Enhancing UE capabilities

Abstract

An apparatus comprising: one or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the apparatus to perform: determining a function that is not executable but is expected to be executable; determining whether a capability to prepare the desired function is available; requesting to provide the desired functionality based on determining that the capability is unavailable; receiving a desired function; making the received function executable; wherein if the device is configured with a configuration and the configuration is under a set of conditions, the functionality comprises the configuration and a method executable by the device; the function also comprises at least one of the following: direct positioning or auxiliary positioning.

Description

Technical Field

[0001] This disclosure relates to UE capability enhancement.

[0002] abbreviation 3GPP 3rd Generation Partner Program 5G / 6G / 7G 5th generation / 6th generation / 7th generation AI (Artificial Intelligence) AIML (Artificial Intelligence - Machine Learning) AoD departure angle BW bandwidth CA carrier aggregation CIR channel impulse response CSI Channel Status Information gNB Next Generation (5G) Node B ID identifier KPIs (Key Performance Indicators) LMC Location Management Component LMF location management function LOS sight distance LPHAP Low-Power High-Precision Positioning LPP LTE positioning protocol MAC Media Access Control ML machine learning NLOS Non-line-of-sight NR New Radio NRPP New Radio Positioning Protocol NW Network NWDAF network data analysis function PRS positioning reference signal PRU positioning reference unit RF radio frequency RRC Radio Resource Control RSTD Reference Signal Time Difference SIB Information Block TDOA arrival time difference ToA Arrival Time UE User Equipment Glossary (partially taken from 3GPP R1-2205695 and 3GPP R1-2300603)

[0003] Background Technology

[0004] In several AI / ML technologies used for positioning purposes, training is performed at a central ML unit, hereinafter referred to as the central unit (or central ML unit). The central unit can be, for example, a Location Management Function (LMF) residing in the core network, which is typically the entity responsible for coordinating positioning. As another example, the central ML unit can be a 5G Network Data Analytics Function (NWDAF), which runs data analytics to generate insights and take actions to enhance the user experience, including positioning use cases.

[0005] Training in the central ML unit is typically performed according to the following process: A group of data collection devices is deployed at carefully selected locations. An example of such a data collection device is a so-called Location Reference Unit (PRU): a PRU is a reference device at a known location that can provide ground truth measurements. PRUs can be adapted to incorporate real-world and label measurement data for AIML-based learning. This PRU data can be used to refine the location of other target UEs in the same area, thereby improving their positioning accuracy. For simplicity, the data collection device will hereafter be referred to as a PRU, although this can be generalized to any type of data collection device. A PRU can be, or can include, for example, a UE or other node devices that may have one or more known characteristics (e.g., known location and / or known LOS / NLOS classification, and / or known positioning measurements based on certain reference signals), which can perform (multiple) positioning measurements based on the reference signals.

[0006] -PRU performs on-site positioning measurements and reports the measurements to the central unit.

[0007] - In addition to measurements collected from the UE / PRU or alternatives, the central unit can use simulation tools to generate (simulated) positioning measurements. For example, the central unit can use simulated / synthetic data to train an initial model.

[0008] - The central unit combines the above positioning measurements to train a general positioning ML model.

[0009] - The AIML model is deployed at a network entity that runs the ML process and / or algorithm. This entity is then referred to as a host type. The host type performing the ML process can be the target UE, PRU, LMF, or a potential radio access network (e.g., gNB and / or location management component LMC) used to enhance positioning accuracy. Summary of the Invention

[0010] The goal is to improve existing technology.

[0011] According to a first aspect, an apparatus is provided, comprising: One or more processors, and memory, the memory storing instructions that, when executed by the one or more processors, cause the device to perform: Preparation function; and at least one of the following Provide the prepared functions to the terminal; or The broadcast indicates the availability of the prepared functions or serves as a notification of the prepared functions; among which If the terminal is configured and configured under a set of conditions, the functionality includes configuration for the terminal and methods that can be executed by the terminal. The functionality also includes at least one of the following: Direct positioning of the terminal, or Terminal-assisted positioning; and The preparation function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Test the ML model; - Validate the trained ML model; or - Update the ML model.

[0012] The ML model is trained based on the data received from the terminal.

[0013] Instructions, when executed by one or more processors, can also cause the device to perform: Receive corresponding data from each of the multiple terminals; The ML model is trained based on data received from multiple terminals.

[0014] Instructions, when executed by one or more processors, can also cause the device to perform: Define a set of conditions; Send a set of conditions together with at least one of the broadcast and provide functions.

[0015] Instructions, when executed by one or more processors, can also cause the device to perform: Receive requests to provide the prepared functions to the terminal; In response to a request, provide the prepared functionality to the terminal; Without a request, prevent the provision of prepared functions to the terminal.

[0016] Instructions, when executed by one or more processors, can also cause the device to perform: Receive one or more conditions from the terminal.

[0017] Instructions, when executed by one or more processors, can also cause the device to perform: The terminal is requested to make the prepared function executable.

[0018] Instructions, when executed by one or more processors, can also cause the device to perform: After the feature has obtained the updated ML model, continue preparing the feature; and at least one of the following: Provide the updated features or incremental updates to the features to the terminal; or Broadcasts indicate the availability of updated features or incremental updates to features, or serve as an indication that the updated features or incremental updates are available.

[0019] According to a second aspect of this disclosure, an apparatus is provided, comprising: One or more processors, and memory storing instructions that, when executed by the one or more processors, cause the device to perform: Receive requests for functions from the terminal; In response to a request, it provides assistance to the terminal, enabling the terminal to have the capability to perform preparation functions; wherein If the terminal is configured and configured under a set of conditions, the functionality includes configuration for the terminal and methods that can be executed by the terminal. And the functionality also includes at least one of the following: Direct positioning of the terminal, or Terminal-assisted positioning; and The preparation function includes at least one of the following: - Collect data for training ML models; - Training ML models; and - Validate the trained ML model.

[0020] Providing assistance may include configuring the terminal to receive a reference signal, and collecting data may include evaluating the received reference signal.

[0021] Instructions, when executed by one or more processors, can also cause the device to perform: A set of conditions for the receiving terminal; Determine whether a function can be executed by the terminal under a set of conditions; If a function cannot be properly performed by the terminal under a given set of conditions, assistance is prohibited.

[0022] According to a third aspect of this disclosure, an apparatus is provided, comprising: One or more processors, and memory storing instructions that, when executed by the one or more processors, cause the device to perform: A function that is determined to be non-executable but is expected to be executable; Determine whether the capability to prepare the desired functionality is available; Based on the premise that the capability is unavailable, request that the desired functionality be provided; Receive the desired functionality; Make the received function executable; where If the device is configured and configured under a set of conditions, the functionality includes configuration and a method executable by the device; the functionality also includes at least one of the following: Direct positioning or assisted positioning; The required capabilities may include at least one of the following: - The ability to collect data that is used for at least one of the following: training, testing, and validation of an ML model; - Processing power, used for at least one of the training, testing, and validation of ML models; - Availability of processing power for at least one of the training, testing, and validation of ML models; - Memory, used for at least one of training, testing and validating the ML model; - Availability of memory for at least one of training, testing, and validating the ML model; or - The ability to obtain true values ​​for validating ML models.

[0023] Instructions, when executed by one or more processors, can also cause the device to perform: Determine one or more conditions for making the function executable; Send a set of one or more conditions along with the request function.

[0024] Instructions, when executed by one or more processors, can also cause the device to perform: In response to receiving a broadcast availability indication that the feature is ready, request the feature. Instructions, when executed by one or more processors, can also cause the device to perform: Before the functionality is implemented, data is provided for preparing the functionality, where the data is based on the measurements performed; When making a function executable, use the data to perform the function.

[0025] Instructions, when executed by one or more processors, can also cause the device to perform: Along with the receiving function, it receives a set of one or more conditions to enable method execution; This is determined by verifying whether one or more conditions are met.

[0026] Instructions, when executed by one or more processors, can also cause the device to perform: Incremental updates to the receiving function; Update functionality based on incremental updates.

[0027] The prepared function can be received by receiving a broadcast message that includes the function.

[0028] Instructions, when executed by one or more processors, can also cause the device to perform: Check whether the required capabilities for the preparation function are available within the scheduled time.

[0029] Instructions, when executed by one or more processors, can also cause the device to perform: Data must not be provided when making the function executable.

[0030] Instructions, when executed by one or more processors, can also cause the device to perform: Determine whether the function is executable when it is received; Based on the premise that the function is determined to be non-executable, make the function executable is prohibited.

[0031] According to a fourth aspect of this disclosure, an apparatus is provided, comprising: One or more processors, and memory storing instructions that, when executed by the one or more processors, cause the device to perform: A function that is determined to be non-executable but is expected to be executable; Determine if the capabilities used to prepare the functionality are available; Based on the determination that the capability is unavailable, request assistance to make the capability available; Receive assistance; Based on assistance, preparation functions; among which If the device is configured and configured under a set of conditions, the function includes configuration and a method executable by the device; and the function also includes at least one of the following: Direct positioning or assisted positioning; The preparation function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Testing ML models; and Validate the trained ML model.

[0032] Providing assistance may include configuring the terminal to receive a reference signal, and collecting data may include evaluating the received reference signal.

[0033] The preparation function may include additionally updating the trained ML model.

[0034] Instructions, when executed by one or more processors, can also cause the device to perform: A notification is used for one or more sets of conditions that enable a function to be executed; wherein the function is suitable for execution under the condition that a set of conditions are met.

[0035] According to the fifth aspect, a method is provided, comprising: Preparation functions; and at least one of the following: Provide the prepared functions to the terminal; or The broadcast indicates the availability of the prepared functions or is used to indicate the availability of the prepared functions; wherein If the terminal is configured and configured under a set of conditions, the functionality includes configuration for the terminal and methods that can be executed by the terminal. And the functionality also includes at least one of the following: Direct positioning of the terminal, or Terminal-assisted positioning; and The preparation function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Test the ML model; - Validate the trained ML model; or - Update the ML model.

[0036] The ML model is trained based on the data received from the terminal.

[0037] The method may also include: Receive corresponding data from each of the multiple terminals; The ML model is trained based on data received from multiple terminals.

[0038] The method may also include defining a set of conditions; Send a set of conditions together with at least one of the broadcast and provide functions.

[0039] The method may also include receiving a request to provide the prepared functionality to the terminal; In response to a request, provide the prepared functionality to the terminal; Without a request, prevent the provision of prepared functions to the terminal.

[0040] The method may also include receiving a set of one or more conditions from the terminal.

[0041] The method may also include requesting the terminal to make the prepared function executable.

[0042] The method may also include continuing to prepare the feature after the updated ML model has been obtained; and at least one of the following: Provide the updated features or incremental updates to the features to the terminal; or Broadcasts indicate the availability of updated features or incremental updates to features, or serve as an indication that the updated features or incremental updates are available.

[0043] According to a sixth aspect of the present invention, a method is provided, comprising: Receive requests for functions from the terminal; In response to a request, assistance is provided to the terminal, enabling it to prepare for functionality; wherein If the terminal is configured and configured under a set of conditions, the functionality includes configuration for the terminal and methods that can be executed by the terminal. And the functionality also includes at least one of the following: Direct positioning of the terminal, or Terminal-assisted positioning; and The preparation function includes at least one of the following: - Collect data for training ML models; - Training ML models; and - Validate the trained ML model.

[0044] Providing assistance may include configuring the terminal to receive a reference signal, and collecting data may include evaluating the received reference signal.

[0045] The method may also include a set of conditions for the receiving terminal; Determine whether a function can be executed by the terminal under a set of conditions; If a function cannot be properly performed by the terminal under a given set of conditions, assistance is prohibited.

[0046] According to a seventh aspect of the present invention, a method is provided, comprising: A function that is determined to be non-executable but is expected to be executable; Determine whether the capabilities needed to prepare the desired functionality are available; Based on the premise that the capability is unavailable, request that the desired functionality be provided; Receive the desired functionality; Make the received function executable; where If the device is configured and configured under a set of conditions, the functionality includes configuration and a method executable by the device; the functionality also includes at least one of the following: Direct positioning or assisted positioning; The required capabilities may include at least one of the following: - The ability to collect data that is used for at least one of the following: training, testing, and validation of an ML model; - Processing power, used for at least one of the training, testing, and validation of ML models; - Availability of processing power for at least one of the training, testing, and validation of ML models; - Memory, used for at least one of training, testing and validating the ML model; - Availability of memory for at least one of training, testing, and validating the ML model; or - The ability to obtain true values ​​for validating ML models.

[0047] The method may also include determining one or more conditions for making the function executable; Send a set of one or more conditions along with the request function.

[0048] The method may also include requesting the functionality in response to receiving a broadcast availability indication indicating that the functionality is ready.

[0049] The method may also include providing data for preparing the function before the function is implemented, wherein the data is based on the measurements performed; When making a function executable, use the data to perform the function.

[0050] The method may also include receiving a set of one or more conditions, together with the receiving function, to enable method executable; This is determined by verifying whether one or more conditions are met.

[0051] The method may also include incremental updates to the receiving function; Update functionality based on incremental updates.

[0052] The prepared functions can be received by receiving broadcast messages that include the functions.

[0053] The method may also include checking whether the required capabilities of the functions to be prepared are available within a predetermined time.

[0054] The method can also include prohibiting the provision of data while the function is executable.

[0055] The method may also include determining whether the function is executable after receiving the function; Based on the premise that the function is determined to be non-executable, make the function executable is prohibited.

[0056] According to the eighth aspect of this disclosure, a method is provided, comprising: A function that is determined to be non-executable but is expected to be executable; Determine if the capabilities used to prepare the functionality are available; Based on the determination that the capability is unavailable, request assistance to make the capability available; Receive assistance; Based on assistance, preparation functions; among which If the device is configured and configured under a set of conditions, the function includes configuration and a method executable by the device; and the function also includes at least one of the following: Direct positioning or assisted positioning; The preparation function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Testing ML models; and Validate the trained ML model.

[0057] Providing assistance may include configuring the terminal to receive a reference signal, and collecting data may include evaluating the received reference signal.

[0058] The preparation function may include additionally updating the trained ML model.

[0059] The method may also include notification of one or more sets of conditions for making a function executable; wherein the function is suitable for being executable under the condition that a set of conditions are met.

[0060] According to a ninth aspect of this disclosure, a computer program product including an instruction set is provided, which, when executed on a device, is configured to cause the device to perform the method according to any one of aspects five through eight. The computer program product may be embodied in a computer-readable medium or may be directly loaded into a computer.

[0061] According to some example embodiments, at least one of the following advantages can be achieved: • The UE can perform functions that it was not previously prepared for; • When providing features, restrictions specific to certain UEs can be considered; • The network can control whether a certain function is available at the UE, because it can provide the UE with the prepared function or not. • It can reduce signaling overhead: • Collect significantly more training data than for each NR UE.

[0062] • Train better models because computational complexity is not limited to UE • Customize the model on demand, taking into account the constraints and hardware limitations of the UE, whenever needed.

[0063] It should be understood that any of the above modifications may be applied individually or in combination to the relevant aspects they address, unless they are explicitly stated to exclude alternatives. Attached Figure Description

[0064] Further details, features, objects, and advantages will become apparent from the following detailed description of preferred exemplary embodiments, taken in conjunction with the accompanying drawings, wherein: Figure 1 The message flow is shown according to some example embodiments; Figure 2 An apparatus according to an example embodiment is shown; Figure 3 A method according to an example embodiment is shown; Figure 4 An apparatus according to an example embodiment is shown; Figure 5 A method according to an example embodiment is shown; and Figure 6 An apparatus according to an example embodiment is shown; Figure 7 A method according to an example embodiment is shown; Figure 8 An apparatus according to an example embodiment is shown; Figure 9 A method according to an example embodiment is shown; and Figure 10 An apparatus according to an example embodiment is shown. Detailed Implementation

[0065] In the following detailed description with reference to the accompanying drawings, certain exemplary embodiments are described, wherein features of the exemplary embodiments may be freely combined with each other unless otherwise described. However, it should be clearly understood that the description of certain exemplary embodiments is given by way of example only and is in no way intended to be construed as limiting this disclosure to the details disclosed.

[0066] Furthermore, it should be understood that the device is configured to perform the corresponding method, although in some cases only the device or only the method is described.

[0067] AIML-based localization (“AIML localization”) is considered to outperform conventional localization without AIML. However, AIML localization is expected to be computationally more challenging for the UE than conventional methods. Similarly, AIML direct localization is inherently more complex than AIML-assisted localization because the former solves a more difficult task (i.e., calculating the UE’s location) than the latter (i.e., extracting localization measurements (and potentially pre-evaluating them)).

[0068] Therefore, not all UEs can support both types of AIML positioning. Some examples of UEs that do not support these two types of AIML positioning include: • RedCap devices (i.e., degraded devices). Some of these devices can perform both direct and assisted localization, but they may not be able to perform model updates on the device.

[0069] LPHAP devices are battery-limited, but they require high-precision positioning. Therefore, they can greatly benefit from AIML positioning methods, provided these methods are delivered to the device with no (or almost no) additional cost (especially battery consumption). That is, it is recommended that LPHAP devices do not require model tuning, model updates, or local storage of large amounts of training data.

[0070] Specifically, some UEs may not be able to prepare AIML positioning, where preparation includes at least one (or all) of the following: - Collect data for training ML models; - Train the ML model; - Test the ML model; - Validate the trained ML model; or - Update the ML model.

[0071] Data collection for training is a lengthy and complex problem, potentially consuming significant entity (UE, gNB, LMF) resources. Therefore, it's advisable to perform data collection for training, model tuning, model updates, and localization only at the central unit, allowing the UE to contribute solely to assisted localization. However, if all UEs require auxiliary signaling and procedures to facilitate the generation / collection of training data and ground truth, network overhead can be substantial. Transmitting training data between network entities can lead to significant overhead and wasted network resources.

[0072] AIML localization models are typically region-of-interest specific and may not work in all regions of the network. Therefore, when a user roams from one region to another, switching from one localization method (e.g., AIML-assisted localization) to another (e.g., AIML direct localization) may require retraining a new model (or multiple models) associated with that new localization method.

[0073] For proprietary models (e.g., UE vendor-specific models), the models are typically trained outside of 3GPP networks and may take into account different datasets (RF environments, synthetic data), which may not be effective for a given scenario or region of interest. In such cases, model retraining or retuning is mandatory to meet positioning requirements.

[0074] According to some example embodiments, the network (e.g., represented by a central unit) can enhance the capabilities of the UE. For example, the network can enable a UE that only supports assisted positioning but is capable of supporting direct positioning (at least under certain conditions) to support direct positioning (and vice versa). Thus, the network and UE systems can obtain the full benefits of the AIML framework.

[0075] Some example implementations provide a framework for enhancing the AIML positioning capabilities of NR UEs. This framework allows the UE to leverage new capabilities (or multiple new capabilities) developed by the NW to enrich its original capability set and adapt to the constraints of the respective UE. Therefore, the NW can switch the UE's configuration between: a. AIML direct positioning and b. AIML-assisted localization For any UE, although the UE may not initially have capabilities a) and b), in some example embodiments, the UE must agree to undergo NW-led capability enhancement to acquire the desired capabilities.

[0076] According to some example embodiments, the ML model is prepared at a central unit (specifically: created, trained, and often validated) before being provided to the UE for inference from the ML model. Preparing the model at the central unit (i.e., via the network) typically has at least one of the following advantages: • The NW can collect a much larger amount of training data than each UE can. Additionally, data availability is better on the NW side.

[0077] Better models with higher complexity can be trained because computational power is not as limited as that of UEs. • Based on the UE's needs, NW can customize the model to fit the UE's constraints, such as hardware and / or software limitations, and also based on available radio resources.

[0078] Figure 1 The actions performed according to an example embodiment are illustrated. In this example embodiment, the LMF is the central unit. The actions are as follows: 1: The LMF requests the UE's positioning capabilities, especially AIML positioning.

[0079] 2: The UE determines its ability to use AIML positioning.

[0080] 3: AIML positioning functions that the UE determines cannot be executed by the UE but are expected to be executable.

[0081] 4. If the UE determines that a function it expects to execute but cannot, it checks whether it can execute the function provided the prepared function is available to it. The UE can also check whether it can prepare the function (at least collect data, train and validate the ML model on which the function is based). In addition to checking whether the UE can fully prepare the function, it can check whether it can prepare the function within a predetermined timeframe. Figure 1 In the example implementation, it is assumed that the UE either does not check whether it can prepare the function, or the result of the check for preparing the function is negative.

[0082] 5: The UE provides feedback to the LMF regarding the request in 1. Specifically, if the UE is able to execute the prepared function (according to the check in 4), the UE indicates to the LMF the AIML function that is expected to be executed but is not yet executable. Additionally, the UE may indicate an AIML positioning function that is executable at the UE.

[0083] 6: The LMF assessment is provided by the UE in the report in 5. It determines whether function Y is expected to be performed but cannot be performed at the UE.

[0084] 7 and 8: LMF prepares the UE (data collection, training, validation) function.

[0085] 9: LMF provides the prepared function Y to UE.

[0086] 10: Optionally, the UE checks whether it actually supports the prepared function. For example, some processing power and / or memory may be occupied by other tasks at the same time, making the prepared function Y no longer available for execution by the UE.

[0087] 11: If the test in 10 is positive (or the test in 10 is not performed), then the UE indicates to the LMF that function Y can now be performed at the UE.

[0088] 12: Upon receiving message 11, the LMF determines the configuration for enabling function Y to be executed on the UE and provides it to the UE.

[0089] 13: AIML localization for UE can now be based on function Y.

[0090] In some example embodiments, the UE can indicate (request) a set of one or more desired AIML positioning modes and methods before receiving a capability request from the central unit. Therefore, in such example embodiments, Figure 1 Actions 1, 5, and 6 in the code are not executed. Figure 1In action 6, in such an example embodiment, the LMF determines only one or more desired positioning capabilities. In such an example embodiment, actions 11 through 13 are optional. For example, actions 12 and 13 may be performed only if the LMF wants to receive positioning information for the UE based on the desired positioning capabilities.

[0091] The capabilities of a UE may include certain conditions that the UE can satisfy based on its hardware configuration and other tasks to be performed by the UE, such as a maximum amount of processing power or a maximum amount of memory available for performing functions. In some example embodiments, the UE may determine these conditions and notify the LMF accordingly. The LMF may take these conditions into account when preparing its functions.

[0092] Figure 1 The actions and scenarios of the example embodiments are explained again below using some different wording. In this scenario, the LMF may need to obtain the location of the i-th UE. In conventional positioning (Rel-17, non-AIML-based positioning), the LMF can request the UE to report direct location information or auxiliary data, which can help the LMF estimate the UE's location. Unlike conventional positioning mechanisms (non-AIML-based positioning), many of the complexities involved in AIML-based positioning frameworks are due to feature training. Here, "feature" refers to AIML direct positioning or AIML-assisted positioning.

[0093] As an example, after receiving a location capability request from the LMF, a UE that may not have direct AIML location capabilities will use the capability information to respond to the LMF request.

[0094] • The UE can use AIML functions or traditional functions to determine a list of intermediate features, such as CIR, AoD, ToA, RSTD, LOS / NLOS indicators, soft information / high resolution of RSTD, etc., by using PRS signals transmitted by multiple gNBs in its neighborhood.

[0095] • The UE can use a relevant set of inputs / outputs to determine the supported functions and the expected functions.

[0096] • The UE can indicate to the LMF a set of available AIML positioning modes, including supported and expected functions.

[0097] The LMF determines the supported and desired functions based on the capability report from a given UE. The LMF can determine / configure configuration parameters for the supported functions. • The LMF can configure the UE with the supported features for ML positioning enabled.

[0098] · If the UE indicates that it cannot prepare the desired function (e.g., ID Y), but can execute the desired function (e.g., ID Y), the LMF can locally instantiate the desired function (ID Y) and trigger data collection and training based on existing feedback from multiple UEs in the network.

[0099] · Once the desired function is trained / tested and verified on the NW side (e.g., LMF), it provides assistance to the UE for enabling / activating the desired function Y. It may include the transfer of the AIML model associated with function Y from the LMF to the UE.

[0100] · Upon receiving the desired function and UE capabilities, the UE can verify that the desired function is supported. In some example embodiments, since the UE provides its conditions to the LMF which takes them into account, the UE can simply assume that the desired function is supported.

[0101] · The UE can indicate to the LMF a change in a set of available AIML positioning modes and methods (including an update to the set of supported functions); the indication may also include the desired function.

[0102] · Based on the indication from the UE, the LMF can determine / configure the configuration / parameters for the newly supported function (e.g., ID Y).

[0103] A function refers to an AI / ML-enabled feature / FG enabled by (one or more) configurations, where the (one or more) configurations are supported based on conditions indicated by UE capabilities. In other words, if the terminal is configured with a configuration and the configuration is under a set of conditions, the function includes the configuration of the terminal and the methods executable by the terminal. If the method is executable, the corresponding function will also be executable. In other words, an executable method means that the function associated with the method is also executable. When the UE has a specific configuration, the execution of the method is realized. For the configuration, the UE may have a set of conditions specific to the UE. Therefore, the function is also restricted by this condition. For example, the configuration is: a neural network with X = 3 hidden layers, where X = 3 is a parameter setting. The UE may have told the NW that it can have at most Xmax = 10 hidden layers, and thus X < Xmax is a condition that the network (central unit) should comply with.

[0104] UE capabilities may include one or more of the following: - The ability to collect sufficient data for at least one of training, testing, and validating an ML model. That is, the amount of data that the UE can collect for preparing the function must be sufficient such that the expected ML model to be trained has the desired quality.

[0105] - The ability to collect appropriate data for at least one of the training, testing, and validation of the ML model; that is, the data collected by the UE must be the data to be input into the ML model.

[0106] - Processing power, used for at least one of the following: training, testing, and validation of the ML model; - Availability of processing capabilities for at least one of the training, testing, and validation of ML models; - Memory, used for at least one of training, testing and validating the ML model; - Availability of the memory for at least one of training, testing, and validating the ML model; or - Gain the ability to validate the ground reality of the ML model.

[0107] In some example embodiments, the desired functionality is prepared by the LMF based on a request from the UE. In some example embodiments, the desired functionality is prepared independently of any request from the UE and is provided to one or more UEs (e.g., via broadcast). In some of these example embodiments, the LMF may fine-tune the prepared functionality based on conditions indicated by a particular UE before it is provided to that particular UE.

[0108] In some example embodiments, the LMF may simply broadcast an availability indication indicating the availability of the prepared feature. Upon receiving the availability indication, the UE can obtain the prepared feature from the LMF indicated in the availability indication.

[0109] For example, LMF can instantiate an AIML model (also referred to as a "reflection model") for AIML direct localization functionality and train the AIML model using measurement reports provided by multiple UEs. How the AIML model is implemented (e.g., the number of inputs, the number of layers, the architecture of the neural network, etc.) is determined by the network. Typically, the network has knowledge of all possible inputs; that is, it can gather information from a set of multiple UEs.

[0110] In this way, the function can be trained using the locations of all multiple UEs, and the function can be trained over a period of time to ensure its generalization. Since the function is a reflection of the UE-side function, it is called a reflection function. This new function can be made visible to the network as a new UE capability.

[0111] Reflection functionality can also be a new feature of the network. For example, the NW can group the UE's Region of Interest (ROI) and capabilities and create related reflection functions.

[0112] Direct AIML positioning-1 indicates group 1-ROI-1 based on ROI-1 and existing assisted positioning within the UE's capabilities.

[0113] Direct AIML positioning-2 refers to group 2-ROI-2 based on ROI-2 and existing assisted positioning within the UE's capabilities.

[0114] During the functional training process, actual ground truth information can be obtained using a subset of available measurement reports provided by the UE through traditional location estimation (based on NR RAT and / or non-NR based). Therefore, the functional training performed at the LMF considers the UE location and the channel characteristics experienced by each of the multiple UEs. Since this function uses auxiliary information as an input list, such as CIR, PDP, AoD, DL-TDOA, etc., the model can be easily used by the UE with minimal effort once obtained externally. Furthermore, due to the large number of training samples, the impact of any UE impairment can be averaged out.

[0115] Since LMF instantiates an AIML model in such an example embodiment, and the corresponding model can be implemented in one or more UEs, the AIML model can also be represented as a "reflection model".

[0116] The LMF can provide an AIML model trained at the LMF to one or more dedicated UEs. Alternatively, the LMF can broadcast the model, making it available to any UE. However, since AIML models can be large in size, in some example embodiments, the LMF only broadcasts an indication of functionality availability, such as via SIB-XX or Pos-SIB. Each UE can then decide whether it wants to download the AIML model. If a UE decides to download the AIML model, it can send a corresponding request to the LMF to receive it. Alternatively, the broadcast message can include a link to a location from which the UE can download the AIML model. This location can be considered to belong to the central unit.

[0117] When the LMF knows that the UE has received an AIML model prepared (created, trained, and validated) by the LMF (e.g., because the UE requested the feature from the LMF), the LMF can request a switch to the feature, such as switching from assisted AIML localization to direct AIML localization. The LMF can provide the UE with a list of suitable inputs for the AIML model. This list can be based on the UE's capabilities. Once the feature is deployed at the UE, initial fine-tuning and inference may be required to ensure functional performance. This is primarily to mitigate potential impairments on the UE and NW sides.

[0118] As an alternative, (multiple) UEs can proactively request new capabilities / features from the network. Based on such requests from UEs (e.g., a subset of them) within a given region of interest, the NW can instantiate a reflection function and train / test / validate it before providing the reflection function to (multiple) corresponding UEs.

[0119] In some example embodiments, when a UE receives a prepared function, the UE checks whether the function is actually executable at the UE before attempting to execute it. For example, some processing power and / or memory deemed available for the function may be occupied by other tasks simultaneously. If the UE observes that its own conditions do not meet the requirements of the function, the UE does not execute the function upon receiving it.

[0120] In some example embodiments, the UE collects data (specifically, performance data) and provides it to the LMF before the function becomes executable at the UE. The LMF uses this data to prepare the function. Once the function becomes executable at the UE, the UE can use this data itself to perform the function at the UE. In some example embodiments, the UE no longer provides this data to the LMF, thereby saving signaling workload.

[0121] Due to the time-varying nature of the wireless environment, reflection functions trained at the LMF and used at UEs in the NW may become obsolete after a period of time. To ensure the desired performance (e.g., estimation accuracy) from the reflection function, the LMF can continue to train the reflection function using measurement reports from UEs that do not have the reflection function implemented. These UEs may have The UE may have recently entered the NW after the function was broadcast, or may be unable to use the function due to minimum capability requirements, or may be able to use the function but still provide auxiliary measurements to monitor the function being used. In some example embodiments, a UE using a reflection function may not provide any auxiliary measurements to the LMF to save signaling workload.

[0122] At certain intervals, features that have been further trained (or an indication of the availability of features that have been further trained) can be broadcast to all UEs to update their respective models. To reduce overhead, LMF can provide differential (incremental) updates to reduce signaling overhead.

[0123] Therefore, compared to a full feature transfer, updates can occur at certain intervals to reduce the overhead for UEs that have already deployed the trained features. Since the features are created using inputs already provided by the UE for auxiliary information, the processing overhead involved in creating the inputs to the model can be quite small. However, the benefits of using new direct AIML features with NW assistance for UEs can be attributed to the following reasons.

[0124] 1. Frequent reporting to NW can be reduced, and reports can be used only when NW requests functional monitoring. 2. It can provide the latest location information to the UE-side application layer without NW signaling, thereby reducing the location estimation error caused by delays due to control plane signaling. In the above example embodiments, the UE indicates to the central unit that it is capable of executing the prepared AIML function. In some example embodiments, the UE may additionally indicate to the LMF whether the UE is capable of preparing the AIML function. In some example embodiments, the UE may generally also be capable of preparing the AIML function (e.g., from the perspective of installed processing power and memory), but the UE lacks some information. In such example embodiments, the central unit may provide the UE with auxiliary information so that the UE can prepare (collect data, train the model, test the model, and validate the model) the AIML function. That is, in these example embodiments, preparation can be performed at the UE after the central unit provides some auxiliary information to the UE. Therefore, the signaling workload is further reduced. After the UE prepares the function, the UE can execute the function.

[0125] For example, auxiliary information could include configuring the UE to measure and evaluate a reference signal that was used for training but has not yet been measured by the UE. As another example, the network could reduce the amount of UE processing power and / or memory used by another task performed at the UE, so that the UE's conditions meet the requirements for readiness functions.

[0126] If the central unit learns from the capability information provided by the UE that the UE cannot prepare the function, it may not provide auxiliary information to the UE.

[0127] Figure 3 An apparatus according to an example embodiment is shown. The apparatus may be a central unit (such as an LMF or NWDAF) or an element thereof. Figure 4 A method according to an example embodiment is shown. Figure 3 The device can perform Figure 4 This method is applicable, but not limited to this method. Figure 4 The method can be derived from Figure 3 The device performs the action, but is not limited to the device performing the action.

[0128] The device includes a preparation component 110, a provisioning component 120, and a sharing component 130. The preparation component 110, the provisioning component 120, and the sharing component 130 can be a preparation component, a provisioning component, and a sharing component, respectively. The preparation component 110, the provisioning component 120, and the sharing component 130 can be a preparer, a provider, and a sharer, respectively. The preparation component 110, the provisioning component 120, and the sharing component 130 can be a preparation processor, a provisioning processor, and a sharing processor, respectively.

[0129] The component 110 used for preparation is prepared by the central unit preparation function (S110). This function is based on an ML model. Preparation may include creating, training, and validating the ML model or a subset of these activities.

[0130] If the functionality is ready, at least one of the following can be executed: The component 120 provided can provide the prepared functions to the terminal for execution by the terminal (S120); or The component 130 for sharing can share (e.g., broadcast) prepared functions or availability indications (S130). The availability indication indicates that the functions are ready.

[0131] Figure 5 An apparatus according to an example embodiment is shown. The apparatus may be a central unit (such as an LMF or NWDAF) or an element thereof. Figure 6 A method according to an example embodiment is shown. Figure 5 The device can perform Figure 6 This method is applicable, but not limited to this method. Figure 6 The method can be derived from Figure 5 The device performs the action, but is not limited to the device performing the action.

[0132] The device includes a receiving component 210 and a providing component 220. The receiving component 210 and the providing component 220 can be a receiving component and a providing component, respectively. The receiving component 210 and the providing component 220 can be a receiver and a provider, respectively. The receiving component 210 and the providing component 220 can be a receiving processor and a providing processor, respectively.

[0133] The receiving component 210 receives a request for a function from the terminal (S210). The providing component 220, in response to the request, provides assistance to the terminal (S220), enabling the terminal to prepare for the function. If the terminal is configured and configured under a set of conditions, the function includes configuration for the terminal and methods executable by the terminal; and the function also includes at least one of the following: Direct positioning of the terminal or auxiliary positioning of the terminal.

[0134] The preparation function includes at least one of the following: - Collect data for training ML models; - Training the ML model; and - Validate the trained ML model; Figure 7 An apparatus according to an example embodiment is shown. The apparatus may be a terminal (such as a UE or MTC device) or a component thereof. Figure 8 A method according to an example embodiment is shown. Figure 7 The device can perform Figure 8 This method is applicable, but not limited to this method. Figure 8 The method can be derived from Figure 7 The device performs the action, but is not limited to the device performing the action.

[0135] The device includes a first component 310 and a second component 320 for determining, a component 330 for requesting, a component 340 for receiving, and a component 350 for enabling. The first component 310 and the second component 320 for determining, the component 330 for requesting, the component 340 for receiving, and the component 350 for enabling can be respectively a first determining component and a second determining component, a requesting component, a receiving component, and an enabling component. The first component 310 and the second component 320 for determining, the component 330 for requesting, the component 340 for receiving, and the component 350 for enabling can be respectively a first determining processor and a second determining processor, a requesting processor, a receiving processor, and an enabling processor.

[0136] The first component 310 for determination determines a function that is not executable but is expected to be executable (S310). The second component 320 for determination determines whether the capability to prepare the expected function is available (S320). If the device is configured and the configuration is under a set of conditions, the function includes the configuration and a method that can be performed by the device. The function also includes at least one of the following: direct positioning or assisted positioning; Component 330, which makes a request, determines that the capability is unavailable (S330) and requests the desired function. Component 340, which receives the desired function (S340). Component 350, which enables the received function, makes it executable (S350).

[0137] Figure 9An apparatus according to an example embodiment is shown. The apparatus may be a terminal (such as a UE or MTC device) or a component thereof. Figure 10 A method according to an example embodiment is shown. Figure 9 The device can perform Figure 10 This method is applicable, but not limited to this method. Figure 10 The method can be derived from Figure 9 The device performs the action, but is not limited to the device performing the action.

[0138] The device includes a first component 410 and a second component 420 for determining, a component 430 for requesting, a component 440 for receiving, and a component 450 for preparing. The first component 410 and the second component 420 for determining, the component 430 for requesting, the component 440 for receiving, and the component 450 for preparing can be respectively a first determining component and a second determining component, a requesting component, a receiving component, and a preparing component. The first component 410 and the second component 420 for determining, the component 430 for requesting, the component 440 for receiving, and the component 450 for preparing can be respectively a first determining unit and a second determining unit, a requesting unit, a receiving unit, and a preparing unit. The first component 410 and the second component 420 for determining, the component 430 for requesting, the component 440 for receiving, and the component 450 for preparing can be respectively a first determining processor and a second determining processor, a requesting processor, a receiving processor, and a preparing processor.

[0139] The first component 410 determines a function that is not executable but is expected to be executable (S410). The second component 420 determines whether the capability to prepare the function is available (S420).

[0140] The requesting component 430, based on the determination that the capability is unavailable (S430), requests assistance to make the capability available. The receiving component 440 receives the assistance (S440). The preparing component 450, based on the assistance (S450), prepares the function.

[0141] If the device is configured and the configuration is under a set of conditions, then the function includes configuration and a method that can be performed by the device. The function also includes at least one of the following: Direct positioning or assisted positioning.

[0142] The preparation function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Test the ML model; and - Validate the trained ML model.

[0143] Figure 11 illustrates an apparatus according to an example embodiment. The apparatus includes at least one processor 810 and at least one memory 820 storing instructions that, when executed by the at least one processor 810, cause the apparatus to perform at least one of the methods according to the following figures and related descriptions: Figure 4 ,or Figure 6 ,or Figure 8 ,or Figure 10 .

[0144] The explanation of 5G (NR) includes some example implementations. However, other example implementations may be used in other 3GPP generations, such as 4G, 6G, 7G, etc.

[0145] UE is an example of a terminal. Other examples are MTC devices. Each terminal can be implemented as a smartphone, mobile phone, laptop computer, sensor device, etc.

[0146] LMF is an example of a central cell. Another example of a central cell is NWDAF. For the purposes of this application, unless otherwise stated or clear from the context, LMF, NWDAF, and central cell may be considered equivalent.

[0147] Some example embodiments utilize location as a related function for explanation. However, some example embodiments may provide another related function, such as CSI compression, particularly CI compression with a two-sided model, BM with a UE-side model, and CSI prediction with a UE-side model.

[0148] A message can be sent from one entity to another in one or more messages. Each of these messages may include additional (different) pieces of information.

[0149] The names of network elements, network functions, protocols, and methods are based on the current standard. In other versions or other technologies, the names of these network elements and / or network functions and / or protocols and / or methods may differ, as long as they provide the corresponding functionality. This also applies to terminals.

[0150] Unless otherwise stated or clearly understood from the context, two entities being distinct statements means that they perform different functions. It does not necessarily mean they are based on different hardware. That is, each entity described in this specification may be based on different hardware, or some or all entities may be based on the same hardware. It does not necessarily mean they are based on different software. That is, each entity described in this specification may be based on different software, or some or all entities may be based on the same software. Each entity described in this specification may be deployed in the cloud.

[0151] Therefore, based on the above description, it is evident that the exemplary embodiments provide, for example, a central unit (such as an LMF or NWDAF) or a component thereof, means embodying such a unit, methods for controlling and / or operating such a unit, and a medium for controlling and / or operating such computer programs and forming a computer program product. Therefore, based on the above description, it is evident that the exemplary embodiments provide, for example, a terminal (such as a UE, MTC device, etc.) or a component thereof, means embodying such a terminal, methods for controlling and / or operating such a terminal, and a medium for controlling and / or operating such computer programs and forming a computer program product.

[0152] By way of non-limiting example, implementations of any of the foregoing boxes, apparatuses, systems, techniques, or methods include implementations as hardware, software, firmware, special-purpose circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof. Each entity described in this specification may be embodied in the cloud.

[0153] It should be understood that the content described above is what is currently considered to be a preferred exemplary embodiment. However, it should be noted that the description of the preferred exemplary embodiment is given by way of example only, and various modifications can be made without departing from the scope of this disclosure as defined by the appended claims.

[0154] Unless otherwise stated, the terms "first X" and "second X" include the same options as "first X" and "second X" as well as options where "first X" differs from "second X". As used herein, "at least one of the following: " and "at least one of " and similar wording, where the list of two or more elements is connected by "and" or "or", means at least any one of the elements, or at least any two or more of the elements, or at least all of the elements.

Claims

1. An apparatus comprising: One or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the device to perform: Preparation function; and at least one of the following: Provide the prepared functions to the terminal; or The broadcast indicates the availability of the prepared features or serves as a notification that the features are prepared. in If the terminal is configured and the configuration is under a set of conditions, then the functionality includes the configuration for the terminal and a method that can be executed by the terminal; Furthermore, the aforementioned functionality also includes at least one of the following: The direct location of the terminal, or The terminal's assisted positioning; and The preparation of the aforementioned function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Test the ML model; - Validate the trained ML model; or - Update the ML model.

2. The apparatus of claim 1, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: The ML model is trained based on the data received from the terminal.

3. The apparatus according to any one of claims 1 and 2, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Receive corresponding data from each of the multiple terminals; The ML model is trained based on the data received from the multiple terminals.

4. The apparatus according to any one of claims 1 to 3, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Determine the set of conditions; The set of conditions is sent together with at least one of the functions of broadcasting and providing the functions.

5. The apparatus according to any one of claims 1 to 4, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Receive a request to provide the prepared functions to the terminal; In response to the request, the prepared functions are provided to the terminal; Without the requested functionality, the prepared functions shall be disabled from being provided to the terminal.

6. The apparatus of claim 5, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Receive the set of one or more conditions from the terminal.

7. The apparatus according to any one of claims 1 to 6, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: The terminal is requested to enable the prepared function to be executed.

8. The apparatus according to any one of claims 1 to 7, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: After the function has obtained the updated ML model, continue preparing the function; and at least one of the following: Provide the updated function or an incremental update of the function to the terminal; or Broadcast the updated feature or the incremental update of the feature, or an update availability indication indicating that the updated feature or the incremental update is available.

9. An apparatus comprising: One or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the device to perform: Receive requests for functions from the terminal; In response to the request, assistance is provided to the terminal, enabling the terminal to prepare for the function; in If the terminal is configured and the configuration is under a set of conditions, then the functionality includes the configuration for the terminal and a method that can be executed by the terminal; Furthermore, the aforementioned functionality also includes at least one of the following: The direct location of the terminal, or The terminal's assisted positioning; and The preparation of the aforementioned function includes at least one of the following: - Collect data for training ML models; - Train the ML model; and - Validate the trained ML model.

10. The apparatus of claim 9, wherein providing the assistance includes configuring the terminal to receive a reference signal, and collecting the data includes evaluating the received reference signal.

11. The apparatus according to any one of claims 9 to 10, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: A set of conditions to receive the terminal; Determine whether the function can be performed by the terminal under the given set of conditions; Based on the determination that the function cannot be properly performed by the terminal under the given set of conditions, the provision of the assistance is prohibited.

12. An apparatus comprising: One or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the device to perform: A function that is determined to be non-executable but is expected to be executable; Determine whether the capability to prepare the desired functionality is available; Based on the determination that the capability is unavailable, a request is made to provide the desired functionality; Receive the desired functionality; To enable the received function to be executed; in If the device is configured and the configuration is under a set of conditions, the function includes the configuration and a method executable by the device; the function also includes at least one of the following: Direct positioning or assisted positioning.

13. The apparatus of claim 12, wherein the required capability comprises at least one of the following: - The ability to collect data, which is used for at least one of the training, testing and validation of the ML model; - Processing power, used for at least one of the training, testing, and validation of the ML model; - The availability of the processing power for at least one of the training, testing and validation of the ML model; - A memory for performing at least one of the training, testing and validation of the ML model; - Availability of the memory for at least one of the training, testing, and validation of the ML model; or - The ability to obtain true values ​​for verifying the ML model.

14. The apparatus according to any one of claims 12 to 13, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Determine one or more sets of conditions for making the function executable; Together with the requested function, send the set of one or more conditions.

15. The apparatus according to any one of claims 12 to 14, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: In response to receiving a broadcast availability indication that the function is ready, the function is requested.

16. The apparatus according to any one of claims 12 to 15, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Before the function is implemented, data for preparing the function is provided, wherein the data is based on the measurements performed; When making the function executable, the data used to perform the function is used.

17. The apparatus according to any one of claims 12 to 16, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: In addition to receiving the aforementioned function, a set of one or more conditions are received to enable the method to perform; Determine whether one or more of the conditions are met.

18. The apparatus according to any one of claims 12 to 17, wherein the instructions, when executed by the one or more processors, further cause the apparatus to perform: Receive incremental updates for the aforementioned function; The function is updated based on the incremental update.

19. An apparatus comprising: One or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the device to perform: A function that is determined to be non-executable but is expected to be executable; Determine whether the capability to prepare the aforementioned function is available; Based on the determination that the capability is unavailable, request assistance to make the capability available; Receive the assistance; Based on the aforementioned assistance, prepare the aforementioned function; in If the device is configured and the configuration is under a set of conditions, the function includes the configuration and a method executable by the device; and the function further includes at least one of the following: Direct positioning or assisted positioning; The preparation of the aforementioned function includes at least one of the following: - Collect data for training ML models; - Train the ML model; - Test the ML model; and Validate the trained ML model.

20. The apparatus of claim 19, wherein providing the assistance includes configuring the terminal to receive a reference signal, and collecting the data includes evaluating the received reference signal.

Images