Methods and apparatuses for data association in a wireless communication network

Abstract

A method (8800) performed by a wireless device (1000) in a wireless communication network is disclosed The method comprises obtaining (801) a configuration for collecting a first set of data. The method further comprises collecting (802) the first set of data, wherein an association between the first set of data and a second set of data is determined by the wireless device or a network node.

Description

[0001] Methods and Apparatuses for Data Association in a Wireless Communication Network

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to the field of wireless communications, and in particular to methods and apparatuses for data association in a wireless communication network such as advanced 5G or 6G networks.

[0004] BACKGROUND

[0005] 5G mobile communications have been driven by the need to provide ubiquitous connectivity for applications as diverse automotive communications, remote control with feedback, video downloads, as well as data applications for Internet-of-Things (loT) devices, machine type communication (MTC) devices, etc.

[0006] Positioning technologies are for example a critical area for 5G or the next generation 6G cellular networks, enabling services such as navigation, asset tracking, emergency response, and location-based applications. Due to the rapidly growing demands for high-accuracy, real-time positioning, there exists a need to further improve positioning accuracy.

[0007] Data driven methods, particularly methods using Artificial Intelligence (Al) / Machine Learning (ML) models, have been increasingly adopted in telecommunications systems to harness vast dataset generated by the wireless network for intelligent decision-making. For data driven methods, to adapt to dynamic, heterogeneous environments and conditions, it is important to make data analysis and association, so that verifiable data can be used for training, validation, and performance monitoring of the machine learning models.

[0008] SUMMARY

[0009] It is an objective of the embodiments herein to provide methods and apparatuses for data association or analysis in a wireless communication network such as advanced 5G or 6G networks.According to a first aspect of some embodiments herein, there is provided a method performed by a wireless device in a wireless communication network. The method comprises:

[0010] - obtaining a configuration for collecting a first set of data; and

[0011] - collecting the first set of data, wherein an association between the first set of data and a second set of data is determined by the wireless device or a network node.

[0012] According to a second aspect of some embodiments herein, there is provided a method performed by a wireless device in a wireless communication network. The method comprises:

[0013] - obtaining a configuration for collecting a first set of data; and

[0014] - collecting the first set of data, wherein the first set of data is combined with one or more other sets of data from one or more other wireless devices, and an association between the combined set of data and a second set of data is determined by the wireless device or a network node.

[0015] According to a third aspect of some embodiments herein, there is provided a method performed by a wireless device in a wireless communication network. The method comprises:

[0016] - obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a first set of data and the second set of data is determined by the wireless device or a network node.

[0017] According to a fourth aspect of some embodiments herein, there is provided a method performed by a network node in a wireless communication network. The method comprises:

[0018] - obtaining a configuration for collecting a first set of data; and

[0019] - collecting the first set of data, wherein an association between the first set of data and a second set of data is determined by a network node or a wireless device.According to a fifth aspect of some embodiments herein, there is provided a method performed by a network node in a wireless communication network. The method comprises:

[0020] - obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a first set of data and the second set of data is determined by a wireless device or a network node.

[0021] According to a sixth aspect of some embodiments herein, there is provided a method performed by a network node in a wireless communication network. The method comprises:

[0022] - obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a combined set of data including a first set of data and the second set of data is determined by a wireless device or a network node.

[0023] According to a seventh aspect of some embodiments herein, there is provided a method performed by a first network node in a wireless communication network. The method comprises:

[0024] - receiving a first set of data from a second network node;

[0025] - generating or receiving a second set of data; and

[0026] - associating between the first set of data and the second set of data;

[0027] According to an eighth aspect of some embodiments herein, there is provided a wireless device comprising a processor and a memory containing instructions executable by said processor. The wireless device is operative to perform the method according to the first aspect.

[0028] According to a ninth aspect of some embodiments herein, there is provided a wireless device comprising a processor and a memory containing instructions executable by said processor. The wireless device is operative to perform the method according to the second aspect.According to a tenth aspect of some embodiments herein, there is provided a wireless device comprising a processor and a memory containing instructions executable by said processor. The wireless device is operative to perform the method according to the third aspect.

[0029] According to an eleventh aspect of some embodiments herein, there is provided a network node comprising a processor and a memory containing instructions executable by said processor. The network node is operative to perform the method according to the fourth aspect.

[0030] According to a twelfth aspect of some embodiments herein, there is provided a network node comprising a processor and a memory containing instructions executable by said processor. The network node is operative to perform the method according to the fifth aspect.

[0031] According to a thirteenth aspect of some embodiments herein, there is provided a network node comprising a processor and a memory containing instructions executable by said processor. The network node is operative to perform the method according to the sixth aspect.

[0032] According to a fourteenth aspect of some embodiments herein, there is provided a network node comprising a processor and a memory containing instructions executable by said processor. The network node is operative to perform the method according to the seventh aspect.

[0033] According to a fifteenth aspect of some embodiments herein, there is provided a computer program comprising instructions which when executed on at least one processor of the wireless device, cause the at least said one processor to carry out the actions or method steps presented herein.

[0034] According to a sixteenth aspect of some embodiments herein, there is also provided a computer program comprising instructions which when executed on at least one processor of the network node, cause the at least said one processor to carry out the actions or method steps presented herein.One advantage of the present invention is to make the data analysis and association adaptable to dynamic, heterogeneous environments and conditions, so that better positioning estimations can be made.

[0035] Further advantages include efficient reporting, reduced ambiguity in AI / ML data collection for training, and improved AI / ML model performance. This, in turn, minimizes unnecessary signaling overhead, optimizing overall system efficiency.

[0036] Additional advantages of the embodiments herein are provided in the detailed description of this disclosure.

[0037] BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Embodiments of the present invention are now described in further detail with reference to the accompanying drawings, in which:

[0039] Fig. 1 shows a diagram illustrating a simplified positioning architecture for UE positioning in a 5G system (5GS) according to some embodiments herein;

[0040] Fig. 2 shows a flow diagram illustrating a UE-based AI / ML positioning method according to some embodiments herein;

[0041] Fig. 3 shows a flow diagram illustrating an LMF-based AI / ML positioning method according to some embodiments herein;

[0042] Fig. 4 shows a diagram illustrating an example where a first set of data and a second set of data are collected at different time points according to some embodiments herein;

[0043] Fig. 5 illustrates an example where a validity window is associated with a label according to some embodiments herein;

[0044] Fig. 6 illustrates an example where a plurality of resources is associated according to some embodiments herein;Fig. 7A and 7B illustrate two examples where one or more labels are associated with / mapped to multiple measurements according to some embodiments herein;

[0045] Fig. 8 illustrates a flowchart of a method performed by a wireless device (e.g., a UE or an loT device) according to some embodiments herein;

[0046] Fig. 9 illustrates a flowchart of a method performed by a network node according to some embodiments herein;

[0047] Fig. 10 is a block diagram depicting a wireless device (e.g., a UE or an loT device) according to exemplary embodiments herein; and

[0048] Fig. 11 is a block diagram depicting a network node according to exemplary embodiments herein.

[0049] DETAILED DESCRIPTION

[0050] In the following, a detailed description of the exemplary embodiments is described in conjunction with the drawings, in various scenarios to enable easier understanding of the solutions described herein.

[0051] It is to be noticed that in the following figures, although a wireless device is often illustrated by a UE, the wireless device may be other devices such as an loT device.

[0052] Figure 1 is a diagram illustrating a simplified positioning architecture for UE positioning in a 5G system (5GS). A positioning protocol is terminated between the Location Management Function (LMF) and the UE, which is known as LTE positioning protocol (LPP). Likewise, a positioning protocol is terminated between NG-RAN node and LMF and is called NRPPa. Further, RRC protocol runs between UE and NG-RAN node, which provides configuration to the UE for transmitting uplink reference signals (e.g. SRS), likewise, there may be layer 2 protocols, such as MAC used for purposes such as activating or deactivating one or more configurations (such as SRS configurations, measurement gap configurations, bandwidth parts configuration, ... etc) or sending some additional information or updates from lower layers of protocol stack between the access network and the UE. Furthermore, there may be SLPP protocol terminated between two UEs or relayed between UE and AMFfor positioning using the sidelink channel. A UE-DN server is shown on the data network, which interfaces directly with the UE, and it may deliver, update, validate the model deployed, to-be deployed or fine-tuned at the UE or in general maintain software updates as needed. The connection to the UE-DN may be through the 5GS or through other means (e.g. WLAN) and it may be transparent to the underlying network. The Non-3GPP Interworking Function (N3IWF) allows connectivity to non-3GPP access to the 5GS. This may allow data connection to be established over WLAN connections to the UE server on data network (i.e. , UE-DN server), thus communicating between the UE and data network (DN) transparently to the network. The AMF is an access and mobility function, and for positioning application AMF is responsible for selecting a suitable LMF for a given UE. Likewise, the UE registers itself with the AMF. In 5GS, the LPP signaling is transmitted transparently via NG-RAN node and AMF as NAS message. There are also other entities like charging function or policy control function, which are not directly relevant for the scope of invention. The UPF is the node for establishing data connection between NG-RAN node and data network for user plane data transmission.

[0053] To support AI / ML positioning methods, in certain embodiments, a network entity in the wireless communication network may send a message to a UE, wherein the message indicates the network support of at least one AI / ML enabled feature to the UE. A network entity may also be referred to as a network node. The message or a subsequent message may further indicate specific details of the at least one AI / ML enabled feature supported to the UE. For example, a network entity, upon receiving REGISTRATION_REQUEST message from the UE, may send

[0054] REGISTRATION-ACCEPT message to the UE. The REGISTRATION_ACCEPT message may indicate AI / ML features supported in the network, which may be signalled as an information element (IE), e.g., 5GS network feature support as a field within the REGISTRATION_ACCEPT message. As an example, one of the bits for 5GS network feature support may signal the UE that an AI / ML feature or a set of AI / ML features are supported by the network.

[0055] In the following example, 5thbit in the sixth octet of the 5GS network feature support has been designed to indicate the support of AI / ML enabled features (e.g. AI / ML enabled positioning) in the network, which is shown below as “AIML” in bold text.8 7 6 5 4 3 2 1

[0056] octet 1

[0057] octet 2

[0058] octet 3

[0059] octet 4*

[0060] octet 5*

[0061] octet 6*

[0062]

[0063] Likewise, in some embodiments, the 6th, 7thor 8thbit (currently set as spare bit) may be encoded to signal, in a similar manner, one or more other features supported by the network to the UE (for example, support of AI / ML enabled beam management, AI / ML enabled mobility, etc).

[0064] Alternatively, in some embodiments, the UE may receive an indication support of Location Services (LCS) in the 5GS network feature support IE, and the support of AI / ML enabled positioning methods by the network may be indicated to the UE as a part of assistance data to the UE. In an example, the message ProvideAssistanceData may be used to indicate to the UE that the network supports labelled data (such as ground truth label) to the provided to the UE. Labelled data may be data that is annotated with tags, or labels, that can provide some extra information of the data, such as information that provides classifications of the data. The terms labelled data, label or labelling data may be used interchangeably in the present disclosure. The terms ground truth, ground truth data and ground truth label may be used interchangeably in machine learning. The terms ground truth, groundtruth data or ground truth label may refer to the correct output or label associated with a dataset, and may be used as a reference for training, evaluating, and monitoring the machine learning model.

[0065] Figures 2 and 3 are flow diagrams illustrating two different examples of using AI / ML models at UE-side or at LMF side respectively. The terms AI / ML model and AI / ML-based model may be used interchangeably in the present invention. In both figures, interactions between network entities and UE in a system for determining the location information (e.g., position, velocity) are depicted. The system may include UE, serving gNB / TRP, neighbouring gNBs / TRPs, and LMF. The ground truth exchange procedure is also indicated in both figures. The ground truth exchange procedure may be implemented by sending additional messages or lEs within the existing LPP messages (e.g. ProvideAssistanceData, RequestLocationlnformation, ProvideLocationlnformation, ProvideCapabilities, etc). Alternatively, the required information may be provided in a dedicated LPP or Radio Resource Control (RRC) message to exchange labelled data (such as ground truth data). In both figures, the procedures / functions shown as dotted boxes may comprise one or more signalling messages exchanged between one or more entities (such as UE and network entities) to perform the procedures / functions depicted within dotted boxes. The one or more messages exchanged may be one or more of the LPP messages described in these two figures or a new message.

[0066] Figure 2 is a flow diagram illustrating a UE-based AI / ML positioning method.

[0067] Although the messages are sent as a sequence of steps, the steps may be repeated, omitted or may be sent in a different sequence from that described in the figure. As it is shown in Figure 2, the LMF may interact with Next Generation Radio Access Network (NG-RAN) nodes to acquire the Downlink Positioning Reference Signal (DL-PRS) configuration from one or more network nodes, such as Transmission-Reception Points (TRPs), gNBs, NG-RAN nodes. This is depicted as step 0 (NRPPa) DL PRS CONFIGURATION INFORMATION EXCHANGE. The following steps 1-7 are explained as follows:

[0068] 1. Location Positioning Protocol (LPP) Capability Transfer: The UE and LMF exchange capability information to determine whether the UE supports a specific AI / ML method (i.e., AI / ML-based method), and what are UE’scapabilities for such support (e.g. whether the UE utilizes AI / ML-based model to obtain the UE’s position or uses AI / ML-based model to estimate one or more parameters). Further examples of UE’s capabilities for supporting AI / ML models may include PRS processing capability, and supported frequency band, etc.

[0069] NR Positioning Protocol A (NRPPa) On-Demand PRS Procedure: Since AI / ML-based method may need certain number of TRPs or certain type of measurements from one or more TRPs, the network may request on-demand PRS transmission to optimize measurement conditions, ensuring PRS is available when and where needed.

[0070] (LPP) Provide Assistance Data: The LMF may provide assistance data to the UE, including PRS configurations (e.g., periodicity, bandwidth), TRP locations, etc., to enable the UE to perform measurements and / or compute its own position.

[0071] (LPP) Request Location Information: The LMF may send a LPP message Request Location Information to the UE to trigger the UE to provide the measurement or location information. The LMF may indicate the UE to provide a UE position, measurements, a ground truth label or a combination of above. DL-PRS Measurements and UE processing

[0072] In this step, UE may perform measurements on one or more reference signals indicated in the assistance data. Furthermore, the AI / ML-based model at the UE may perform inference based on the performed measurement and / or assistance data (e.g., provided in ProvideAssistanceData message) and / or information provided in Req uestLocation Information message.

[0073] (LPP) Provide Location Information: The UE may send a message to the LMF to provide the location information as requested by the gNB. The location information may comprise UE position, measurements, a ground truth label or a combination of above. Furthermore, there may be information associating one ground truth label with data from multiple measurements, or associating multiple ground truth labels with a single measurement. This enables the network to compute the associated ground truth label at the time instant of the measurement by combining at least two ground truth labels associated with a single measurement or a group of measurements.7. In some cases, there may be a dedicated message exchanged between the LMF and the UE, wherein the LMF may provide a ground truth label to the UE. This ground truth label may be inferred from one or more information provided in ProvideLocationlnformation and / or based on network-based methods (such as AI / ML based position computation at the LMF-side or based on external ground truth labels.). In other cases, this procedure may be realized by transmitting the ground truth label to the UE by adding an IE to ProvideAssistanceData or to the Request Location Information message.

[0074] In some embodiments, the ground truth label may be collected at the same entity (e.g. the UE) utilizing the AI / ML-based model for inference operation. It is to be noticed that AI / ML inference is the process of running data into a machine learning model to calculate an output, and this process may also be referred to as putting a machine learning model into production. This may, for example, be implemented using one or more positioning methods (e.g. DL-TDOA, DL-AoD, GNSS, and so on), and the outcome of second positioning method used as a label. Likewise, for the uplink based methods, the association may be done based on one or more positioning methods computed at the network side, e.g., Downlink Time Difference Of Arrival (DL-TDOA) that is network (NW) based, multi-Round-Trip Time (multi-RTT) or independent labelling mechanisms such as camera-based methods. For camerabased methods, the network may be equipped with cameras that are locating objects and mapping the objects to ground truth locations for the UEs. For example, if a robot containing a UE is moving and the robot has a marker that can be identified by the camera in deployment scenario, the identified information relating to the marker may be provided to the network (such as an LFM or a mobile edge computing node), e.g., via Network Exposure Function (NEF), so that the LMF may associate the ground truth label with the measurements made by a network node or the UE.

[0075] As an example, a UE for data collection may be mounted on a platform, which may have some markers (like QR code, fluorescent reflector etc.), which may be detected by an independent system (e.g. a camera). A location of the UE may be obtained by the marker related information, and this information may be fed to the network (e.g., LMF) through an intermediate node enabling the independent system to provide ground truth labels for a UE. The ground truth labels from the independent system(e.g., an external system) may be associated with the measurement data by a network node (e.g. Network Data Analytics Function (NWDAF), LMF, etc.)

[0076] In some embodiments, the ground truth label may be collected by a different entity (e.g. a UE) than the inference entity (e.g. the LMF). A UE may provide labelled data (such as ground truth data), and a NG-RAN node may provide measurements, which are associated at a network entity (e.g. an LMF).

[0077] In a further embodiment, the measurement may be made by the UE and reported to the UE-DN server (or a vendor specific server reachable via the internet), or in other words an Over The Top (OTT) server. The UE-DN server may be a vendor specific server that interacts with UE over an internet connection, with contents that may be transparent to the 3GPP network. The UE-DN server may be assumed to be a server in internet, which the UE can interact with for managing the AI / ML models deployed, activated or enabled with the UE. This may interact with the 5GC network or Public Land Mobile Network (PLMN) through application function or network exposure function or through the Gateway Mobile Location Centre (GMLC). Network exposure function is a network node that separates Core Network (CN) from data network (DN), allowing the network capabilities to be exposed to entities in DN in a secure manner, and allowing external entities to influence the network’s behaviour in a controlled manner.

[0078] The UE-DN server may initiate a Mobile Terminal-Location Request (MT-LR) via the CN or the UE may initiate a Mobile Originating-Location Request (MO-LR) to acquire labels for training or monitoring of the AI / ML model’s performance.

[0079] In some embodiments, the UE may make use of supplementary service MO-LR to request assistance data for enabling AI / ML positioning in UE-based mode. In some embodiments, the UE may indicate the network a self-location initiation, e.g., by setting deferred-Mo-lrSelfLocationlnitiation. In other words, the UE initiates a network request, so that the network provides the assistance data. By indicating a selflocation initiation, the UE indicates to the network that the client for location service is at the same UE.The UE may during MO-LR request signal an indication containing a request for ground truth labels. The UE may further indicate the purpose of the request, e.g., training an AL / ML model or monitoring an AL / ML model.

[0080] The request for ground truth labels may indicate the quality of the ground truth labels. The quality may be indicated by specifying accuracy, latency and / or integrity parameters (e.g. protection limit, alert limit, target integrity risk) etc. of the estimated position of the UE. A certain quality (such as a configurable quality) needs to be maintained so that the estimated UE position can be used as ground truth labels. The quality of the ground truth labels may be indicated by a quality indicator.

[0081] In some embodiments, the UE may request assistance data from the network (i.e. , a network node such as an LFM or a gNB) using the LPP RequestAssistanceData message. In Location Services (LCS) framework, the location request may be initiated by the network as Network Induced Location Request (NI-LR), Mobile Originated Location Request (MO-LR), or Mobile Terminated Location Request (MT-LR). The requested assistance data may indicate the network to provide ground truth labels. The UE may indicate the required Quality of Service (QoS), e.g., accuracy and / or update frequency of ground truth labels. For example, the UE may request periodic assistance data delivery, where the network provides the ground truth labels at periodic intervals. The ground truth labels provided by the network at certain interval may be of certain quality. The UE (e.g. by implementation) may be able to utilize its sensors (such as motion sensors) to determine the ground truth labels for lifecycle management (LCM) of AI / ML models (e.g. monitoring, training data collection etc.).

[0082] In some embodiments, the UE may request assistance data from the network to enable high quality ground truth labels, e.g. by using Global Navigation Satellite Systems (GNSS). The position calculated in GNSS may need to be corrected for different effects, such as ionospheric propagation etc. In LPP, mechanisms to correct for the calculated position are specified as Real Time Kinematic (RTK), State Space Representation (SSR) etc. The correction data for GNSS and / or built-in sensor measurements may be used to create labels for LCM of UE-side AL / ML model. The UE may use a first positioning method for collecting the first set of data (such as measurement data). Optionally, the UE may use a second positioning method for collecting or generating a second set of data (such as labelled data). For example,the UE may receive an indication from the network to utilize location estimates (such as position, velocity, time) from the second positioning method configured for the UE as a ground truth label.

[0083] In some embodiments, the network may be equipped with additional monitoring capabilities. There may be advanced sensing capabilities or additional sensors (e.g., camera-based sensors), which enable data collection and association for training and monitoring. The network may be able to indicate the ground truth labels and associated quality of the ground truth labels to the system. The system may be any node in the network. Optionally, the UE may receive ground truth labels if the network is able to get good quality labels.

[0084] Alternatively, the network may provide the ground truth labels in requestLocationlnformation LPP message. The UE may compare the ground truth information provided in requestLocationlnformation with the computed position from the AI / ML model, and provide a location and / or a Key Performance Indicator (KPI) which may be based on error information, in a subsequent provideLocationlnformation message.

[0085] Figure 3 is a flow diagram illustrating an LMF-based AI / ML positioning method. It is shown in Figure 3 an example of sequences of events for AI / ML based positioning at the network side. In summary, the network node (e.g. the LMF) requests a certain configuration needed for determining position of the UE. The serving cell (such as a gNB) may determine a suitable Uplink Sounding Reference Signal (UL SRS) configuration and provide the configuration to the UE. Similarly, the serving cell may indicate this information to the LMF using the PositioninglnformationResponse message in NRPPa. The configuration may be periodic, semi-persistent or aperiodic. In case of configuration where a transmission is triggered after receiving a second message (e.g., a Medium Access Control- Control Element (MAC-CE)), there may be a request from the LMF to activate the provided configuration (5a) and the NG-RAN node may activate the UE to start sending the SRS. The NG-RAN node may indicate one or more TRPs to perform measurement on the signals. The request (e.g., MEASUREMENT REQUEST) may contain one or more lEs to indicate certain types of measurements (e.g. a certain number of samples, a certain measurement characteristic or reporting characteristics). The NG-RAN node may perform the measurement and report the measurement to the LMF. The LMF processing stepmay comprise performing an inference on the AI / ML model.

[0086] The ground truth exchange procedure may be a dedicated message to transmit ground truth label associated with the measurement or this may be embedded into a ProvideLocation Information message. In some embodiments, a UE may be configured with one or more positioning methods, wherein at least one method is used as a ground truth label for the network side AI / ML positioning. For example, the UE’s location reported by the UE using Assisted-GNSS (A-GNSS) may be used as a label. The association between the ground truth label and the UE transmission may be indicated.

[0087] A network entity may configure a UE to provide one or more locations at future time instances. A UE may be configured to provide its location using one or more UE-based positioning methods.

[0088] The ground truth label obtained at the UE and / or at the network may differ from the time instance at which the network and / or the UE performed measurements. To estimate the ground truth label applicable to the time instance at which the measurement is made, the network may configure the UE to:

[0089] 1) Utilise a certain positioning method configured to the UE, e.g., DL-TDOA, GNSS, Downlink Angle of Departure (DL-AoD), or sensor-based methods; or 2) Estimate ground truth at the time instance indicated by the network entity, wherein the UE may interpolate between two subsequent location estimates obtained with a UE-based positioning method configured by the network. Such interpolation may be done using Kalman filter, where the sensor reading may be combined with position estimates to obtain the ground truth label in between two ground truth labels; and / or

[0090] 3) indicate the motion characteristics or motion parameters of the UE to the network. This indication may enable the network to determine the ground truth label applicable to the current time instance based on the information reported by the UE.

[0091] The motion characteristics or motion parameters may include any one of the following:

[0092] - Linear displacement reported by the UE, e.g., based on UE-sensors like accelerometer,- Vertical displacement reported by the UE, e.g., based on barometer reading,

[0093] - Steps detected, e.g., in case of pedestrian motion,

[0094] - Vehicular motion, e.g., accelerating, decelerating, etc.,

[0095] - Raw or processed sensor reading values, e.g., inertial and / or barometric sensors x-Acc, y-Acc, z-Acc, x-Gyr, x-Gyr, y-Gyr, z-Gyr, delta-Pressure, - Orientation vector or the UE, e.g., as roll, pitch, yaw or quaternions.

[0096] In some embodiments, the UE may utilize one or more UE-based methods to determine a location, e.g., a hybrid between GNSS and sensors. The network may indicate the information that may be combined to provide the hybrid results.

[0097] In some embodiments, the UE may indicate to the network, the source of ground truth label and confidence in the ground truth labels.

[0098] In the following description, various examples of how the ground truth labels may be obtained in different scenarios are provided.

[0099] Scenario 1: UE-based AI / ML positioning

[0100] Case 1: LMF provides ground truth labels for UE-side model monitoring, with UE utilizing quality metrics provided by a network node (e.g., LMF) for monitoring.

[0101] Case 2: LMF provides ground truth labels for UE-side model monitoring, with UE interpolating the ground truth labels with information available at UE (e.g. on-board sensors).

[0102] Case 3: UE provides inference labels with LMF performing the monitoring functionality. Optionally, LMF may use its ground truth labels, and compare those ground truth labels against the UE provided inference labels.

[0103] Case 4: UE provides inference labels and a movement profile to the LMF to enable monitoring. Optionally, LMF may use its ground truth labels, and the movement profile of the UE to interpolate and associate. The UE may indicate that it is stationary, and this information may be used as part of the movement profile of the UE. Optionally, the movement profile at LMF may also be inferred by the LMF based on one or more measurements (e.g., carrier phase) made by a network entity (e.g., a gNB).Scenario 2: LMF-based AI / ML positioning

[0104] Case 1: UE provides the ground truth from alternative mechanism, e.g., GNSS or Radio Access Technology (RAT)-based from the application layer (e.g. QR codes). Case 2: UE provides the movement profile to the LMF to determine validity of labels at future instants. Movement profiles may be signalled as enumerations {Stationary, uniformMotionVehicle, acceleratingVehicle, pedestrian, ...}

[0105] Case 3: UE provides a displacement that is already in LPP. Adding orientation -> displacement and orientation information at LMF allows it to monitor against the ground truth. The displacement may be a linear displacement such as a vertical or horizontal displacement.

[0106] For UE-based and / or LMF-based AI / ML positioning methods, it is important to have verifiable data for training or performance monitoring of machine learning models. In the description below, various ways of data association have been provided to associate measurement data with labeled data to improve the accuracy and performance of the AI / ML positioning methods.

[0107] In some embodiments, a first set of data, which may also be denoted as Part A data or Part A, is to be associated with a second set of data, which may be denoted as Part B data or Part B. At least part of the first set of data (such as measurement data) may be used as input data to one or more machine learning models, and the accuracy or performance of the one or more machine learning models may be verified or validated by the second set of data, which may include labelled data.

[0108] In some embodiments, the association between the first set of data and the second set of data is based on time stamps or identifiers of the wireless device (e.g., a UE). However, variations in measurement conditions, such as differing antenna configurations, resource allocations, or spatial parameters, may introduce discrepancies between the first set of data and the second set of data. The terms data association and data pairing may be used interchangeably in the present invention.

[0109] In some embodiments, the data association is based on the source of the measurement data. Some methods may assume the first set of data (i.e., Part A) andthe second set of data (i.e., Part B) originate from the same entity, while others address cases where different entities generate these datasets.

[0110] While time-stamp-based association / pairing is a common approach, it does not fully account for mismatches arising from changes in measurement conditions between the data collection of Part A and Part B. Further, measurement configurations may differ across TRPs, SRS / PRS resources, or frequency layers, impacting data consistency. Addressing these variations is essential for improving ML model accuracy and ensuring reliable data association in training datasets. The different embodiments for data association will be described as below.

[0111] Example of signalling messages for acquiring ground truth label (Part B) Examples for direct AI / ML positioning at the UE side:

[0112] 1) Capabilities exchange procedure:

[0113] A UE may provide its capabilities to the network by providing its capabilities by transmitting the LPP message ProvideCapabilities to the network node (LMF) either on its own initiative or in response to a Requestcapabilities message from the network node.

[0114] The Requestcapabilities message may be enhanced with a new IE NR-AIML- RequestCapabilities-r19, added to Requestcapabilities message to UE capabilities with respect to AIML based positioning methods.

[0115] [[

[0116] nr-AIML-RequestCapabilities-r19 NR-AIML-RequestCapabilities-r19 OPTIONAL - Need ON

[0117] ]]

[0118]

[0119] The IE NR-AIML-RequestCapabilities-rl9 may further indicate the UE to report on its specific UE capabilities, such as its capability to provide ground truth label using one or more positioning methods specified in 3GPP or based on alternate sources (e.g. by scanning QR code, detecting Bluetooth beacons, RFID (radio frequency identification) tags, Ambient loT devices) etc. Furthermore, the network may inquire the UE whether it has its capabilities to support finetuning of AI / ML model on the device.

[0120] The ProvideCapabilities message may be provided unsolicited or in response to the Requestcapabilities message sent by the network node (LMF).

[0121] The ProvideCapabilities message may be enhanced with a new IE NR-AIML- ProvideCapabilities-rl9, added to ProvideCapabilities message from UE tonetwork node (LMF) to provide UE capabilities with respect to AIML based positioning methods.

[0122] The NR-AIML-ProvideCapabilities-rl9 IE shall contain a sequence of one or more capabilities that are relevant for AIML-based positioning. The ASN.l snippet shows an excerpt where the UE reports which AIML modes it supports, UE-based direct positioning, UE-assisted LMF-based AI / ML positioning, etc. Further, it shows what type of GTL acquisition is supported by UE. In the example below, the signalling shows an example where the UE can indicate the network whether or not it supports UE-based RAT (radio access technology) dependent positioning methods configured to the UE as ground truth label, GNSS as a label, hybrid method deployed at the UE (which may be implementation specific as to which of the methods are used as hybrid - e.g. sensor fusion using Inertial sensors with updates from UE-based or network-based UE location), ground truth obtained by scanning QR (quick-response) code, RFID detection, BT (Bluetooth)-beacons, use of ambient loT to read our stored location at A-IoT (Ambient- loT) device or detect a configured landmark.

[0123]

[0124] 2) Assistance Data from the network:

[0125] A network node (e,g LMF) may indicate whether a network node can provide the ground truth labels for the UE, and for which purposes (monitoring or data collection for training).

[0126] Option 1: The network may have advanced AI / ML based methods optimized for the given area, and the network can make these labels available to the UE.

[0127] ■=> Periodic deliveiy of ground truth labels by periodic delivery of assistance data within a positioning session.

[0128] Option 2: the network may provide assistance data, enabling the UE to detect secondary sources of ground truth (e.g. QR codes, landmarks, range to other UE using sidelink, RFID tags etc), subject to capabilities of the UE.

[0129] Option 3: There may be option for finetuning (UE-capability), and the NW may provide data forfinetuning correspondingto this environment.A UE may request assistance data from the network transmitting the LPP message RequestAssistanceData to the network node (LMF). The network node (LMF) sends ProvideAssistanceData to the UE either on its own or in response to a RequestAssistanceData message from the network node.

[0130] The existing RequestAssistanceData message may be enhanced with a new IE NR- AIML - RequestAssistanceData-rl9, added to RequestAssistanceData message to provide assistance data with respect to AIML based positioning method.

[0131] [[

[0132] nr-AIML-RequestAssistanceData-rl9 NR-AIML-RequestAssistanceData-rl9

[0133] OPTIONAL - Need ON

[0134] The ASN.l snippet below shows an example of the request from UE to the network node (LMF) to provide assistance data of different types (e.g. DL-PRS, additional information for computing position using AIML direct positioning, assistance data for determining ground truth label by UE with the help from network, or requesting ground truth label from the network node (LMF) itself.

[0135]

[0136] The existing ProvideAssistanceData message may be enhanced with a new IE NR- AIML - ProvideAssistanceData-rl9, added to ProvideAssistanceData message to provide assistance data with respect to AIML based positioning method.

[0137] nr-AIML-ProvideAssistanceData-rl9 NR-AIML-ProvideAssistanceData-rl9 OPTIONAL - Need ON

[0138] An example of how the assistance information may be structured is depicted. The examples are supposed to indicate one way of providing AD (assistance data) to utilize a certain type of reference.

[0139] -- AS1I1START

[0140]

[0141]

[0142] In the above, the IE Locationcoordinates is one of the GAD (geometric area descriptor) shapes defined in TS 23.032. The identifier identifies the QR code as registered in the system. There may be a fixed value (e.g. 0) which indicates that the UE is reportingthe QR code detected, which was not indicated in the AD. The checksum ensures that the data read by the UE was not tampered.

[0143]

[0144] The field rf id- Identifier represents the unique identifier of the RFID tag (such as UID (unique identifier) or EPC (electronic product code)). It is used to uniquely identify a specific RFID tag used for GTL (ground truth label) in the positioning system provisioned. Likewise the field rf id-FrequencyBand-rl 9 specifies the frequency band on which the RFID tag operates. This helps define the communication range and technology used for the RFID tag, such as LF (125-134 kHz), HF (13.56 MHz), or UHF (860-960 MHz). The field rf id-ExpectedLocationCoordinates-rl 9 represents the expected geographic location (latitude, longitude, and optionally altitude) where the RFID tag is anticipated to be located. This helps in guidingthe detection process by indicating the approximate area for tag detection. Finally the field rf id-DetectionRadius-r!9defines the detection radius around the expected location of the RFID tag. This radius provides an approximate distance within which the tag is expected to be detected. The field rf id-DetectionThreshold-rl9 indicates the proximity detection threshold, after which the indicated RFID can be used as a GTL.

[0145]

[0146] The field bt-Beacon- Identifier is the unique identifier of the Bluetooth beacon, typically using UUID (universally unique identifier), major, and minor values, or a similar identifier to uniquely identify the beacon. Likewise, the field bt-ExpectedLocationCoordinates-rl9 indicates the expected geographic location (latitude, longitude, and optionally altitude) of the Bluetooth beacon. The field bt-Detect ionRadius-r 19 defines the detection radius around the expected location. The field bt-DetectionRadius-rl9 together with bt-ExpectedLocationCoordinates-rl9 defines the approximate range within which the beacon can be detected. The field bt-SignalStrengthThreshold-rl9 represents the minimum RSSI (signal strength threshold) required for the beacon to be detected. If the signal strength is below this threshold, it may not be considered for GTL reporting.

[0147]

[0148] The Ambient loT detection information may be similar to the BT beacon detection.

[0149] In case camera is used to capture the surrounding and use this for ground truth information, and the UE indicates its capability for such support, the UE may receive assistance data from the network to determine its location based on detection of landmark or features.

[0150]

[0151]

[0152] The camera based landmark detection may be used together with QR-based GTL to cross validate whether the QR information and the location of QR matches the expected location. This enables tampering to be reduced, as the network or the UE can validate the information of QR with the location of QR as detected by the UE as opposed to as deployed by the network.

[0153] The IE NR-PeriodicGTL-r19 is an example of GTL delivered by the network periodically to enable the UE to perform LCM (lifecycle management) operations (e.g. collecting data and labels or monitoring its model). In some cases, the network may have advanced capabilities (such as a more accurate AI / ML based positioning method deployed at LMF and / or NWDAF nodes). Furthermore, the network may have additional information which improves UE location. The Network node may provide GTL to the UE to fine-tune its model or to monitor its model or simply to collect labels for data collection.

[0154] Likewise, there may also be a location corresponding to the time when the UE made a measurement. The GTL request corresponding to the time when the UE made the measurement or is expected to make measurement can be conveyed (for example, in RequestAssistanceData message)

[0155] 3) Monitoring of UE model (RequestLocationlnformation, ProvideLocationlnformation)

[0156] Option 1 : The NW provides the GTL, and the UE computes difference to the GTL (or KPI), and indicates the NW to reconfigure the UE with a different method if AI / ML method is not performing.

[0157] Option 2: The UE reports the inference data, and the NW (with the ground truth available at the NW side), indicates the model performance statistics and / or switches to a different positioning method.Option 3: The NW provides GTL at multiple time instants, and the UE interpolates the timestamp to perform monitoring.

[0158] Option 4: The NW provides assistance data to the UE to obtain the ground truth label. The UE utilizes the ground truth label at one or more time instants and / or its sensor data to create a new GTL at the measurement instant. The UE may use the ground truth label or indicate the GTL to the network.

[0159] Option 5: The NW provides assistance data to enable the UE to acquire GTL. The UE obtains its GTL using the method and / or its own sensor data. The UE then computes the KPIs and provides the KPI to the network.

[0160] The message Req uestLocation Information can be extended to introduce procedures to retrieve information to enable monitoring of UE model or to collect data (e.g. for crowdsourcing of data) for training (e.g. at NWDAF, network data analytics function). As an example, an IE NR-AIML- RequestLocationlnformation-r19 can be signalled to the UE, indicating the information requested by the network.

[0161]

[0162] The ProvideLocationlnformation message can be extended to introduce NR-AIML- ProvideLocationlnf ormation-rl 9

[0163]

[0164]

[0165] The ProvideLocationlnformation message may contain KPI computed by the UE, or GTL computed by the UE if ground truth label or KPI (key performance indicator) is requested by the UE. The source of GTL may be indicated. The source of GTL may be network provided, network provided with sensor fusion, UE acquired GTL, RAT dependent, RAT independent positioning methods, Hybrid. Hybrid may indicate which methods have been used to determine the GTL labels.

[0166] The ProvideLocationlnformation may link the GTL with measurement time. This may be further linked to sensor measurement. The GTL at GTL acquisition time (t1 ) or times (tn), with the time instance where the transmission or reception is made (as configured) (tm1 ), with sensor information providing information linking the GTL at transmission or reception with GTL acquisition time.

[0167] Alternatively, the UE computes internally the GTL at the configured time instance based on GTL acquired at one or more reference times.

[0168]

[0169]

[0170] The UE may raise an error if the GTL cannot be provided, despite the UE having the capability due to several reasons, such as overheating, low battery, high battery drainage rate, etc. The UE can indicate in errorthe reason and / or it may indicate how long the UE is unable to acquire or provide GTL. The network may provide the UE with NW generated GTL if it has or the network may swich the positioning method (for example, it cannot monitor the model performance in absence of GTL).

[0171] In some examples, the UE may be able to take a picture with its onboard camera but may not be able to process it to obtain the ground truth label. The UE may provide a picture or a series of pictures with timestamp to a server (UE server on Data network or to a network node), where the GTL may be acquired by a network node and communicated to the UE or used as GTL at the network side. In some examples, the UE server on data network may be able to provide GTL to network node (e.g. via AF (application function) or NEF (network exposure function) ) correspondingto a certain time, or vice versa.In some examples, mobile edge computing (MEC) may be deployed to locate UE by camera based processing (e.g. by use of visual markers or image processing) and indicate GTL to a network node, where the network node associates the information provided by the edge computing node with the measurement performed by the UE or network node or provides the GTL to the UE. The entity receiving the GTL may use it for data collection for training, or for monitoring of models.

[0172] Examples for LMF-sided Direct Positioning (Case 3b)

[0173] Most of the signalling discussed for GTL acquisition can be reused here. Some further enhancements that may be performed are discussed below:

[0174] 1) Capabilities of UE: (Request / Provide UE capabilities)

[0175] An IE (e.g., NR-AIML-RequestCapabilities-r19) may indicate the UE to provide its capabilities or may further indicate the UE to report on its specific UE capabilities, such as its capability to provide ground truth label using one or more positioning methods specified in 3GPP, using advanced processing (such as use of Kalman Filters), or based on alternate sources (e.g. by scanning QR code, detecting Bluetooth beacons, RFID tags, Ambient loT devices) etc.

[0176] The UE sends an IE NR-AIML-ProvideCapabilities-r19 may further indicate the UE to report on its specific UE capabilities, such as its capability to provide ground truth label using one or more positioning methods specified in 3GPP, using advanced processing (such as use of Kalman Filters), or based on alternate sources (e.g. by scanning QR code, detecting Bluetooth beacons, RFID tags, Ambient loT devices) etc.

[0177] Similar details as detailed out in UE-based direct positioning may be used.

[0178] 2) Assistance Data from the network: (Request / Provide Assistance Data) Network may provide assistance data to acquire ground truth labels, enabling the UE to detect secondary sources of ground truth (e.g. QR codes, landmarks, range to SL-UE, RFID tags etc), subject to capabilities of the UE.

[0179] l ' lThe procedure would be similar to the one for direct UE positioning.

[0180] ) Req uestLocation Information

[0181] In case of LMF-based AI / ML positioning, the network may request the UE provides ground truth label corresponding to one or more of the following: (i) Time the UE acquired the ground truth label

[0182] (ii) Time the UE transmitted at least one SRS (sounding reference signal) corresponding to SRS resource from the SRS resource set.

[0183] (iii) Preconfigured time configured by the network.

[0184] Likewise, the network may further configure the UE to provide, one or more of the following:

[0185] (i) Movement profile of the UE

[0186] (ii) Sensor information

[0187] (iii) Information associating the ground truth label at first time instance to an event (e.g. SRS transmission) at a second time instance. For example, the displacement or displacement vector of the UE since the GTL was last measured by the UE. Alternatively, Sensor information (Acc-x, Acc- y, Acc-z, Gyr-x, Gyr-y, Gyr-z, Pressure) since the last GTL may be requested.

[0188] a. NW requests the UE to provide ground truth label corresponding to i. SRS transmission time

[0189] ii. Preconfigured time

[0190] b. Furthermore, the NW may request the UE to provide:

[0191] i. Movement profile

[0192] ii. Sensor information

[0193] ) Provide Location Information

[0194] UE Provides label, or information as requested by the LMF.

[0195] ■=> The label and / or motion profile is used by NW to determine label for monitoring or training.A UE may be configured to collect, or request assistance data to acquire ground truth labels and / or report data to the network, if certain events are triggered. The events may be:

[0196] • UE travelling in a preconfigured track, where the NW intends to collect data.

[0197] • UE entering a certain area, where the area may be indicated as a RAN area, tracking area (TA), registration area, positioning service area, cell, group of cells, or a certain geometric shape (e.g. ellipsoid with a certain center(s)Zfoci), certain distance from a certain reference area, certain orientation from a certain location or am area.

[0198] • UE detecting certain signals above a certain threshold.

[0199] • UE performance degrading beyond a certain threshold.

[0200] Collection of the first set of data by a measurement entity

[0201] In some embodiments, the first set of data comprises measurement data. In some embodiments, the measurement data comprises channel measurement data.

[0202] Channel measurement data may be referred to as channel measurement.

[0203] In some embodiments, the first set of data comprises channel measurement data, an associated quality indicator of the channel measurement, and an associated time stamp of the channel measurement data.

[0204] In some embodiments, the first set of data is collected by a measurement entity. The measurement entity may be a wireless device or a network node in a wireless communications network. The measurement entity may be configured to report, within a single channel measurement data report, a plurality of channel measurements, corresponding quality indicators, and associated time stamps.

[0205] In some embodiments, the measurement entity may be configured to report, within each channel measurement data report, a channel measurement, a corresponding quality indicator, and an associated timestamp.

[0206] In some embodiments, the plurality of channel measurements correspond to measurements collected across multiple resources, transmission reception points (TRPs), beams, bandwidth parts, or frequency layers. In some embodiments, theplurality of channel measurements corresponds to multiple measurements of the same resource, such as measurements obtained through receiver diversity, different measurement instances, or different frequencies in a carrier aggregation scenario. In some embodiments, data may be collected on at least one of a first resource set associated with a first TRP and a second resource set associated with a second TRP, wherein the first resource set and the second resource set may be configured on different frequency layers or may be allocated to different beam configurations.

[0207] In some embodiments, the measurement entity may be configured to include one or more identifiers (e.g., resource ID, resource set ID, or indicator on the reference signal (RS) or channel) when reporting channel measurement data. The reported measurement data is associated with one or more resource IDs, allowing differentiation based on resource allocation.

[0208] In some embodiments, the measurement entity may be configured to collect radio measurements from one or more reference signals transmitted on different resources, resource sets, TRPs, or frequency layers. The measurements may be performed based on Uplink reference signals, such as SRS, Random-Access Channel (RACH), sidelink reference signals, or downlink reference signals, such as PRS, Channel State Information Reference Signa (CSI-RS), Tracking Reference Signal (TRS) or Synchronization Signal Block (SSB).

[0209] In one example, the first set of data may comprise beam-level signal-strength measurements, such as L1-RSRP measurements, performed by the wireless device on a configured set of reference signal resources (e.g., CSI-RS resources or SS / PBCH block resources) associated with a first resource setting. These beam-level measurements may serve as input to an AI / ML model for predicting future beam conditions, including the identification of beams expected to be strongest at one or more future time instances.

[0210] In some embodiments, the one or more resources may be associated with one or more resource sets, where each resource set corresponds to a predefined grouping of resources configured for measurement collection. The measurement entity may report measurement data mapped to specific resource sets.Further, according to some embodiments, the network may dynamically configure resources and / or resource sets based on the measurement entity's reporting capabilities.

[0211] In some embodiments, the measurement entity may be configured to receive assistance data, wherein the assistance data comprises information on the one or more channel measurement. The assistance data for the channel measurement data may comprise information on at least one measurement resource.

[0212] Collection of the second set of data

[0213] In some embodiments, the second set of data comprises labelled data such as ground truth label data. Ground truth label data may be collected using a reference positioning technique different from that applied for collecting the channel measurement data, wherein the reference positioning technique includes at least one of GNSS-based positioning, multi-TRP multilateration, landmarks, configured position or sensor fusion-based localization. In some embodiments, the ground truth label may be collected or generated by a labelling entity. The labelling entity may be the same entity as the measurement entity or a different entity (e.g., a UE, a network node, a gNB, or an external localization system). The conditions under which the measurements and ground truth labels are obtained may differ due to factors such as channel dynamics, mobility, and measurement availability across frequency layers or TRPs.

[0214] In one implementation, the second set of data may comprise predicted beam-related quantities generated by the wireless device using an AI / ML model. For example, the second set of data may include predicted beam identifiers indicating which reference signal resources or synchronization signal blocks are predicted to have the best signal quality at one or more future time instances, and optionally predicted signalstrength values associated with the predicted beam identifiers.

[0215] As a further example, the second set of data may comprise predicted precoding matrix indicators (PM Is) generated by the wireless device using an AI / ML model at a second time stamp. The predicted PM I indicates predicted precoder matrices for one or more future time instances, derived from channel measurements on CSI-RS resources collected at earlier time instances.Potential mismatches between the first set of data and the second set of data In some embodiments, the measurement conditions and labeling conditions may differ in one or more of the following aspects:

[0216] • Antenna Configuration:

[0217] The wireless device (such as a UE) may perform channel measurements using a first antenna (or beam or rx chain), while the ground truth label may be derived from data obtained using a different antenna configuration (e.g., another panel or a full-array combination).

[0218] • Reference Signal Resources and Frequency Layers:

[0219] The wireless device (such as a UE) may measure RS from a first set of TRPs at a first frequency layer, whereas the ground truth label is generated at a different frequency layer. Similarly, the wireless device (such as a UE) may measure RS from a first set of TRPs, whereas the ground truth label may be generated using RS from a second set of TRPs at the same or at a different frequency layer.

[0220] • Timing and Environmental Conditions:

[0221] The channel measurements may be performed under a first set of mobility or propagation conditions (such as a slow-fading channel at a specific time instance), while the labeled position as labelled data may be derived from a second set of conditions with dynamic multipath changes.

[0222] Optionally, the measurement and labeling processes occur at different time stamps, where the labeling entity estimates the position using a past or future observation window relative to the measurement instant.

[0223] Figure 4 is a diagram illustrating an example where the first set of data (such as the measurement data) and the second set of data (such as the labelling data) are collected at different time points. Time points may be referred to as time instances. In this figure, an example is illustrated where a UE moves along a track, and measurement events correspond to UL reception at TRP(s), SL reception, or DL reception by the UE. Independent label points at different time points are generated, which may be provided by a different entity or by the same entity at different timestamps. The number of measurement events and label points may differ which introduces challenges in data association if a validation indication is not present. For example, at measurement times tA1 and tB 1 , the time stamps are very close, and due to user mobility or orientation change, the channel condition exhibits high spatial correlation between these two time points. However, at measurement time tA2, the closest timestamp label at tB2 does not provide a valid association to the measurement event due to environmental factors or positioning inconsistencies. At measurement times tA3, tB3, and tB4, channel variations, user mobility, or trajectory data may allow for a predicted label to be associated with tA3. Additionally, at measurement time tA4, a valid association may be established with label time tB5, while at measurement time tA5, no valid label may be present.

[0224] The measurement entity (e.g. UE, TRP) may identify, for example based on history measurements, the changes in measurement information that may be related to a window or a range. The measurement entity may indicate or utilize a validity window or a valid range, wherein the validity window or range may be determined by a validation entity. The validity window may be referred to as a validity range, a validity time window, or a validation time window.

[0225] In some embodiments, the validation entity may be the same as the measurement entity, which determines a measurement validity window or range, wherein the measurement validity window defines a time period, frequency range, or spatial consistency condition over which the measurement data is expected to remain valid. The measurement validity window may be dynamically adjusted based on the rate of channel variation, user mobility state, and environmental conditions.

[0226] Similarly, a label validity window may be introduced to indicate the time range, spatial boundary, or measurement conditions under which a label remains applicable. The label validity window may be configured based on positioning accuracy, network-assisted corrections, or mobility-based label prediction. A validity window may be a measurement validity window, a label validity window, or a combination of the above.

[0227] [Time Correlation Mechanism]

[0228] In some embodiments, a Time Correlation Mechanism enables alignment between measurement data (of Part A) and labeling data (of Part B). Labelling data may bereferred to as labelled data. The time correlation mechanism may operate based on at least one of the following:

[0229] • Direct Time Matching: the system aligns Part A and Part B data points by comparing time stamps,

[0230] • Predictive Time Association: if time stamps do not align precisely, a predictive algorithm interpolates or extrapolates the labeled data based on historical movement patterns or propagation conditions, or

[0231] • Dynamic Time Window Association: a flexible time window is applied to determine whether a label is valid for a given measurement. In some examples, the dynamic time window association may be considered as a function where the matching degrades toward the window edges.

[0232] [validation time window indication] In some embodiments, the measurement entity provides information on the validation time window, wherein the measurement validity window may be dynamically adjusted based on at least one of the following:

[0233] • Channel variability

[0234] • Network conditions

[0235] • label and / or measurement accuracy

[0236] Figure 5 illustrates an example where a validity window is associated with a label data. A UE performs multiple channel measurements reporting values such as RSSI, ToA, CIR, path delay or CSI values. The network provides labels (labell, Iabel2, labels, and Iabel4) at different time points (or time instants or time stamps) tA1 , tA2, tA3 and tA4. At tA1 for example, the times tamps match closely, allowing direct association of the measurement data and the labelled data. If the validity window of tA1 covers one or more measurements, the label remains applicable (or valid). At tA2, the system detects rapid mobility, thus reducing the validity window for Iabel2 at tA2. At the time point (or time instant or time stamp) tA2, no label within the validity window can be associated with the measurements. It is to be noticed that the terms time point, time stamp, timestamp, or time instant may be used interchangeably. Some embodiments of the Label generating entity are the following:

[0237] UE: The UE may generate labels based on GNSS positioning, UE-based positioning, sensor fusion, or external positioning sources.• Positioning Reference Unit (PRU): A dedicated PRU may generate reference positions.

[0238] • Location Management Function (LMF): The LMF may provide labels based on network-assisted, UE-assisted Methods, GNSS, multilateration, or hybrid positioning methods.

[0239] • External (Over-the-Top) Sources: Labels may originate from third-party services such as different positioning systems or non-3GPP positioning services.

[0240] The validity time window may be expressed as at least one of the following:

[0241] • A time-based window, where a label remains valid within a predefined time range (e.g., ±T seconds around the measurement event).

[0242] • A spatial zone-based window, where the label applies within a specific geographic area, adjusting based on estimated UE trajectory.

[0243] • A probabilistic distribution (e.g., Gaussian-like model), where the label’s confidence decays over time or distance, providing a weighted probability of association.

[0244] The label validity range may be expressed as expectedLabel-range, representing the uncertainty in the reference label's applicability over time.

[0245] • the validity range may be realized as a Gaussian-like function, a truncated Gaussian function, or a fixed time window centered around a reference time (T_REF), with a confidence level defining the probability of association within the range.

[0246] • In some implementations, the validity range may be a fixed window defined as [-X x Ts, T_REF, +Y x Ts] , where X and Y can be equal or different values, representing the permissible deviation from T_REF.

[0247] • The interpretation of the validity range can be left to implementation-specific configurations, allowing for either a probabilistic confidence-based association or a hard decision approach where data points outside the range are deemed invalid.

[0248] • The time resolution may be expressed in different units, such as Ts (symbol duration), Tc (chip duration), milliseconds, or microseconds, based on network deployment scenarios.The validity window for a measurement, expectedMeasurement-range, defines the expected time range in which a measurement remains relevant before it is considered outdated due to channel variations, mobility, or network dynamics.

[0249] • The validity window may be represented as a Gaussian-like or truncated Gaussian function, or alternatively, as a fixed validity window defined as [-X x Ts, T_REF, +Y x Ts], where X and Y can be set dynamically by the device or Network node. The maximum values forX and Y can be set predefined in the report configuration or measurement configuration.

[0250] • The validation can be one example handled as a soft association with confidence weighting or alternatively as a hard binary decision (valid / invalid) based on the measurement’s placement within the window.

[0251] • The window is centered around the measurement event, with an associated confidence interval determining how likely a measurement remains valid within a given time frame.

[0252] • The reporting range may define different multi-scale resolution, where different timing units (Ts, Tc, milliseconds, or microseconds) are applied dynamically based on network conditions.

[0253] The expectedLabel-Uncertainty and / or expectedMeasurement-Uncertainty define a search window for associating Part A (measurement) with Part B (label). If the uncertainty of either Part A or Part B exceeds a predefined threshold, the association may be discarded or re-weighted in machine learning training or network optimization tasks.

[0254] In one scenario, the AI / ML model resides at the UE, and the UE is responsible for collecting Part A (measurements), while the LMF provides Part B (labels). The UE collects channel measurements (e.g., RSRP, CSI, ToA, AoA) from TRPs or neighboring UEs. The UE may optionally report Part A to the LMF via uplink signaling (e.g., NRPPa / LPP), including measurement data, timestamps, and quality indicators. The LMF determines Part B, generating a ground truth label. A validity range or uncertainty window is introduced to ensure accurate measurement-label association, allowing the UE to correctly align Part A and Part B for AI / ML model training.In a second scenario, the AI / ML model resides at the LMF, while the gNB collects Part A (measurements), and Part B is provided by either the LMF or the UE. The gNB collects network-side measurements and transmits Part A to the LMF. Part B (label) is generated by the LMF. The LMF must ensure alignment between measurements (Part A) and labels (Part B), particularly when the UE provides labels asynchronously. A validity range or uncertainty window is applied at the LMF to resolve temporal discrepancies and network delays, preventing mismatches in AI / ML model training. In a third scenario, the AI / ML model resides at the LMF, while the UE collects Part A (measurements). The UE collects Part A and reports it to the LMF. The LMF processes Part A, applies AI / ML algorithms, and generates a network-assisted label (Part B). A validity window for measurements is required to determine how long a UE-collected measurement remains relevant for AI / ML processing at the LMF. The LMF integrates Part A and Part B using a search window approach to ensure associations across different reporting latencies and UE mobility conditions.

[0255] In an example, the measurement data corresponds to Layer 1 Reference Signal Received Power (L1-RSRP) measurements. The UE collects this measurement data based on a first set of reference signal resources, which may be configured via a first Resource Setting (e.g., resourcesForChannelMeasurement comprising CSI-RS or SS / PBCH block resources). Furthermore, the labelled data, which may comprise predicted channel state quantities such as Predicted CRI (P-CRI), Predicted SSBRI (P-SSBRI), or Predicted L1-RSRP (P-L1-RSRP), may be generated based on a second set of reference signal resources. This second set of reference signal resources may be configured via a distinct second Resource Setting (e.g., resourcesForSetA-r19). By utilizing two distinct resource settings, the network can independently control the measurement inputs and the predicted (labelled) outputs. In a further example, the validity range of the association between the measurement data and labelled data is determined temporally. In one example, this is implemented as a function of a reference time, which the UE determines as the slot containing the latest transmission occasion of the reference signals (e.g., CSI-RS or SSB resources) utilized for channel measurement. The boundaries of this temporal validity range may be established by one or more configured time offsets. For instance, an initial boundary of this temporal validity range may be defined by a configured time gap (TimeGap-r19) between the reference time and a prediction time instance. Furthermore, a terminalboundary of this temporal validity range may require that an associated report (e.g., a first CSI Reporting Setting) has a minimal slot offset no larger than a predefined threshold (for example 64 slots) from the slot of the measurement transmission occasion. If the time offset exceeds this boundary, the association falls outside the validity range.

[0256] In some embodiments, the validation of the data association may be performed based on a consistency check between the channel measurement data and the ground truth label data, wherein the consistency check includes verifying whether the reference signal resource configuration of the collected channel measurement data corresponds within the spatial configuration of the ground truth label data. The consistency check may include spatial consistency and / or temporal consistency. The method may include dynamically adapting the pairing validation mechanism based on network conditions, wherein network conditions include variations in propagation delay, mobility state of the user equipment (UE), or interference levels in the allocated frequency bands.

[0257] In some embodiments, the measurement entity or validation entity may determine a validation window, range, or validity period for a label using the current and / or previous measurement history. This mechanism enables adaptive validation, improving the reliability of measurement-label pairing.

[0258] The validation window may be derived based on at least one of the following:

[0259] • Temporal consistency: Evaluating timestamp correlations between past and current measurements to determine the expected validity duration of a label.

[0260] • Spatial correlation: Analyzing historical positioning estimates to define the expected region where a label remains applicable.

[0261] • Channel conditions: Using past signal measurements (e.g., Doppler shift, received power variations) to estimate when a label becomes outdated. The measurement entity can see the interval of time of the measurement (for example CIR changes). This may be indirectly related to the channel changes and user movement and type of movement (cycling, driving, walking, standing etc.). The measurement entity may have an AI / ML model to predict the validation based on such events.Measurement confidence decay: Assigning a probability-based validity function (e.g., Gaussian decay model) where a label’s reliability decreases over time or distance.

[0262] According to an embodiment of the disclosure, a base station includes a transceiver and a controller, the controller being configured to transmit, to a terminal, measurement configuration information for at least one resource, and / or resource set; and to receive, from the terminal, measurement reports containing channel measurement data, wherein the channel measurement data may be collected under a first measurement condition, and ground truth label data may be collected under a second measurement condition.

[0263] According to an embodiment of the disclosure, a terminal includes a transceiver and a controller, the controller being configured to receive, from a base station or the network, configuration information for channel measurements; to collect channel measurement data based on the received configuration information; and to validate a pairing of the collected channel measurement data with ground truth label data using at least one of timestamp correlation, spatial consistency verification, or frequency domain matching.

[0264] In some embodiments, a method for data collection and validation in machine learning-based positioning is provided, wherein a pairing between the collected channel measurement data and the ground truth label data is determined based on a preconfigured threshold of validity, the threshold being dynamically adjusted based on environmental conditions or measurement discrepancies.

[0265] Further, according to an embodiment of the disclosure, a method is provided wherein different types of measurement and labeling techniques are adaptively supported based on the availability of a positioning reference, wherein the positioning reference may be obtained using at least one of multi-TRP observations, GNSS corrections, or a network-assisted positioning scheme.

[0266] Further, according to an embodiment of the disclosure, it is possible to provide a method by a terminal for efficiently identifying valid measurement-label pairs based on resource-specific configurations, ensuring accurate machine learning dataset generation for wireless positioning applications.[Adaptive Triggering of Future Measurements]

[0267] According to some embodiments, the collection of future measurements may be aligned with the availability of validity window associated with a label, in particular a time-based window or a spatial zone-based window. For instance, the density of future measurements may be increased in case a label having a large validity window is available. Optionally, the data may sometimes be more easily to be associated / paired since the validity window is large.

[0268] Depending on the scenario, the entities as measurement entity, label entity may differ as seen below:

[0269]

[0270] In one embodiment, the measurement entity may be a base station (gNB or TRP) that collects uplink (UL) or sidelink (SL) measurements from a UE, or DL from other gNB, or NTN, while the label generating entity is a network node that determines the ground truth label using for example network-assisted and / or GNSS-based

[0271] positioning.In another embodiment, the measurement entity may be a wireless device (such as a UE) that collects downlink (DL) measurements from multiple TRPs or SL or additional signals from other UEs, while the network acts as the label-generating entity.

[0272] Throughout this disclosure, the term 'network' is used broadly to encompass various network entities that participate in measurement collection, validation, or label generation. Depending on the specific use case and network architecture, the term 'network' may refer to:

[0273] • Location Management Function (LMF)

[0274] • Network Data Analytics Function (NWDAF)

[0275] • Core Network Entities: Including the AMF, SMF, or other control-plane functions involved in managing measurements and analytics

[0276] . gNB

[0277] • RAN-Level Components: Including TRPs, anchor UE in sidelink mode, relay nodes.

[0278] • External Systems: The network may also interface with third-party location services, cloud-based AI / ML frameworks, or external databases that contribute to measurement validation and label refinement.

[0279] In a further embodiment, the measurement entity may be a gNB or TRP that collects UL or SL measurements, while the UE itself serves as the label-generating entity by reporting a position label.

[0280] In some implementations, multiple UEs may act as measurement entities, each collecting DL, UL, or SL measurements. A network node aggregates and processes multi-user positioning data to generate a unified label.

[0281] In an alternative embodiment, a Positioning Reference Unit (PRU) functions as both the measurement entity and the label-generating entity, collecting DL, UL, or SL measurements and generating a ground truth reference position based on controlled environmental conditions.

[0282] The association between the measurement entity and label-generating entity may be dynamically determined based on network configuration, resource allocation, or the availability of reference positioning sources. The system may select different entitiesfor measurement and labeling depending on mobility conditions, service requirements, or the presence of multiple TRPs.

[0283] Examples of AI / ML-Based Positioning and Entity Associations can be seen below: Direct AI / ML Positioning

[0284] • Case 1: UE-based positioning with UE-side model, direct AI / ML positioning o Part B: LMF

[0285] o Part A: UE

[0286] o Association entity: UE

[0287] o Validation entity: UE I LMF

[0288] • NG-RAN node assisted positioning with LMF-side model, direct AI / ML positioning

[0289] o Part B: LMF / UE

[0290] o Part A: gNB

[0291] o Association entity: LMF

[0292] o Validation entity: LMF

[0293] • UE-assisted / LMF-based positioning with LMF-side model, direct AI / ML positioning

[0294] o Part B: LMF / UE

[0295] o Part A: UE

[0296] o Association entity: LMF

[0297] o Validation entity: LMF

[0298] AI / ML Assisted Positioning

[0299] • NG-RAN node assisted positioning with gNB-side model, AI / ML assisted positioning

[0300] o Part B: LMF / UE

[0301] o Part A: gNB

[0302] o Association entity: gNB I LMF

[0303] o Validation entity: UE I LMF

[0304] • UE-assisted / LMF-based positioning with UE-side model, AI / ML assisted positioning

[0305] o Part B: LMF / UE

[0306] o Part A: UEo Association entity: LMF

[0307] o Validation entity: LMF

[0308] In some embodiments, the validation entity ensures the correct association between Part A (measurement data) and Part B (label, cost, or reward data). The validation entity may operate in multiple configurations depending on network deployment and AI / ML processing strategies.

[0309] In some embodiments, the validation entity receives both Part A and Part B from their respective sources and determine whether the measurement-label association is valid. The validation entity may be performed at multiple network entities (e.g., at both UE and LMF). Alternatively, a single network node (e.g., LMF) aggregates all Part A and Part B data and determines the optimal association.

[0310] In another embodiment, the validation entity provides an output to an action entity, which may be a network node (e.g., gNB, LMF) or an AI / ML agent responsible for optimizing future network decisions. The action entity may use the validation entity’s output to:

[0311] • Refine measurement selection criteria for Part A data collection.

[0312] • Adjust label generation strategies for Part B association.

[0313] • Trigger network reconfiguration events, such as activation, de-activation, switching or Fallback for the functionalities or AI / MI models

[0314] In a further embodiment, the validation entity may itself generate an action based on the validated measurement-label pairs. In this case, the validation entity configured to perform one or more of the following:

[0315] • request for a label or label configuration report. In one option the label request is configured to be on demand.

[0316] • modifying UE measurement reporting intervals, wherein the request is configured to be on demand.

[0317] Figure 6 illustrates an example where a plurality of resources is associated. In this figure, multiple TRPs (A, B, C, D) are illustrated, each operating with multiple beamsor resources. The UE, equipped with multiple transmission and / or reception antennas (e.g., beams, RF chains, or timing error groups), moves along a track and performs transmissions and / or receptions at five time instants (t1 , t2, t3, t4, t5).

[0318] Optionally, the channel measurement report may include an indication of resource ID, Timing Error Group (TEG), and antenna information in addition to the measurement data, quality indicators, and timestamps. This indication is explicitly associated with the reported measurement to ensure correct data interpretation.

[0319] In some embodiments, the measurement report may further include:

[0320] • Antenna Port ID: Specifies the port on which the measurement was performed.

[0321] • Timing Error Group (TEG): Identifies the hardware chain used for signal reception or transmission. A TEG is associated with the transmissions or reception of one or more resources, which have the Tx and or Rx timing errors within a certain margin.

[0322] • Transmit or Receive Beam ID: Identifies the beam direction associated with the measurement.

[0323] • Measurement Report ID: Uniquely identifies a channel measurement event. The measurement report may further contain a resource ID, resource set ID, and TEG group ID, along with timestamp information and associated quality indicators. [Channel Measurement Configuration]

[0324] In some embodiments, the UE or TRP may receive a channel measurement configuration from the network. The configuration assists the device in identifying the appropriate resources for reporting within a measurement report. The measurement configuration granularity may be higher than the reporting granularity, and the network may provide additional information to enable efficient report.

[0325] The configuration may include at least one of the following:

[0326] • Resources and resource set associations

[0327] • TRP ID for distinguishing between multiple TRP transmissions

[0328] • Reporting criteria, such as:

[0329] • Selection of the top-N strongest resourcesResources with the highest received power

[0330] Resources fulfilling a predefined measurement quality threshold [Configuration for a Label Report]

[0331] In some embodiments, the network may provide a label report configuration to the label-generating entity, which defines the structure, content, and association of label reports with channel measurement data. In some cases, the reporting format is predefined or pre-configured.

[0332] The label report configuration may specify at least one of the following:

[0333] • The format and parameters required for label reporting

[0334] • Positioning accuracy, uncertainty indicators, and confidence levels.

[0335] • The validity window or range over which the label remains applicable (to ensure alignment with measurement conditions).

[0336] • The reporting periodicity is generated continuously, on-demand, or at scheduled intervals.

[0337] • Label Report ID: Specifies the reference label assigned to a given measurement.

[0338] In some embodiments, the label report configuration may include associations with one or more measurement reports, ensuring that each label is linked to corresponding channel measurement data. The association may be achieved through explicit measurement references, where each label report includes identifiers corresponding to previously reported channel measurements. The association may be achieved through implicit mapping rules, wherein the association of labels is based on time stamp alignment, TRP identification, or spatial consistency constraints. The solution supports mapping multiple channel measurement reports can be linked to one or more label reports. The mapping process may be based on:

[0339] • Direct Measurement-to-Label Linking: A single PartA(e.g., a channel measurement report from a TRP or UE) is explicitly linked to a single Part B (e.g., a GNSS-based position label or network-estimated location).

[0340] • One-to-Many Association: A single Part A measurement event may be associated with multiple Part B labels (different estimation sources, accuracy levels, or reference systems).• Many-to-One Association: Multiple Part A measurements (e.g., from different TRPs, beams, or frequency layers) can be aggregated and associated with a single Part B label.

[0341] • Many-to-Many Association: A collection of multiple Part A measurements may be mapped to multiple Part B labels, where association logic depends on timestamp alignment, spatial correlation, TRPs, beams or mobility constraints.

[0342] Figure 7A and 7B illustrate two examples where one or more labels are associated with / mapped to multiple measurements. In Figure 7A, channel data measurements (UL measurement data and DL measurement data) are associated with one NW label. Figure 7B illustrates an example where a UE label is associated with / mapped to three different measurement data, that is, UL measurement TRP A, B1 , UL measurement TRP A, B3, and DL measurement. B1 and B3 may indicate that there are two different beams.

[0343] [ Indication to Part B (Labeling Association)]

[0344] In some embodiments, the measurement report may further include an indication related to Part B (labeling data). If the measurement report originates from the same entity responsible for label generation, the report may include an explicit indication linking the measurement to the corresponding label.

[0345] The indication to Part B ensures that the measurement-label association is retained within the same entity, avoiding potential mismatches during post-processing. This indication may be at least one of the following:

[0346] • A direct label reference, identifying the applicable label instance.

[0347] • An association ID, linking the measurement event to a specific ground truth entry.

[0348] In some embodiments, association to reference ground truth label, e.g., a certain Antenna reference point (ARP), to which other ground truth may be associated. The ARP is to indicate which antenna or which reference point in antenna is the reference for measurement or association. A UE may signal relative ground truth locations, with respect to different ARPs.

[0349] In some embodiments, the information of data association between the first set of data and the second set of data may be stored at NW entity (e.g. UDR) or at RAN-node or at AMF, as a part of UE context or UE information stored in a database at network.

[0350] Besides positioning, the association between the first set of data (Part A) and the second set of data (Part B) may be used for other use cases as follows:

[0351] [Part A / Part B in AI / ML Beam Management]

[0352] For moving UEs (BM2 case), there may be two configurations of the input and output parameters of the AI / ML model that need to be properly tuned:

[0353] • Input: measurements of Set B (or different Set B between time-steps) to be included, that are relevant to the prediction of best beam(s)

[0354] • Output: the number of time-steps into the future the AI / ML model prediction (e.g., for the top-K beams) is valid / reliable. A time-step may refer to a time period between 2 measurements.

[0355] Since the AI / ML beam prediction algorithm is trained (or monitored) using a dataset consisting of Part A / Part B (input / output) data points, the validating entity needs to determine the following:

[0356] • Part A: how many past Set B beam measurements are required for accurate future beam prediction.

[0357] • Part B: for a given Part A, for how many future time-steps can Part B (e.g., top- K beams in Set A) be reliably predicted.

[0358] Proper configuration of these parameters depends on the general conditions of the (radio) environment (e.g., SNR levels) and the movement patterns of the UE.

[0359] [Part A / Part B in AI / ML CSI prediction and compression]

[0360] This case may be identical to the beam management use case above.

[0361] For moving UEs there may be two configurations of the input and output parameters of the AI / ML model that need to be properly tuned:

[0362] • Input (applies to both CSI prediction and compression cases): the length of past measurements of CSIs, that are relevant to the prediction of future CSI • Output (applies only to the CSI prediction case): the number of time-steps into the future the AI / ML model CSI prediction is reliableSince the CSI prediction (or compression) algorithm is trained (or monitored) using a dataset consisting of Part A / Part B (input / output) datapoints, the validating entity needs to determine the following:

[0363] • Part A: how many past CSI measurements are required for accurate future CSI prediction (or CSI reconstruction)

[0364] • Part B: for a given Part A:

[0365] o for how many future time-steps can we reliably predict CSI values o how accurately can we reconstruct the input CSI

[0366] Proper configuration of these parameters depends on the general conditions of the (radio) environment (e.g., SNR levels) and the movement patterns of the UE.

[0367] [Part A / Part B in RRM optimization processes]

[0368] In accordance with an embodiment, collecting ground truth label data (Part B) comprises obtaining reference information associated with a measurement event, wherein the reference information may not include ground truth labels (e.g., positioning labels) as defined in supervised learning approaches, but quantities that are iteratively minimized or maximized through a learning or optimization process, such as quality indicators, performance metrics and constraint violations.

[0369] The latter interpretation is applicable to data collection tasks (for training or monitoring) for radio resource management (RRM) processes. Such problems can be seen as sequential decision making (or planning) problems, where the AI / ML model (called agent in these settings) can make decisions that affect the overall network. In this case, the part A consists of channel (and other) measurements, as well as the AI / ML model outputs (actions), while Part B contains a notion of reward or cost, reflecting how well the AI / ML model performs.

[0370] Example use cases for Part A / Part B discussion are provided in the Table below. The fourth column indicates to which extend the AI / ML model outputs (actions, included in part A) have long-term consequences in the system performance or cost (Part B). Therefore, correct association of Part A and Part B by the validation entity is crucial for training (or monitoring) an AI / ML model for each task.

[0371] To add to the complexity, the validation entity must associate measurements and actions from potentially more than one UEs, that span several time-steps and arelogged at different frequencies or time-intervals (Part A), with Part B or more than one “labels” in part B, like in the admission control use case, that may be provided by combining reports from several UEs, e.g., for fairness calculation. Fairness may be calculated to maximize the allocation of resources to the most poorly treated sessions or the worst-case UE. Admission control refers in general to a mechanism (or agent) that manages network resources in order to enforce congestion control and QoS requirements. A new access request (e.g., a UE at a BS) is accepted if enough resources are available, so no QoS violations in the already served UEs are observed.

[0372] The table below shows the different examples of Part A and Part B association in RRM optimization processes:

[0373]

[0374]

[0375]

[0376] Use of ground-truth label for other AI / ML enabled use cases:

[0377] In some embodiments, a UE may deploy an AI / ML based model for determining one or more network parameters and / or measurements. For example, the measurement (e.g. RSRP, RSRQ, CQI, ... etc) or parameters (e.g. best beams) may depend on location and / or orientation.

[0378] Some implementations of the label generating entity in other AI / ML enabled use cases are the following:

[0379] • Measurement Entity as Label Provider: In reinforcement learning-based systems, the measurement entity may generate labels (rewards or cost functions) based on observed network performance metrics.

[0380] • Validation Entity as Label Provider: The validation entity may derive labels based on cross-validated measurement results, performance benchmarks, or additional network information.

[0381] • Multi-Source Label Aggregation: Labels may be computed by combining inputs from multiple sources, such as UE reports or network-level observations. In some embodiments, a UE may be provided a first configuration or a set of first configurations, wherein the first configuration provides the information about the reference signal, whose measurement is obtained by performing physical measurement of the first set of configurations.In some embodiments, a UE may be provided a second configuration or a set of second configurations, wherein the second configuration provides the information about reference signal, whose measurement is obtained by using one or more AI / MI models on the first set of configurations.

[0382] In some embodiments, a UE may be provided with a third set of configurations or a set of third configuration, where the third configuration may be a subset of first and / or the second configuration. A UE performs physical measurement on the third configuration.

[0383] The measurement performed on the third set of configurations is the ground truth label, where the measurement on first set of configuration forms an input and the second set of configuration forms an output. In some cases, the measurement may be determined by consolidating one or more third set of measurements. For example, estimated based on two resources provided in third set of configurations. Two resources may indicate Reference signals that are arranged into Resource sets and resources within a set.

[0384] In some embodiments, a UE may be configured by the network to store and / or report at least one of the following:

[0385] • input measurements (performed on first set of configurations),

[0386] • predicted measurements (predicted on a second set of configuration)

[0387] • ground truth label (measurement performed on a third set of configurations).

[0388] • UE location obtained using GNSS

[0389] • UE sensor information

[0390] • UE orientation information

[0391] • UE velocity information

[0392] In some examples, the third set of configurations may be a subset of second set of configurations.

[0393] The collected information may be transferred to the NW using an RRC message (e.g. measurement reports).

[0394] In some implementations, the UE may be configured to store and / or report subject to certain criteria. The criteria may be:1 ) A configured measurement quantity exceeds a certain threshold 2) A configured measurement quantity exceeds a certain threshold but remains below a second threshold.

[0395] 3) A configured measurement quantity exceeds a fall below threshold

[0396] 4) Events such as beam failure, handover failure.

[0397] 5) The UE buffer exceeds a certain threshold. The threshold may be based on U E-capability.

[0398] 6) The UE determines that a low power condition is reached.

[0399] In some embodiments, a network entity may configure a UE to store the above information for a certain period of time. The collected information may be associated with a timestamp and if the reporting criteria is not triggered, the stored information is discarded.

[0400] In other examples, the UE may be configured to store and report the stored information in an on-demand basis. The measurement is reported as soon as ground truth is available.

[0401] [Use of Positioning Reference Unit Data in AI / ML-Based Positioning]

[0402] In an illustrative example, the wireless device obtains a configuration for collecting a first set of data, wherein the configuration comprises reference signal configuration information provided by the network node. The reference signal configuration information may include parameters for one or more reference signal resource sets, such as but not limited to: periodicity, bandwidth, resource identifiers, and quasi colocation information. The wireless device may collect the first set of data by performing measurements on the configured reference signals. By way of example, the first set of data may comprise one or more of a reference signal time difference, a reference signal received power, a reference signal received path power, a receivetransmit time difference, or combinations thereof. Each measurement may be associated with a timestamp and one or more quality indicators.

[0403] As a further example, the second set of data may comprise measurements and location information of the positioning reference unit. The measurements of the positioning reference unit may comprise any of the measurement types that the wireless device collects, and may further comprise carrier phase measurements. Thelocation information of the positioning reference unit may comprise geographic coordinates of the positioning reference unit.

[0404] In some embodiments, the network node provides the second set of data to the wireless device as part of assistance data for an AI / ML-based positioning method. For example, the network node may include the second set of data in the same message or information structure as the reference signal configuration used by the wireless device to collect the first set of data. This structural association may act to link the first set of data and the second set of data to the same positioning context. In some embodiments, the association between the first set of data and the second set of data is determined by the network node by providing both the reference signal configuration and the positioning reference unit data within a common assistance data structure. The network node may select the positioning reference unit data that corresponds to the same reference signal resources (identified, e.g., by resource identifiers) that the wireless device is configured to measure.

[0405] In some embodiments, the association between the first set of data and the second set of data is determined by the wireless device by matching shared identifiers present in both data sets. For example, the first set of data and the second set of data may each contain one or more of a reference signal identifier, a resource set identifier, a resource identifier, and a timestamp. The wireless device may use these shared identifiers to associate its own measurements with the corresponding positioning reference unit measurements. In some embodiments, the wireless device uses both the first set of data and the second set of data as inputs to one or more machine learning models for position estimation. The positioning reference unit data may serve as reference information that enables the one or more machine learning models to improve the accuracy of the position estimate, for example by providing correction terms or calibration information derived from a reference device at a known location, which may be referred to as a positioning reference unit.

[0406] Referring to Figure 8, there is illustrated a method (800) performed by a wireless device (1000) according to some of the previously described embodiments. The method comprises:

[0407] - obtaining (801 ) a configuration for collecting a first set of data; and- collecting (802) the first set of data, wherein an association between the first set of data and a second set of data is determined by the wireless device or a network node.

[0408] In some embodiments, the second set of data is collected or generated by a network node or the wireless device.

[0409] In some embodiments, the network node for collecting or generating the second set of data is the same as the network node for determining the association between the first set of data and the second set of data.

[0410] In some embodiments, the network node for collecting or generating the second set of data is different from the network node for determining the association between the first set of data and the second set of data.

[0411] In some embodiments, the method further comprising receiving the second set of data from the network node where the second set of data is collected.

[0412] In some embodiments, the method further comprising:

[0413] - obtaining a configuration for reporting the first set of data; and

[0414] - sending one or more reports related to the first set of data based on the configuration for reporting the first set of data.

[0415] In some embodiments, it is disclosed a method performed by a wireless device in a wireless communication network. The method comprises:

[0416] - obtaining a configuration for collecting a first set of data; and

[0417] - collecting the first set of data, wherein the first set of data is combined with one or more other sets of data from one or more other wireless devices, and an association between the combined set of data and a second set of data is determined by the wireless device or a network node.

[0418] In some embodiments, the second set of data is collected or generated by a network node or the wireless device.

[0419] In some embodiments, the network node for collecting or generating the second set of data is the same as the network node for determining the association between the combined set of data and the second set of data.In some embodiments, the network node for collecting or generating the second set of data is different from the network node for determining the association between the combined set of data and the second set of data.

[0420] In some embodiments, the method further comprises receiving the second set of data from the network node where the second set of data is collected.

[0421] In some embodiments, the method further comprises:

[0422] - obtaining a configuration for reporting the first set of data or the combined set of data; and

[0423] - sending one or more reports related to the first set of data or the combined set of data based on the configuration for reporting the first set of data or the combined set of data.

[0424] In some embodiments, a method performed by a wireless device in a wireless communication network is disclosed. The method comprises:

[0425] - obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a first set of data and the second set of data is determined by the wireless device or a network node.

[0426] In some embodiments, the first set of data is collected by a network node or the wireless device.

[0427] In some embodiments, the network node for collecting the first set of data is the same as the network node where the association between the first set of data and the second set of data is determined.

[0428] In some embodiments, the network node for collecting the first set of data is different from the network node where the association between the first set of data and the second set of data is determined.

[0429] In some embodiments, the method further comprises receiving the first set of data from the network node where the first set of data is collected.

[0430] In some embodiments, the method further comprises:

[0431] - obtaining a configuration for reporting the second set of data; and- sending one or more reports related to the second set of data based on the configuration for reporting the second set of data.

[0432] In some embodiments, the first set of data is collected at a first time stamp, and the second set of data is collected or generated at a second time stamp.

[0433] In some embodiments, the first set of data comprises measurement data.

[0434] In some embodiments, the first set of data comprises one or more quality indicators of the measurement data.

[0435] In some embodiments, the first set of data comprises one or more time stamps of the measurement data.

[0436] In some embodiments, the measurement data comprises channel measurement data. In some embodiments, the measurement data is used as input data to one or more machine learning models.

[0437] In some embodiments, the method is for estimating or determining a position.

[0438] In some embodiments, the second set of data comprises labelled data.

[0439] In some embodiments, the second set of data comprises one or more quality indicators of the labelled data.

[0440] In some embodiments, the second set of data comprises one or more time stamps of the labelled data.

[0441] In some embodiments, the measurement data is collected based on a first set of reference signal resources, and the labelled data is collected or generated based on a second set of reference signal resources that is different from the first set of reference signal resources.

[0442] In some embodiments, a request is transmitted by the wireless device for an update of the first set of data and / or the second set of data based on changes in measurement conditions.

[0443] In some embodiments, the association between the first set of data and the second set of data is determined based on one or more validation criteria.In some embodiments, the changes in the measurement conditions include updates of the one or more validation criteria.

[0444] In some embodiments, at least one of the one or more validation criteria is a validity range of the association.

[0445] In some embodiments, the validity range of the association is determined based on at least one of a mobility state of the wireless device, channel conditions of the wireless device, surrounding environment conditions of the wireless device, a time correlation, a spatial correlation, or a confidence level associated with association.

[0446] In some embodiments, the validity range of the association is based on a time window or a parameter associated with the first set of data and / or the second set of data. In some embodiments, the validity range of the association is defined as a function based on a reference time and one or two time offsets, the one or two time offsets defining boundaries of the validity range of the association.

[0447] In some embodiments, the boundaries of the validity range of the association are specified, either preconfigured or configured by a network node, and the wireless device selects and reports validity values within the specified boundaries of the validity range of the association.

[0448] In some embodiments, the validity range of the association is represented as a Gaussian-like or truncated Gaussian function centred around a reference time.

[0449] In some embodiments, the validity range is defined within a range as [-X x Ts, T_REF, +Y x Ts], where X and Y are equal or different values, and the maximum values for X and Y are predefined or fixed in a report configuration or a measurement configuration. In some embodiments, the validity range of the second set of data (such as labelled data) includes a spatial dimension, defining a geographical area within which the second set of data remains valid.

[0450] In some embodiments, the validity range of the association includes an angular direction, wherein the angular direction is within a defined angle range relative to a reference direction.

[0451] In some embodiments, the validity range is associated with a beam identifier (beam ID) or a set of beam IDs, wherein the first and / or second set of data is valid only whenone or more measurements are performed on a beam or a set of beams corresponding to the beam ID or the set of beam IDs.

[0452] Accordingly, the validity range of the data association may be tied to specific beam identifiers. In a specific 5G implementation, a beam ID may correspond to a CSI-RS Resource Indicator (CRI) or an SS / PBCH Block Resource Indicator (SSBRI). For example, during Reference Signal Prediction Accuracy Indicator (RS-PAI) reporting, the predicted beam instance is considered as an accurate prediction or valid only when L1-RSRP measurements performed on a monitored beam (e.g., nrofBestBeamforMonitoring-r19) successfully map to a previously reported predicted beam ID (e.g., P-CRI or P-SSBRI). Furthermore, the validity range may be defined by spatial boundaries, which in the context of NR, are enforced via downlink spatial domain transmission filters or Quasi Co-Location (QCL) Type-D spatial receive parameters associated with the specific CSI-RS or SSB resources.

[0453] For example, in an AI / ML beam prediction scenario, the beam identifiers may correspond to predicted beam indices (e.g., predicted reference signal resource indicators or predicted synchronization signal block indices) reported by the wireless device. The set of predicted beam identifiers constitutes the set of beam IDs to which the validity range is tied. The validity of the association between the measurement data (which may be referred to as the first set of data) and the predicted beam data (which may be referred to as the second set of data) is conditioned on whether actual beam measurements confirm the predicted beam identifiers: the wireless device may monitor the prediction accuracy by performing L1-RSRP measurements on a set of monitoring beams at one or more monitoring occasions and determine whether the best measured beams match the predicted beam identifiers. The predicted beam data may be treated as valid only when the monitored beams correspond to the predicted beam identifiers, thereby defining a validity range in the beam domain. If the predicted beam identifiers do not match any of the best measured beams, the prediction instance is counted as inaccurate and the association between the measurement input and the prediction output is not validated for that monitoring occasion.

[0454] In some embodiments, the validity range of the association is defined by spatial boundaries, including coordinate-based limits, sectorized coverage areas, orpredefined network-defined regions, within which the first set of data and / or the second set of data (such as labelled data) is valid.

[0455] In some embodiments, the validity range is a numerical value determined by the wireless device or the network node.

[0456] In some embodiments, the mobility state of the wireless device comprises at least one of a moving speed of the wireless device, a moving direction of the wireless device, or a trajectory of the wireless device within the wireless communication network.

[0457] In some embodiments, the time correlation is determined based on correlations of time stamps associated with the first set of data and / or the second set of data.

[0458] In some embodiments, the spatial correlation is determined based on correlations of position estimations.

[0459] In order to perform the previously described process or method steps performed by the wireless device, there is also provided a wireless device. The wireless device may be a UE or an loT device. Figure 10 illustrates a simplified block diagram depicting a wireless device 1000. The wireless device 1000 comprises a processor 1010 or processing circuit or a processing module or a processor means 1010; a receiver circuit or receiver module 1040; a transmitter circuit or transmitter module 1050; a memory module 1020, a transceiver circuit or transceiver module 1030 which may include the transmitter circuit 1050 and the receiver circuit 1040. The wireless device 1000 further comprises an antenna system 1060 which includes antenna circuitry for transmitting and receiving signals to / from at least the network node or other wireless device(s). The antenna system employs beamforming as previously described.

[0460] A wireless device (WD) may be a user equipment (UE), non-access-point station (non-AP STA), a mobile terminal, a wireless terminal, a mobile station, an loT device, a machine type communication (MTC) device, etc. loT devices may include wireless sensors, software, actuators, and computer devices. The loT devices can be imbedded into mobile devices, motor vehicles, industrial equipment, environmental sensors, medical devices, aerial vehicles and more, as well as network connectivity that enables these devices to collect and exchange data across an existing network infrastructure. In some embodiments, the non-limiting terms wireless device (WD), non-access-point station (non-AP STA) or a user equipment (UE) are used interchangeably.The wireless device 1000 may belong to any radio access technology including 4G or LTE, LTE-A, 5G, advanced 5G or a combination thereof that support beamforming technology. The wireless device comprising the processor and the memory contains instructions executable by the processor, whereby the wireless device 1000 is operative or is configured to perform any one of the embodiments related to the wireless device as previously described.

[0461] The processing module / circuit 1010 includes a processor, microprocessor, an application specific integrated circuit (ASIC), field programmable gate array (FPGA), or the like, and may be referred to as the “processor.” The processor 1010 controls the operation of the wireless device and its components. Memory (circuit or module) 1020 includes a random-access memory (RAM), a read only memory (ROM), and / or another type of memory to store data and instructions that may be used by processor 1010. In general, it will be understood that the wireless device 1000 in one or more embodiments includes fixed or programmed circuitry that is configured to carry out the operations in any of the embodiments disclosed herein.

[0462] In at least one such example, the processor 1010 includes a microprocessor, microcontroller, DSP, ASIC, FPGA, or other processing circuitry that is configured to execute computer program instructions from a computer program stored in a non-transitory computer-readable medium that is in or is accessible to the processing circuitry. Here, “non-transitory” does not necessarily mean permanent or unchanging storage, and may include storage in working or volatile memory, but the term does connote storage of at least some persistence. The execution of the program instructions specially adapts or configures the processing circuitry to carry out the operations disclosed in this disclosure relating to the wireless device. Further, it will be appreciated that the wireless device 1000 may comprise additional components.

[0463] The wireless device 1000 by means of processor 1010 executes instructions contained in the memory 1020 whereby the wireless device is operative to perform any one of the previously described embodiments related to the actions performed by the wireless device, some of which are presented in appended claims.

[0464] There is also provided a computer program comprising instructions which when executed by the processor 1010 of the wireless device cause the processor 1010 tocarry out the method according to any one of the previously described embodiments.

[0465] Referring to Figure 9, there is illustrated a method performed by a network node (1100) in a wireless communication network according to some of the previously described embodiments. The method comprises:

[0466] • obtaining (901 ) a configuration for collecting or generating a second set of data, and

[0467] • collecting or generating (902) the second set of data, wherein an association between a first set of data and the second set of data is determined by a wireless device or a network node.

[0468] In some embodiments, the first set of data is collected by a wireless device or a network node.

[0469] In some embodiments, the network node for collecting the first set of data is the same network node for collecting or generating the second set of data.

[0470] In some embodiments, the network node for collecting the first set of data is different from the network node for collecting or generating the second set of data.

[0471] In some embodiments, the network node for collecting the first set of data is the same network node for determining the association between the first set of data and the second set of data.

[0472] In some embodiments, the network node for collecting the first set of data is different from the network node for determining the association between the first set of data and the second set of data.

[0473] In some embodiments, the method further comprises receiving the first set of data from the wireless device or the network node where the first set of data is collected.

[0474] In some embodiments, the method further comprises:

[0475] - obtaining a configuration for reporting the second set of data; and

[0476] - sending one or more reports related to the second set of data based on theconfiguration for reporting the second set of data.

[0477] In some embodiments, a method performed by a network node in a wireless communication network is disclosed. The method comprises:

[0478] - obtaining a configuration for collecting a first set of data; and

[0479] - collecting the first set of data, wherein an association between the first set of data and a second set of data is determined by a network node or a wireless device.

[0480] In some embodiments, the second set of data is collected or generated by a wireless device or a network node.

[0481] In some embodiments, the network node for collecting or generating the second set of data is the same network node for collecting the first set of data.

[0482] In some embodiments, the network node for collecting or generating the second set of data is different from the network node for collecting the first set of data.

[0483] In some embodiments, the network node for collecting or generating the second set of data is the same network node for determining the association between the first set of data and the second set of data.

[0484] In some embodiments, the network node for collecting or generating the second set of data is different from the network node for determining the association between the first set of data and the second set of data.

[0485] In some embodiments, the method further comprises receiving the second set of data from the wireless device or the network where the second set of data is collected or generated.

[0486] In some embodiments, the method further comprises:

[0487] - obtaining a configuration for reporting the first set of data; and

[0488] - sending one or more reports related to the first set of data based on the configuration for reporting the first set of data.

[0489] In some embodiments, a method performed by a network node in a wireless communication network is disclosed. The method comprises:- obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a combined set of data including a first set of data and the second set of data is determined by a wireless device or a network node.

[0490] In some embodiments, the first set of data is collected by a wireless device, and the combined set of data is collected from one or more wireless devices.

[0491] In some embodiments, the network node for collecting or generating the second set of data is the same as the network node for determining the association between the combined set of data including the first set of data and the second set of data.

[0492] In some embodiments, the network node for collecting or generating the second set of data is different from the network node for determining the association between the combined set of data including the first set of data and the second set of data.

[0493] In some embodiments, the method further comprises receiving the combined set of data including the first set of data from one or more wireless devices including the wireless device where the first set of data is collected.

[0494] In some embodiments, the method further comprises:

[0495] - obtaining a configuration for reporting the second set of data; and

[0496] - sending one or more reports related to the second set of data based on the configuration for reporting the second set of data.

[0497] In some embodiments, a method performed by a first network node in a wireless communication network is disclosed. The method comprises:

[0498] - receiving a first set of data from a second network node;

[0499] - generating or receiving a second set of data; and

[0500] - associating between the first set of data and the second set of data;

[0501] In some embodiments, the first network node is a core network (CN) node, and the second network node is a radio access network (RAN) node.

[0502] In some embodiments, the CN node is a Location Management Function (LMF).To perform the previously described process or method steps performed by the network node there is also provided a network node. The term ‘network node’ used herein may be any kind of network node comprised in a network which may further comprise any of base station (BS), radio base station (RBS), base transceiver station (BTS), base station controller (BSC), radio network controller (RNC), g Node B (gNB), evolved Node B (eNB or eNodeB), Node B, multiple transmission point (multi-TRP), 5G access nodes, multistandard radio (MSR) radio node such as MSR BS, multi-cell / multicast coordination entity (MCE), integrated access and backhaul (IAB) node, relay node, donor node controlling relay, radio access point (AP), access point station (AP STA), transmission points, transmission nodes, Remote Radio Unit (RRU), Remote Radio Head (RRH), a core network node (e.g., mobile management entity (MME), self-organizing network (SON) node, a coordinating node, a positioning node, an external node (e.g., 3rd party node, a node external to the current network), nodes in distributed antenna system (DAS), a spectrum access system (SAS) node, an element management system (EMS), etc. The network node may also comprise test equipment. The network nodes serve users within a cell. Figure 11 illustrates a block diagram depicting a network node 1100. Optionally, the network node may be a gNodeB, gNB, a Transmission-Reception Point, TRP, or a network server. Optionally, the network node is a core network (CN) node. Optionally, the CN node is a Location Management Function (LMF).

[0503] The network node 1100 comprises a processor 1110 or processing circuit or a processing module or a processor means 810; a receiver circuit or receiver module 1140; a transmitter circuit or transmitter module 1150; a memory module 1120, a transceiver circuit or transceiver module 1130 which may include the transmitter circuit 1150 and the receiver circuit 1140. The network node 1100 further comprises an antenna system 1160 which includes antenna circuitry for transmitting and receiving signals to / from at least the wireless device. The antenna system employs beamforming as previously described.

[0504] The network node 1100 may belong to any radio access technology including 4G or LTE, LTE-A, 5G, advanced 5G or a combination thereof that support beamforming technology. The network device comprising the processor and the memory contains instructions executable by the processor, whereby the network node 1100 is operativeor is configured to perform any one of the embodiments related to the network node 1100 as previously described.

[0505] The processing module / circuit 1110 includes a processor, microprocessor, an application specific integrated circuit (ASIC), field programmable gate array (FPGA), or the like, and may be referred to as the “processor.” The processor 1110 controls the operation of the network node and its components. Memory (circuit or module) 1120 includes a random-access memory (RAM), a read only memory (ROM), and / or another type of memory to store data and instructions that may be used by processor 810. In general, it will be understood that the network node in one or more embodiments includes fixed or programmed circuitry that is configured to carry out the operations in any of the embodiments disclosed herein.

[0506] In at least one such example, the processor 1110 includes a microprocessor, microcontroller, DSP, ASIC, FPGA, or other processing circuitry that is configured to execute computer program instructions from a computer program stored in a non-transitory computer-readable medium that is in or is accessible to the processing circuitry. Here, “non-transitory” does not necessarily mean permanent or unchanging storage, and may include storage in working or volatile memory, but the term does connote storage of at least some persistence. The execution of the program instructions specially adapts or configures the processing circuitry to carry out the operations disclosed in this disclosure relating to the wireless device. Further, it will be appreciated that the wireless device 800 may comprise additional components. The network node 800 may also be viewed as a Transmitter and Receiver Point (TRP).

[0507] The network node 1100 by means of processor 1110 executes instructions contained in the memory 1120 whereby the network node 1100 is operative to perform any one of the previously described embodiments related to the actions performed by the network node.

[0508] There is also provided a computer program comprising instructions which when executed by the processor 1110 of the network node cause the processor 1110 to carry out the method according to some embodiments.Reference throughout this specification to “an example” or “exemplary” means that a particular feature, structure, or characteristic described in connection with the example is included in at least one embodiment of the present technology. Thus, appearances of the phrases “in an example” or the word “exemplary” in various places throughout this specification are not necessarily all referring to the same embodiment.

[0509] Throughout this disclosure, the word "comprise" or “comprising” has been used in a non-limiting sense, i.e. meaning "consist at least of". Although specific terms may be employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation. The embodiments herein may be applied in any wireless systems including LTE or 4G, LTE-A (or LTE-Advanced), 5G, advanced 5G, WiMAX, WiFi, satellite communications, TV broadcasting etc.

Claims

1. CLAIMS1. A method (800) performed by a wireless device (1000) in a wireless communication network, the method comprising:- obtaining (801 ) a configuration for collecting a first set of data; and- collecting (802) the first set of data, wherein an association between the first set of data and a second set of data is determined by the wireless device or a network node.

2. The method according to claim 1 , wherein the second set of data is collected or generated by a network node or the wireless device.

3. The method according to claim 2, wherein the network node for collecting or generating the second set of data is the same as the network node for determining the association between the first set of data and the second set of data.

4. The method according to claim 2, wherein the network node for collecting or generating the second set of data is different from the network node for determining the association between the first set of data and the second set of data.

5. The method according to any of claims 2-4, the method further comprising receiving the second set of data from the network node where the second set of data is collected.

6. The method according to any of claims 1-5, the method further comprising:- obtaining a configuration for reporting the first set of data; and- sending one or more reports related to the first set of data based on the configuration for reporting the first set of data.

7. A method performed by a wireless device in a wireless communication network, the method comprising:- obtaining a configuration for collecting a first set of data; and- collecting the first set of data, wherein the first set of data is combined with one or more other sets of data from one or more other wireless devices, and an association between the combined set of data and a second set of data is determined by the wireless device or a network node.

8. The method according to claim 7, wherein the second set of data is collected or generated by a network node or the wireless device.

9. The method according to claim 8, wherein the network node for collecting or generating the second set of data is the same as the network node for determining the association between the combined set of data and the second set of data.

10. The method according to claim 8, wherein the network node for collecting or generating the second set of data is different from the network node for determining the association between the combined set of data and the second set of data.

11. The method according to any of claims 8-10, the method further comprising receiving the second set of data from the network node where the second set of data is collected.

12. The method according to any of claims 7-11 , the method further comprising:- obtaining a configuration for reporting the first set of data or the combined set of data; and- sending one or more reports related to the first set of data or the combined set of data based on the configuration for reporting the first set of data or the combined set of data.

13. A method performed by a wireless device in a wireless communication network, the method comprising:- obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a first set of data and the second set of data is determined by the wireless device or a network node.

14. The method according to claim 13, wherein the first set of data is collected by a network node or the wireless device.

15. The method according to claim 14, wherein the network node for collecting the first set of data is the same as the network node where the association between the first set of data and the second set of data is determined.

16. The method according to claim 14, wherein the network node for collecting the first set of data is different from the network node where the association between the first set of data and the second set of data is determined.

17. The method according to any of claims 14-16, the method further comprising receiving the first set of data from the network node where the first set of data is collected.

18. The method according to any of claims 13-17, the method further comprising:- obtaining a configuration for reporting the second set of data; and- sending one or more reports related to the second set of data based on the configuration for reporting the second set of data.

19. The method according to any of claims 1 -18, wherein the first set of data is collected at a first time stamp, and the second set of data is collected or generated at a second time stamp.

20. The method according to any of claims 1 -19, wherein the first set of data comprises measurement data.

21. The method according to claim 20, wherein the first set of data comprises one or more quality indicators of the measurement data.

22. The method according to claim 20 or 21 , wherein the first set of data comprises one or more time stamps of the measurement data.

23. The method according to any of claims 20-22, wherein the measurement data comprises channel measurement data.

24. The method according to any of claims 20-23, wherein the measurement data is used as input data to one or more machine learning models.

25. The method according to any of claims 1-24, wherein the method is for estimating or determining a position.

26. The method according to any of claims 1-25, wherein the second set of data comprises labelled data.

27. The method according to claim 26, wherein the second set of data comprises one or more quality indicators of the labelled data.7028. The method according to claim 26 or 27, wherein the second set of data comprises one or more time stamps of the labelled data.

29. The method according to claim 20 or 26, wherein the measurement data is collected based on a first set of reference signal resources, and the labelled data is collected or generated based on a second set of reference signal resources that is different from the first set of reference signal resources.

30. The method according to any of claims 1-29, wherein a request is transmitted by the wireless device for an update of the first set of data and / or the second set of data based on changes in measurement conditions.

31. The method according to any of claims 1-30, wherein the association between the first set of data and the second set of data is determined based on one or more validation criteria.

32. The method according to claims 30 and 31, wherein the changes in the measurement conditions include updates of the one or more validation criteria.

33. The method according to claim 31 or 32, wherein at least one of the one or more validation criteria is a validity range of the association.

34. The method according to claim 33, wherein the validity range of the association is determined based on at least one of a mobility state of the wireless device, channel conditions of the wireless device, surrounding environment conditions of the wireless device, a time correlation, a spatial correlation, or a confidence level associated with association.

35. The method according to claim 33 or 34, wherein the validity range of the association is based on a time window or a parameter associated with the first set of data and / or the second set of data.

36. The method according to any of claims 33-35, wherein the validity range of the association is defined as a function based on a reference time and one or two time offsets, the one or two time offsets defining boundaries of the validity range of the association.

37. The method according to claim 36, wherein the boundaries of the validity range of the association are specified, either preconfigured or configured by a network node,71and the wireless device selects and reports validity values within the specified boundaries of the validity range of the association.

38. The method according to any of claims 33-37, wherein the validity range of the association is represented as a Gaussian-like or truncated Gaussian function centred around a reference time.

39. The method according to any of claims 33-38, wherein the validity range is defined within a range as [-X x Ts, T_REF, +Y x Ts], where X and Y are equal or different values, and the maximum values for X and Y are predefined or fixed in a report configuration or a measurement configuration.

40. The method according to claim 35, wherein the validity range of the second set of data (such as labelled data) includes a spatial dimension, defining a geographical area within which the second set of data remains valid.41.The method according to any of the claims 33-40, wherein the validity range of the association includes an angular direction, wherein the angular direction is within a defined angle range relative to a reference direction.

42. The method according to any of claims 33-41, wherein the validity range is associated with a beam identifier (beam ID) or a set of beam IDs, wherein the first and / or second set of data is valid only when one or more measurements are performed on a beam or a set of beams corresponding to the beam ID or the set of beam IDs. 43 The method according to any of claims 33-42, wherein the validity range of the association is defined by spatial boundaries, including coordinate-based limits, sectorized coverage areas, or predefined network-defined regions, within which the first set of data and / or the second set of data (such as labelled data) is valid.

44. The method according to any of claims 33-43, wherein the validity range is a numerical value determined by the wireless device or the network node.

45. The method according to claim 34, wherein the mobility state of the wireless device comprises at least one of a moving speed of the wireless device, a moving direction of the wireless device, or a trajectory of the wireless device within the wireless communication network.

46. The method according to any of the claims 34-45, wherein the time correlation is determined based on correlations of time stamps associated with the first set of data and / or the second set of data.

47. The method according to any of the claims 34-46, wherein the spatial correlation is determined based on correlations of position estimations.

48. A method performed by a network node in a wireless communication network, the method comprising:- obtaining a configuration for collecting a first set of data; and- collecting the first set of data, wherein an association between the first set of data and a second set of data is determined by a network node or a wireless device.

49. The method according to claim 48, wherein the second set of data is collected or generated by a wireless device or a network node.

50. The method according to claim 48, wherein the network node for collecting or generating the second set of data is the same network node for collecting the first set of data.

51. The method according to claim 48, wherein the network node for collecting or generating the second set of data is different from the network node for collecting the first set of data.

52. The method according to any of claims 48-51, wherein the network node for collecting or generating the second set of data is the same network node for determining the association between the first set of data and the second set of data.

53. The method according to any of claims 48-52, wherein the network node for collecting or generating the second set of data is different from the network node for determining the association between the first set of data and the second set of data.

54. The method according to claim 49, the method further comprising receiving the second set of data from the wireless device or the network where the second set of data is collected or generated.

55. The method according to any of claims 48-55, the method further comprising:73- obtaining a configuration for reporting the first set of data; and- sending one or more reports related to the first set of data based on the configuration for reporting the first set of data.

56. A method (900) performed by a network node (1100) in a wireless communication network, the method comprising:- obtaining (901 ) a configuration for collecting or generating a second set of data; and- collecting or generating (902) the second set of data, wherein an association between a first set of data and the second set of data is determined by a wireless device or a network node.

57. The method according to claim 56, wherein the first set of data is collected by a wireless device or a network node.

58. The method according to claim 57, wherein the network node for collecting the first set of data is the same network node for collecting or generating the second set of data.

59. The method according to claim 57, the network node for collecting the first set of data is different from the network node for collecting or generating the second set of data.

60. The method according to any of claims 57-59, wherein the network node for collecting the first set of data is the same network node for determining the association between the first set of data and the second set of data.

61. The method according to any of claims 57-60, wherein the network node for collecting the first set of data is different from the network node for determining the association between the first set of data and the second set of data.

62. The method according to any of claims 57 -61, the method further comprising receiving the first set of data from the wireless device or the network node where the first set of data is collected.

63. The method according to any of claims 56-62, the method further comprising:- obtaining a configuration for reporting the second set of data; and- sending one or more reports related to the second set of data based on the configuration for reporting the second set of data.

64. A method performed by a network node in a wireless communication network, the method comprising:- obtaining a configuration for collecting or generating a second set of data; and - collecting or generating the second set of data, wherein an association between a combined set of data including a first set of data and the second set of data is determined by a wireless device or a network node.

65. The method according to claim 64, wherein the first set of data is collected by a wireless device, and the combined set of data is collected from one or more wireless devices.

66. The method according to claim 64 or 65, wherein the network node for collecting or generating the second set of data is the same as the network node for determining the association between the combined set of data including the first set of data and the second set of data.

67. The method according to any of claims 64-66, wherein the network node for collecting or generating the second set of data is different from the network node for determining the association between the combined set of data including the first set of data and the second set of data.

68. The method according to any of claims 65-67, the method further comprising receiving the combined set of data including the first set of data from one or more wireless devices including the wireless device where the first set of data is collected.

69. The method according to any of claims 64-68, the method further comprising:- obtaining a configuration for reporting the second set of data; and- sending one or more reports related to the second set of data based on the configuration for reporting the second set of data.

70. A method performed by a first network node in a wireless communication network, the method comprising:- receiving a first set of data from a second network node;- generating or receiving a second set of data; and- associating between the first set of data and the second set of data;71. The method according to claim 70, wherein the first network node is a core network (CN) node, and the second network node is a radio access network (RAN) node.

72. The method according to claim 71, where the CN node is a Location Management Function (LMF).

73. A wireless device (1000) comprising a processor (1010) and a memory (1020) containing instructions executable by said processor (1010), whereby the wireless device (1000) is operative to perform the method according to any of claims 1-47.

74. The wireless device (1000) according to claim 73, where the wireless device (1000) is a User Equipment, UE.

75. The wireless device (1000) according to claim 73, where the wireless device (1000) is an Internet of Things (loT) device.

76. A network node (1100) comprising a processor (1110) and a memory (1120) containing instructions executable by said processor (1110), whereby the network node (1100) is operative to perform the method according to any of claims 48-72.

77. The network node (1100) according to claim 76, wherein the network node (1100) is a gNodeB, gNB, a Transmission-Reception Point, TRP, or a network server.

78. The network node (1100) according to claim 76, wherein the network node (1100) is a core network (CN) node.

79. The network node (1100) according to claim 78, wherein the CN node is a Location Management Function (LMF).

80. A computer program comprising instructions which when executed on at least one processor of the wireless device (1000), cause the at least said one processor to carry out the method according to any of the claims 1-47.7681. A computer program comprising instructions which when executed on at least one processor of the network node (1100), cause the at least said one processor to carry out the method according to any of the claims 48-72.77

Images