to the reference data, and the second type of information included in the third information includes incremental information corresponding to the location information belonging to the non-reference data; The method of claim 3.

9. The AI ​​model is applied to channel state information (CSI) prediction, the first network element includes an access network device, and the second network element includes a terminal device; the training data includes CSI estimation results obtained by the second network element based on a channel state information reference signal (CSI-RS) coming from the first network element, and the CSI estimation results include CSI estimation results belonging to the reference data and CSI estimation results belonging to the non-reference data; the first type of information includes complete information of the CSI estimation result belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation result belonging to the non-reference data. The method of claim 3.

10. The AI ​​model is applied to channel state information (CSI) feedback, the first network element includes an access network device, and the second network element includes a terminal device; the training data includes a CSI estimation result obtained by the second network element based on a channel state information reference signal (CSI-RS) coming from the first network element, or the training data includes a CSI estimation result obtained by the second network element based on the CSI-RS and quantization compression information corresponding to the CSI estimation result; The method of claim 3.

11. the training data includes the CSI estimation results, the CSI estimation results include a CSI estimation result belonging to the reference data and a CSI estimation result belonging to the non-reference data, the first type of information includes complete information of the CSI estimation results belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation results belonging to the non-reference data; or, the training data includes the CSI estimation result and the quantized compression information of the CSI estimation result, the CSI estimation result includes the CSI estimation result belonging to the reference data and the CSI estimation result belonging to the non-reference data, the first type of information includes complete information of the CSI estimation result belonging to the reference data and complete information of quantized compression information of the CSI estimation result belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation result belonging to the non-reference data and incremental information corresponding to the quantized compression information of the CSI estimation result belonging to the non-reference data. The method of claim 10.

12. 1. A method for transmitting training data for an AI model, comprising: transmitting, by a second network element, first information to a first network element, the first information including a first type of information and a second type of information, the first type of information and the second type of information indicating training data for the AI ​​model provided by the second network element; Including, the training data includes one or more reference data and non-reference data, the first type of information includes complete information of the one or more reference data, and the second type of information includes incremental information of the non-reference data relative to reference data corresponding to the non-reference data; A method for sending training data for an AI model.

13. The method of claim 12 , wherein the one or more reference data are determined from the training data based on a first clustering algorithm.

14. Prior to the step of transmitting the first information by the second network element to the first network element, the method further comprises: receiving, by the second network element, first indication information from the first network element, the first indication information being used to determine the first clustering algorithm; determining, by the second network element, the first clustering algorithm based on the first indication; 14. The method of claim 13, comprising:

15. The AI ​​model is applied to downlink positioning, the second network element includes a terminal device, and the first network element includes a Location Management Function (LMF) network element; the training data includes channel measurement results and / or location information of the second network element, and the channel measurement results are obtained by the second network element based on a positioning reference signal coming from a third network element; The method of claim 14.

16. If the training data includes the channel measurement results, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data; or or, If the training data includes the location information of the second network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data, the first type of information includes complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the location information belonging to the non-reference data; or or, When the training data includes the channel measurement results and the location information of the second network element, the channel measurement results include the channel measurement results belonging to the reference data and the channel measurement results belonging to the non-reference data, the location information includes the location information belonging to the reference data and the location information belonging to the non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data and the complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data and incremental information corresponding to the location information belonging to the non-reference data.

16. The method of claim 15.

17. The AI ​​model is applied to uplink positioning, the second network element includes an access network device, and the first network element includes a location management function (LMF) network element; the training data includes channel measurement results, the channel measurement results are acquired by the second network element based on a sounding reference signal coming from a third network element, the channel measurement results include channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data, the first type of information includes complete information of the channel measurement results belonging to the reference data, and the second type of information includes incremental information corresponding to the channel measurement results belonging to the non-reference data; The method of claim 14.

18. The AI ​​model is applied to uplink positioning, the second network element includes a terminal device, and the first network element includes a location management function (LMF) network element; the training data includes location information of the second network element, the location information includes location information belonging to the reference data and location information belonging to the non-reference data, the first type of information includes complete information of the location information belonging to the reference data, and the second type of information includes incremental information corresponding to the location information belonging to the non-reference data; The method of claim 14.

19. The AI ​​model is applied to uplink positioning, the second network element includes an access network device, and the first network element includes a location management function (LMF) network element; the training data provided by the second network element includes channel measurement results, the channel measurement results including the channel measurement results belonging to the reference data and channel measurement results belonging to the non-reference data; the first type of information included in the first information includes complete information of the channel measurement result belonging to the reference data, and the second type of information included in the first information includes incremental information corresponding to the channel measurement result belonging to the non-reference data; The method of claim 14.

20. the AI ​​model is applied to CSI prediction, the second network element includes a terminal device, and the first network element includes an access network device; the training data includes CSI estimation results obtained by the second network element based on a channel state information reference signal (CSI-RS) coming from the first network element, and the CSI estimation results include CSI estimation results belonging to the reference data and CSI estimation results belonging to the non-reference data; the first type of information includes complete information of the CSI estimation result belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation result belonging to the non-reference data. The method of claim 14.

21. the AI ​​model is applied to CSI feedback, the second network element includes a terminal device, and the first network element includes an access network device; the training data includes a CSI estimation result obtained by the second network element based on a channel state information reference signal (CSI-RS) coming from the first network element, or the training data includes a CSI estimation result obtained by the second network element based on the CSI-RS and quantization compression information corresponding to the CSI estimation result; The method of claim 14.

22. the training data includes the CSI estimation results, the CSI estimation results include a CSI estimation result belonging to the reference data and a CSI estimation result belonging to the non-reference data, the first type of information includes complete information of the CSI estimation results belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation results belonging to the non-reference data; or, the training data includes the CSI estimation result and the quantized compressed information of the CSI estimation result, the CSI estimation result includes a CSI estimation result belonging to the reference data and a CSI estimation result belonging to the non-reference data, the first type of information includes complete information of the CSI estimation result belonging to the reference data and complete information of quantized compressed information of the CSI estimation result belonging to the reference data, and the second type of information includes incremental information corresponding to the CSI estimation result belonging to the non-reference data and incremental information corresponding to the quantized compressed information of the CSI estimation result belonging to the non-reference data.

22. The method of claim 21.

23. 23. A communications device configured to implement any one of claims 1 to 22.

24. 23. A processor coupled to a memory, the processor configured to invoke computer program instructions stored in the memory to perform the method of any one of claims 1 to 22.

2. A communication device comprising:

25. A computer-readable storage medium storing instructions that, when executed on a computer, enable the computer to perform the method of any one of claims 1 to 22. A computer-readable storage medium.

26. 23. A computer program product, wherein the computer-readable storage medium stores instructions that, when executed on a computer, enable the computer to perform the method of any one of claims 1 to 22.

27. A communication system comprising a communication device according to claim 23 or 24.

Images