Training and inference for AI-based positioning

Abstract

A user equipment (UE) is configured to receive, from a location management function (LMF), a configuration for an artificial intelligence (AI)-based UE positioning method using a neural network (NN), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for downlink (DL) reference signal (RS) reception on the one or more positioning cells; estimating a channel response for the DL RS received from the one or more positioning cells; sending the channel response as an NN inference input to the LMF, wherein the LMF estimates a UE location; and receiving a UE position estimate from the LMF.

Description

Technical Field

[0001] The present application relates generally to wireless communications, and in particular to training and inference for AI-based positioning. Background Art

[0002] 5G New Radio (NR) introduces many radio access network (RAN) and core network (CN) enhancements as well as an enhanced security architecture. Artificial intelligence (AI) and / or machine learning (ML) processes, such as deep learning neural networks, can be used to enhance the operation of the air interface. Use cases for AI / ML for the air interface include channel state information (CSI) feedback enhancements (e.g., overhead reduction, improved accuracy, prediction); beam management (e.g., beam prediction in the time and / or spatial domain for overhead and latency reduction, beam selection accuracy improvement); and positioning accuracy enhancements for different scenarios, including, for example, those with heavy no-site line (NLOS) conditions. Summary of the invention

[0003] Some exemplary implementation aspects relate to a processor of a user equipment (UE), the processor being configured to perform operations. The operations include: receiving a configuration for an artificial intelligence (AI)-based UE positioning method using a neural network (NN) from a location management function (LMF), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells; estimating a channel response for the DL RS received from the one or more positioning cells; sending the channel response as an NN inference input to the LMF, wherein the LMF estimates a UE position; and receiving a UE position estimate from the LMF.

[0004] Other exemplary embodiments relate to a location management function (LMF) of a cellular core network, the location management function (LMF) being configured to perform operations. The operations include: training a neural network (NN) using a first training data set for an artificial intelligence (AI) based user equipment (UE) positioning method; receiving a request for AI based positioning from a UE; transmitting to the UE a NN inference input request including a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells; receiving from the UE a channel response for DL ​​RS received from one or more positioning cells, wherein the channel response is a NN inference input; and estimating a UE position using the NN based on the NN inference input.

[0005] A further exemplary embodiment relates to a processor of a user equipment (UE) configured to perform operations including: receiving a configuration for an artificial intelligence (AI) based UE positioning method using a neural network (NN) from a location management function (LMF), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for uplink (UL) reference signal (RS) transmission to one or more positioning cells; transmitting UL RS to one or more positioning cells; and receiving a UE position estimate from the LMF. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 Network arrangements according to various exemplary embodiments are shown.

[0007] Figure 2 An exemplary UE according to various exemplary embodiments is shown.

[0008] Figure 3 A signaling diagram for AI-based UE positioning according to a first option is shown, wherein the training and inference phases of the AI ​​process are performed at the UE, according to various exemplary embodiments.

[0009] Figure 4 A signaling diagram for AI-based UE positioning according to the second option is shown according to various exemplary embodiments, where the training and inference phases of the AI ​​process are performed at the Location Management Function (LMF).

[0010] Figure 5 A signaling diagram for AI-based UE positioning according to the third option according to various exemplary embodiments is shown, wherein the training phase of the AI ​​process is performed at the LMF and the inference phase of the AI ​​process is performed at the UE.

[0011] Figure 6 A signaling diagram for channel response acquisition at a UE for identifying neural network (NN) inference inputs for UE positioning according to various exemplary embodiments is shown.

[0012] Figure 7 A signaling diagram for channel response acquisition at one or more positioning transmit and receive points (TRPs) / next generation NodeBs (gNBs) for identifying NN inference inputs for UE positioning at an LMF is shown according to various exemplary embodiments.

[0013] Figure 8 Methods for AI-based UE positioning using NN according to various exemplary embodiments are shown, wherein the training and inference phases of the AI ​​process are performed at the UE.

[0014] Fig. 9 A method for AI-based UE positioning using a NN according to various exemplary embodiments is shown, wherein the training and inference phases of the AI ​​process are performed at the LMF.

[0015] Fig.10 A method for AI-based UE positioning using a neural network (NN) according to various exemplary embodiments is shown, wherein a training phase of the AI ​​process is performed at the LMF and an inference phase of the AI ​​process is performed at the UE. DETAILED DESCRIPTION

[0016] The exemplary aspects may be further understood with reference to the following description and associated drawings, in which similar elements have the same figure numbers. The exemplary aspects describe operations for supporting artificial intelligence (AI) and / or machine learning (ML) based channel estimation in wireless networks. In some aspects, a deep learning neural network (NN) is used to enhance a positioning method for locating a user equipment (UE). According to various embodiments, a training phase of the NN and an inference phase of the trained NN may be performed at a location management function (LMF) of a UE and / or a 5G new radio (NR) core network (CN). In other aspects, operations for obtaining a channel response used as an input for the inference phase of a trained NN are described.

[0017] As described above, exemplary embodiments are described with reference to the LMF of the 5G New Radio (NR) Core Network (CN). However, it should be understood that the operations described herein are not limited to the LMF residing in the core network. For example, some positioning operations may be performed via a sidelink (SL) connection (e.g., SL positioning) in out-of-coverage scenarios. In these scenarios or any other scenarios, the LMF may reside outside the core network (e.g., SL LMF).

[0018] The exemplary aspects are described with reference to a UE. However, the use of a UE is provided for illustration purposes. The exemplary aspects can be utilized with any electronic component that can establish a connection with a network and is configured with hardware, software, and / or firmware for exchanging information and data with the network. Therefore, a UE as described herein is used to represent any electronic component that can access a wireless network and perform an AI-based positioning method to locate the UE.

[0019] The exemplary aspects are described with reference to the network being a 5G New Radio (NR) network and the base station being a next generation Node B (gNB). However, for illustrative purposes, the use of a 5G NR network and a gNB is provided. The exemplary aspects may be applicable to any type of network utilizing similar functionality. For example, some positioning methods described herein may be RAT independent.

[0020] Exemplary embodiments are further described with respect to location estimation based on artificial intelligence (AI) and / or machine learning (ML). Depending on the UE and network implementation, any number of different AI / ML models may be used. For example, a deep learning neural network may be used. In addition, various types of models may use different types of channel response data for training the model, and different types / densities of RS for the inference phase of location estimation. Therefore, for illustrative purposes, references to any particular AI-based location estimation model are provided. The exemplary aspects are applicable to any type of AI-based location estimation model using a training phase and an inference phase that may be performed at a UE and / or a core network element (e.g., a location management function (LMF)).

[0021] An exemplary embodiment is further described with respect to a location management function (LMF) of a 5G core network (5GC) for supporting location determination of a target UE. As will be further described below, in the exemplary aspects described herein, the LMF may perform operations to facilitate AI-based estimation of the UE location, including: performing either or both of an inference phase and a training phase of a positioning NN for locating the target UE; configuring a RAN node and / or UE to send / receive a reference signal (RS) for channel response acquisition; and receiving / sending position estimation and / or channel response estimation feedback to / from the target UE. It should be understood that the LMF is not required to be in the core network. For example, the LMF may reside on a separate server connected to the 5GC or may reside within the 5G RAN.

[0022] Throughout the specification, it will be described that the channel response can be used to train the AI ​​model, as an input to the AI ​​model, etc. It should be understood that the exemplary embodiments are not limited to the channel response. When the term channel response is used, it may also refer to other types of data that can be used to train the AI ​​model or as an input to the model, such as channel layer 1 received reference signal power (L1-RSRP), channel power delay distribution, a combination of descriptive inputs, or any other input useful for determining the location of the UE.

[0023] In addition, throughout the specification, a gNB may be referred to as a "serving cell". A gNB acting as a serving cell is a cell to which the UE is currently connected, for example, the UE may be in a radio resource control (RRC) connected state with the gNB and may actively exchange data and / or control information with the base station. A gNB may also be referred to as a "positioning gNB", "positioning node" or "positioning cell". A gNB acting as a positioning cell is a base station that assists in positioning the UE, for example, sending a positioning reference signal (PRS) to the UE to assist in positioning the UE. A gNB may simultaneously act as a serving cell and a positioning cell relative to the UE, or may only act as a positioning cell for the UE. In addition, throughout the specification, a gNB may be referred to as a "neighboring cell" or "adjacent cell". According to the present disclosure, a neighboring cell may not act as a serving cell for the UE, however certain signals may be exchanged between the neighboring cell and the UE without entering an RRC connected state. One or more neighboring cells may act as additional positioning gNBs to assist in positioning the UE.

[0024] A serving cell or a neighboring cell acting as a positioning gNB may include one or more transmit and receive points (TRPs), for example, a first TRP and a second TRP. One or more of the TRPs located at a specific positioning gNB may be used in the exemplary positioning method and may be referred to as a "positioning TRP". Multiple positioning TRPs may be located at a single positioning gNB.

[0025] In addition, throughout the specification, various positioning methods based on UE and / or based on UE-assisted LMF are described. These positioning methods may include downlink (DL)-based positioning, uplink (UL)-based positioning, or combined DL+UL-based positioning, including but not limited to: Assisted Global Navigation Satellite System (A-GNSS) (GPS); Wireless Local Area Network (WLAN); Terrestrial Beacon System (TBS); Downlink Time Difference of Arrival (DL-TDOA); DL Angle of Departure (DL-AoD); and Multiple Round Trip Time (Multiple RTT). In some positioning methods, a positioning reference signal (PRS) is sent to a UE on the DL from each of a plurality of network nodes (i.e., positioning transmission and reception points (TRPs)) so that the DL arrival timing of each PRS at the UE can be determined, and the position of the UE is determined based thereon. In addition, other types of reference signals may be used on the DL to locate the UE, including a channel state information reference signal (CSI-RS) and / or a synchronization signal block (SSB), such as a demodulation RS (DMRS). Uplink (UL) reference signals may also be used to locate the UE, including a sounding reference signal (SRS). Those skilled in the art will appreciate that these reference signals may also be used for purposes other than locating the UE. Therefore, the RS described herein is not limited to any particular type of reference signal.

[0026] Figure 1An exemplary network arrangement 100 according to various exemplary embodiments is shown. The exemplary network arrangement 100 includes a user equipment (UE) 110. Those skilled in the art will appreciate that a UE may be any type of electronic component configured to communicate via a network, such as a mobile phone, a tablet, a smart phone, a phablet, an embedded device, a wearable device, a Cat-M device, a Cat-M1 device, an MTC device, an eMTC device, other types of Internet of Things (IoT) devices, etc. It should also be understood that an actual network arrangement may include any number of UEs used by any number of users. Therefore, for purposes of illustration, only an example with a single UE 110 is provided.

[0027] UE 110 may communicate directly with one or more networks. In the example of network configuration 100, the networks with which UE 110 may wirelessly communicate are 5G NR radio access network (5G NR-RAN) 120, LTE radio access network (LTE-RAN) 122, and wireless local area network (WLAN) 124. Therefore, UE 110 may include a 5G NR chipset that communicates with 5G NR-RAN 120, an LTE chipset that communicates with LTE-RAN 122, and an ISM chipset that communicates with WLAN 124. However, UE 110 may also communicate with other types of networks (e.g., legacy cellular networks), and UE 110 may also communicate with the network via a wired connection. Referring to the exemplary aspect, UE 110 may establish a connection with 5G NR-RAN 122.

[0028] 5G NR-RAN 120 and LTE-RAN 122 may be parts of cellular networks that may be deployed by cellular providers (e.g., Verizon, AT&T, T-Mobile, etc.). These networks 120, 122 may include, for example, cells or base stations (NodeB, eNodeB, HeNB, eNBS, gNB, gNodeB, macrocell base stations, microcell base stations, small cell base stations, femtocell base stations, etc.) configured to transmit and receive traffic from UEs equipped with appropriate cellular chipsets. WLAN 124 may include any type of wireless local area network (WiFi, hotspot, IEEE 802.11x network, etc.).

[0029] UE 110 may be connected to 5G NR-RAN via at least one of next generation nodeB (gNB) 120A and / or gNB 120B. Reference to two gNBs 120A, 120B is for illustrative purposes only. The exemplary aspects may be applied to any suitable number of gNBs.

[0030] In addition to the networks 120, 122, and 124, the network arrangement 100 also includes a cellular core network 130, the Internet 140, an IP multimedia subsystem (IMS) 150, and a network service backbone 160. The cellular core network 130 (e.g., the 5GC of the 5G NR network) can be viewed as an interconnected collection of components that manage the operation and traffic of the cellular network. The cellular core network 130 also manages the traffic flowing between the cellular network and the Internet 140. The core network 130 may include a location management function (LMF) 131 to support location determination of the UE, as will be further described below.

[0031] The IMS 150 may be generally described as an architecture for delivering multimedia services to the UE 110 using the IP protocol. The IMS 150 may communicate with the cellular core network 130 and the Internet 140 to provide multimedia services to the UE 110. The network service backbone 160 communicates directly or indirectly with the Internet 140 and the cellular core network 130. The network service backbone 160 may be generally described as a collection of components (e.g., servers, network storage arrangements, etc.) that implement a set of services that may be used to extend the functionality of the UE 110 to communicate with various networks.

[0032] Figure 2 An exemplary UE 110 is shown according to various exemplary embodiments. Figure 1 100 is used to describe the UE 110. The UE 110 may represent any electronic device and may include a processor 205, a memory arrangement 210, a display device 215, an input / output (I / O) device 220, a transceiver 225, and other components 230. The other components 230 may include, for example, an audio input device, an audio output device, a battery providing a limited power source, a data acquisition device, a port for electrically connecting the UE 110 to other electronic devices, a sensor for detecting the status of the UE 110, etc. In addition, the UE 110 may be configured to access the SNPN.

[0033] The processor 205 may be configured to execute multiple engines of the UE 110. For example, the engines may include a position estimation engine 235 for performing operations related to executing a training phase and / or an inference phase of a positioning neural network (NN) and obtaining a channel response as an input to the inference phase, which will be described in more detail below.

[0034] The above-described engine as an application (e.g., program) executed by the processor 205 is merely exemplary. The functionality associated with the engine may also be represented as an independently integrated component of the UE 110, or may be a modular component coupled to the UE 110, such as an integrated circuit with or without firmware. For example, an integrated circuit may include an input circuit for receiving a signal and a processing circuit for processing the signal and other information. The engine may also be embodied as an application or an independent application. In addition, in some UEs, the functions described for the processor 205 are split between two or more processors such as a baseband processor and an application processor. The exemplary aspects may be implemented in any of these or other configurations of the UE.

[0035] The memory 210 may be a hardware component configured to store data related to operations performed by the UE 110. The display device 215 may be a hardware component configured to display data to a user, and the I / O device 220 may be a hardware component that enables a user to perform input. The display device 215 and the I / O device 220 may be separate components or may be integrated together (such as a touch screen). The transceiver 225 may be a hardware component configured to establish a connection with the 5G-NR RAN 120, the LTE RAN 122, etc. Therefore, the transceiver 225 may operate on a variety of different frequencies or channels (e.g., a set of continuous frequencies).

[0036] An exemplary network base station, in this case gNB 120A, may represent a serving cell for UE 110. gNB 120A may represent any access node belonging to a 5G NR network that UE 110 may use to establish a connection and manage network operations. gNB 120A may include a processor, a memory arrangement, an input / output (I / O) device, a transceiver, and other components. Other components may include, for example, an audio input device, an audio output device, a battery, a data acquisition device, a port for electrically connecting gNB 120A to other electronic devices, and the like. Functionality associated with the processor of gNB 120A may also be represented as a separate integrated component of gNB 120A, or may be a modular component coupled to gNB 120A, such as an integrated circuit with or without firmware. For example, an integrated circuit may include input circuitry for receiving signals and processing circuitry for processing signals and other information. In addition, in some gNBs, the functionality described for the processor is split between multiple processors (e.g., a baseband processor, an application processor, etc.). The exemplary aspects may be implemented in any of these or other configurations of the gNB.

[0037] The memory may be a hardware component configured to store data related to operations performed by the UE 110, 112. The I / O device may be a hardware component or port that enables a user to interact with the gNB 120A. The transceiver may be a hardware component configured to exchange data with the UE 110 and any other UE in the system 100. The transceiver may operate on a variety of different frequencies or channels (e.g., a set of contiguous frequencies). Thus, the transceiver may include one or more components (e.g., a radio component) to enable data exchange with various networks and UEs.

[0038] As described above, artificial intelligence (AI) and / or machine learning (ML) (e.g., deep learning neural networks (NN)) may be used to enhance operations over the air interface, including positioning accuracy enhancements for different scenarios, including, for example, those with heavy no-site line (NLOS) conditions.

[0039] The general process of neural network-based operation includes: a training phase, in which training data is input into the NN to train the model; and an inference phase, in which inference data is input into the trained NN for the NN to perform estimation based on the input data. In addition, a calibration phase can be used to evaluate the error of the trained NN, and if the error is high enough, retrain the NN using different training data.

[0040] As described above, LMF refers to a 5G core network (5GC) entity that supports location determination of a target UE. The LMF may perform various operations to facilitate AI-based location estimation, including: configuring the UE for AI-based UE positioning; performing NN training / inference operations for UE positioning; configuring the RAN node (serving cell and / or positioning cell / TRP) and / or the UE to send / receive reference signals (RS) acquired for channel response; sending / receiving feedback from the UE and / or gNB to inform the location estimate; and other operations are described in detail below. It should be understood that the LMF is a core network entity, and messages between the UE and the LMF sent via the positioning protocol are routed via the gNB. In addition, in some aspects, messages are sent directly from the UE or LMF to the gNB (serving gNB or positioning cell), or received from the gNB at the UE or LMF. Therefore, the signaling exchange may refer to the LMF or gNB as a sending / receiving entity for communicating with the UE.

[0041] According to various exemplary embodiments described herein, the training and / or inference phase of AI-based channel measurement may be performed at the LMF or the UE. In a first option, both training and inference are performed at the UE. In a second option, both training and inference are performed at the LMF. In a third option, training is performed at the LMF and inference is performed at the UE. In a fourth option, training is performed at the UE and inference is performed at the LMF.

[0042] According to a first option, both training and inference are performed at the UE.Performing both phases at the UE provides maximum UE privacy with respect to further options to be described below.

[0043] Figure 3 A signaling diagram 300 for AI based UE positioning according to a first option is shown, wherein the training and inference phases of the AI ​​process are performed at the UE, according to various exemplary embodiments. Diagram 300 shows one example of the first option, however, various alternative embodiments will also be described.

[0044] Block 305 shows a capability signaling exchange between the UE and the gNB / LMF to indicate the UE and network capabilities. For example, LMF 131 may signal support for AI-based positioning to UE 110, and UE 110 may report the AI-based positioning capability to LMF 131. The LMF may configure the UE for the AI-based positioning procedure.

[0045] Box 310 shows the training phase of the AI-based positioning neural network (NN). The UE and LMF may enter an initial training phase after the capability exchange. In some embodiments, as shown in 310, the UE transmits a training data request to the LMF. In response, the UE receives a training data set from the LMF, and the UE trains the NN based on the received training data set. The training data set may include information related to a specific location, channel conditions at the location, positioning parameters for various positioning methods including reference signals, etc.

[0046] In one embodiment, after the initial capability exchange of 305, the UE may request a first training data set, e.g., data set index x=0, and the LMF may provide data set x=0 in response to the request. In another embodiment, the initial request is optional, and the LMF may transmit the first training data set to the UE without a specific request from the UE. As will be described in further detail below, if retraining of the NN is desired, the request step 310 may be repeated and a different training set x (x=0, ..., N) may be indicated to calibrate the NN and reduce model error. It should be understood that the various training data sets that may be requested and received in 310 may be similar or different in composition.

[0047] Box 310 of Figure 300 shows an example of a UE requesting and receiving training data for positioning a NN from an LMF via a gNB of a 5G RAN. However, other options may be used to train the NN. In an alternative embodiment, the UE trains the NN using historical data. For example, the historical data may relate to parameters of a positioning method previously used by the UE. In another alternative embodiment, the UE may download training data from a reference point (e.g., a positioning reference defined in the network and associated with a location) or a kiosk. To provide an illustrative example, the UE may access a kiosk / terminal in a location (e.g., an airport) that may have poor positioning performance when AI is not used. The UE may communicate with the kiosk to receive a high-speed download of training data for the NN. This may be performed while the UE is connected to the 5G RAN or offline relative to the 5G RAN. When the UE identifies the training data, the UE may train the NN.

[0048] Box 315 shows an optional calibration phase for the AI-based positioning neural network. This calibration can be performed in various scenarios, for example, when the UE is recently powered on in a new location, and does not need to be performed after each training phase. For the calibration phase, the UE can use existing positioning mechanisms to determine a calibration position for comparison with the AI-based position estimate. The accuracy of the AI ​​model relative to the calibration position can be evaluated within a certain error threshold. For example, in some embodiments, the UE may identify a calibration position [X, Y] based on either a UE-based RAT-independent technology (e.g., GNSS or GPS) or (b) a RAT-dependent technology (e.g., time difference of arrival (TDOA)) defined in 3GPP Rel-16 or Rel-17. In other embodiments, the calibration position [X, Y] may be manually downloaded, for example, from a positioning reference unit defined in the network and associated with the position described above.

[0049] To perform AI-based position determination, the UE requests neural network inference input for a calibration phase including channel response (e.g., L1-RSRP, beam direction, etc.). Figure 6 Described in more detail, the request may indicate a specific TRP, frequency, or RS to be sent to the UE so that the UE can estimate the channel response. The channel response may be used as input data for the calibration inference phase of the NN to estimate the UE position and the error relative to the calibrated position. The UE may determine a single position, or the determination may be based on multiple tests of the NN. If the error is greater than a predetermined threshold, the UE may return to the training phase of 310 and send a new request for a training data set (e.g., data set index x=1).

[0050] In alternative embodiments, the reception of the training and / or calibration positions may be done offline.In some embodiments, the LMF may transmit the training data and the calibration data (eg, calibration positions) in a combined transmission.

[0051] Box 320 shows the inference phase of the AI-based positioning neural network. After the NN is trained and optionally calibrated, the NN can be used for UE-based positioning. Similar to the optional calibration phase 315 discussed above, the UE sends a request for NN inference input, for example, some reference signals from which the UE can derive channel response parameters that are input into the trained NN. The UE receives the NN inference input from the LMF and estimates the UE position [X, Y], which will be explained in more detail below.

[0052] The RS used for NN inference input can be periodic, aperiodic, or semi-persistent. After one or more positioning estimates, the UE can report the preferred TRP / cell ID of the reference signal and the type of reference signal, such as CSI-RS for CSI, CSI-RS for beam management (BM), or PRS. The LMF instance used for training can be the same LMF instance as the LMF instance used for inference or a different LMF instance.

[0053] According to the second option of training / inference, both training and inference are performed at the LMF. Using the second option can reduce UE processing relative to the first option. The LMF as a trusted entity can preserve UE privacy, although not as secure as the first option.

[0054] Figure 4 A signaling diagram 400 for AI based UE positioning according to the second option is shown, wherein the training and inference phases of the AI ​​process are performed at the LMF, according to various exemplary embodiments. Diagram 400 shows one example of the second option, however, alternative embodiments will also be described.

[0055] Block 405 shows a capability signaling exchange between the UE and the gNB / LMF to indicate the UE and network capabilities. The capability exchange of 405 may be similar to the capability exchange of 305 of FIG. 300 .

[0056] Box 410 shows the training phase of the AI-based positioning neural network. The LMF may initiate the training phase after capability exchange. In this option, no signaling is required between the UE and the LMF to initiate the training phase because the training is performed at the LMF. The LMF may use a first training data set (e.g., data set index x=0) for initial training, and may repeat the training phase with a different training data set (e.g., data set index x=1) if the following calibration phase results in an error exceeding a threshold.

[0057] Box 415 illustrates an optional calibration phase for the AI-based positioning neural network. For the calibration phase, the LMF may use another positioning mechanism to compare with the AI-based position determination to evaluate the accuracy of the AI ​​model within a certain error threshold. For example, in some embodiments, the LMF may identify the calibration position [X, Y] based on either (a) LMF-assisted RAT-independent technology (e.g., GNSS or GPS) defined in 3GPP Rel-16 or Rel-17 or (b) RAT-dependent technology (e.g., time difference of arrival (TDOA)). In other embodiments, the calibration position [X, Y] may be manual, for example, a specific position defined in the network as a positioning reference and reported by the UE to the LMF.

[0058] In this example, to perform AI-based position determination, the LMF requests the neural network to infer inputs (e.g., channel response, L1-RSRP, beam direction) and estimates the position and error based on the NN inferred inputs received from the UE. The LMF may request the NN inferred inputs (e.g., channel response) and calibrated position [X, Y] from a specific reference UE or positioning reference cell. If the error is greater than a predetermined threshold, the LMF returns to the training phase of 410 and trains the NN using a new training data set (e.g., data set index x=1).

[0059] In one example, the channel response may be obtained at the UE (from DL RS sent by one or more positioning gNB / TRPs) and the UE may send channel response feedback to the LMF as NN inference input. In another example, the channel response may be obtained at the gNB / TRP (from UL RS sent from the UE) and the gNB / TRP may send channel response feedback to the LMF. Figure 6 to Figure 7 These examples are described in more detail.

[0060] In an alternative embodiment, the training and calibration position determination of 410 to 415 may be done offline, for example, when the UE evaluates the positioning reference points as described above. The UE used to train the model may be different from the UE used for calibration and inference.

[0061] Box 420 shows the inference phase of the AI-based positioning neural network. After the NN is trained and optionally calibrated, the NN can be used for LMF-based positioning of the UE. The UE sends a request for AI-based positioning and receives a NN inference input request from the LMF. The UE can send the NN inference input (e.g., channel response feedback) to the LMF and the LMF estimates the UE position [X, Y]. The RS used for the NN inference input can be periodic, aperiodic, or semi-persistent.

[0062] Signal 425 shows the optional step of the LMF transmitting the determined position to the UE.

[0063] According to the third option for training / inference, training is performed at the LMF and inference is performed at the UE. Relative to the first and second options, the third option provides a compromise between the processing steps performed at the UE and the LMF.

[0064] Figure 5 A signaling diagram 500 for AI based UE positioning according to the third option is shown, wherein the training phase of the AI ​​process is performed at the LMF and the inference phase of the AI ​​process is performed at the UE according to various exemplary embodiments. Diagram 400 shows an example of the third option, however, alternative embodiments will also be described.

[0065] Block 505 shows a capability signaling exchange between the UE and the gNB / LMF to indicate the UE and network capabilities. The capability exchange of 505 may be similar to 305 of FIG. 300 and / or 405 of FIG. 400 .

[0066] Block 510 shows the training phase of the AI-based positioning neural network. The LMF may initiate the training phase after the capability exchange. Figure 4 410, in this option, no signaling is required between the UE and the LMF to initiate the training phase, because the training is performed at the LMF. The LMF can use the first training data set (e.g., data set index x=0) for initial training, and if the following calibration phase results in an error exceeding a threshold, the training phase can be repeated with a different training data set (e.g., data set index x=1).

[0067] Block 515 shows an optional calibration phase for the AI-based positioning neural network. Figure 4 415, for the calibration phase, the LMF may use another positioning mechanism to compare with the AI-based position determination to evaluate the accuracy of the AI ​​model within a certain error threshold. For example, in some embodiments, the LMF may identify the calibration position [X, Y] based on either of the LMF-assisted RAT-independent technology (e.g., GNSS or GPS) defined in 3GPP Rel-16 or Rel-17 or (b) RAT-dependent technology (e.g., time difference of arrival (TDOA)). In other embodiments, the calibration position [X, Y] may be manual, for example, a specific location defined as a positioning reference in the network. The calibration position may be provided as feedback from the UE.

[0068] To perform AI-based position determination, the LMF requests the neural network to infer inputs (e.g., channel response, L1-RSRP, beam direction) and estimates the position and error based on the NN inference input received from the UE. The LMF may request the NN inference input (e.g., channel response) and the associated position [X, Y] from a specific reference UE or positioning reference cell. If the error is greater than a predetermined threshold, the LMF returns to the training phase of 510 and trains the NN using a new training data set (e.g., data set index x=1).

[0069] Block 520 shows the UE receiving the trained and optionally calibrated NN. After the NN is trained and optionally calibrated, the NN may be provided from the LMF to the UE for UE-based positioning at the UE. Figure 5 In the example shown, the NN download is initiated by the UE transmitting a request for NN download to the LMF. However, in other embodiments, the NN download may be initiated by the LMF. In one embodiment, the NN may be downloaded offline from a positioning reference unit (e.g., a kiosk) via, for example, a high-speed wired link, a Bluetooth connection, a near field communication (NFC) link, a WiGig link (e.g., 802.11ad / ay), an ultra-wideband (UWB) link, etc.

[0070] Box 525 shows the inference phase of the AI-based positioning neural network. In this example, the UE sends a request for AI-based positioning and also sends a request for NN inference input. The UE receives NN inference input from the LMF, such as some reference signals, from which the UE can derive parameters for input into the trained neural network. The UE can estimate the UE position [X, Y] based on the received RS. The RS used for NN inference input can be periodic, aperiodic, or semi-persistent.

[0071] According to the fourth option, training is performed at the UE and inference is performed at the LMF. Currently, the fourth option is not considered particularly feasible. However, those skilled in the art will determine that the fourth option can be used if necessary.

[0072] As described above, depending on the various options for training / inference for AI-based positioning, the NN inference input may include a channel response for a DL reference signal (obtained at the UE) or a channel response for a UL reference signal (obtained at one or more gNBs).

[0073] When a DL reference signal is used to obtain a channel response, the RS may be, for example, a CSI-RS or a PRS. The UE obtains channel responses from multiple gNBs by receiving RS from each of the multiple gNBs. The LMF controls the RS configuration process.

[0074] Figure 6A signaling diagram 600 for channel response acquisition at a UE for identifying NN inference input for UE positioning according to various exemplary embodiments is shown. Diagram 600 shows LMF 131 and three gNBs (gNB 1, 2, 3) configured as positioning cells for UE 110. However, one skilled in the art will determine that a greater or lesser number of gNBs may be used for UE positioning depending on the type of NN, the type of channel response used as input to the NN, etc.

[0075] Box 605 shows a request (optional) and configuration of a DL RS for channel response acquisition. In some embodiments, as shown in 605, the UE may initiate a channel acquisition process by requesting an RS from the LMF. In this example, the UE transmits a request to the LMF using a positioning protocol, and the LMF configures RSs at different gNBs (e.g., gNB 1, gNB 2, and gNB 3). Any positioning protocol may be used, such as the Long Term Evolution (LTE) Positioning Protocol (LPP), the New Radio Positioning Protocol A (NRPPa), etc.

[0076] In an alternative embodiment, the UE transmits a request to an associated gNB (e.g., a serving gNB). The gNB relays the request to the LMF, and the LMF configures the RS at a different gNB. In other embodiments, no UE request for RS is required, and the LMF can initiate RS configuration.

[0077] RS can be configured by LMF according to any of the following three options. In the first option, LMF transmits the configuration of RS to each positioning TRP / gNB separately and transmits the configuration of RS directly to UE using positioning protocol. In the second option, LMF transmits RS configuration to each positioning TRP / gNB, and each gNB transmits the corresponding configuration to UE. For the second option, UE needs to be connected to the gNB that configures RS. In the third option, LMF transmits RS configuration to each positioning TRP / gNB and transmits all configurations to the serving gNB associated with the UE. The associated gNB transmits the configuration of all RS from all positioning TRP / gNBs to the UE.

[0078] Block 610 shows that the UE receives RS from the gNB. The gNB transmits a reference signal based on the configuration from the LMF. The UE estimates the channel response based on the received RS.

[0079] If the UE is performing the inference phase of the NN, the UE may use the channel response estimate as input to the training network. If the LMF is performing the inference phase of the NN, the UE may send channel response feedback to the LMF.

[0080] In some embodiments, for example, when the UE performs the training and / or inference phase of AI-based positioning, the UE may use the channel response estimate directly as an input to the NN inference. When the UE estimates the channel response during calibration of the NN, the UE may estimate the UE position and further determine an error value to evaluate the quality of the trained NN. When the UE is in the inference phase, the UE may estimate the UE position in [X, Y] coordinates.

[0081] In other embodiments, for example, when the LMF performs the inference phase of AI-based positioning, the UE may provide the channel response estimate as feedback to the LMF. The LMF is capable of triggering feedback of the UE channel response. In one option, the feedback may be provided directly to the LMF using the positioning protocol. In another option, the feedback may be provided to an associated gNB (e.g., a serving gNB), which relays feedback for all gNBs to the LMF. In this option, the feedback may be provided using the existing L1 CSI feedback process. In yet another option, feedback for each gNB is provided to the corresponding gNB that sends RS using the existing L1 CSI feedback process, and each of these gNBs relays the corresponding feedback to the LMF. In yet another option, feedback is provided to each gNB using a quantized time domain channel impulse response.

[0082] When the UL reference signal is used to obtain a channel response, the RS may be, for example, a sounding reference signal (SRS) or a positioning SRS (P-SRS). In some embodiments, the SRS may be a non-precoded SRS. One or more TRPs / gNBs obtain a channel response from the RS sent from the UE. The LMF controls the RS configuration process.

[0083] Figure 7 A signaling diagram 700 for channel response acquisition at one or more positioning TRPs / gNBs for identifying NN inference input for UE positioning at the LMF according to various exemplary embodiments is shown. Diagram 700 shows the LMF 131 and three gNBs (gNB 1, 2, 3) configured as positioning cells for the UE 110. However, one skilled in the art will determine that a greater or lesser number of gNBs may be used for UE positioning depending on the type of NN, the type of channel response used as input to the NN, etc.

[0084] Block 705 illustrates the request (optional) and configuration of RS for channel acquisition in the AI-based positioning method. In some embodiments, the UE may initiate a channel response acquisition procedure by requesting RS from the LMF, as shown in 705. In this example, the UE transmits the request directly to the LMF using the positioning protocol, and the LMF configures the reception of RS at different gNBs (e.g., gNB 1, gNB 2, and gNB3).

[0085] In an alternative implementation, the UE transmits a request to an associated gNB (e.g., serving gNB). The gNB relays the request to the LMF, and the LMF configures RS reception at a different gNB.

[0086] In other embodiments, a UE request for RS is not required and the LMF can initiate RS configuration. The LMF can configure RS by transmitting the RS configuration to each positioning TRP / gNB separately and transmitting the RS configuration directly to the UE using the positioning protocol, or by transmitting the RS configuration to each positioning TRP / gNB and transmitting all configurations to the serving gNB associated with the UE. The associated gNB transmits the uplink RS configuration to the UE.

[0087] Block 710 illustrates the UE transmission of RS to gNB based on the configuration. Each gNB estimates the channel response based on the received RS. In one embodiment, for example, if the gNB beams are aligned or in a single sector scenario, the UL RS may be sent to each gNB simultaneously. In another embodiment, the UL RS may be sent to each gNB at different times, for example, to accommodate beam directivities of different gNBs.

[0088] Box 715 shows the channel response feedback from the TRP / gNB to the LMF. In these embodiments, the LMF is performing the inference phase of the NN and can use the channel response estimate as input to the training network.

[0089] In the following, the above-mentioned training / inference option is also associated with the channel response acquisition option in the above-mentioned methods 800 to 1000.

[0090] Figure 8 A method 800 for AI-based UE positioning using a neural network (NN) is shown, according to various exemplary embodiments, wherein the training and inference phases of the AI ​​process are performed at the UE.

[0091] At 805, the UE receives a configuration for AI-based UE positioning from the LMF (e.g., via a serving gNB). The configuration may be received after the UE positioning capabilities are exchanged with the LMF.

[0092] At 810, the UE identifies training data for training the NN. In one option, the training data is identified based on historical data stored at the UE. For example, one or more training data sets may be generated based on results of a previous UE positioning method (e.g., using a legacy method such as TDOA). In another option, the training data is received from the LMF via a download from a serving gNB. The training data may be received in response to a UE request. In yet another option, the training data is received from a positioning reference unit (e.g., a kiosk) that is configured to provide training data via (e.g.) a high-speed wired link, a Bluetooth connection, a near field communication (NFC) link, a WiGig link (e.g., 802.11ad / ay), an ultra-wideband (UWB) link.

[0093] At 815, the UE trains the neural network with the identified training data.

[0094] At 820, the UE requests NN inference inputs. The inference inputs for the NN include channel response estimates determined based on DL RSs received at the UE from one or more gNBs. The channel response may include, for example, L1-RSRP, beam direction, etc. Thus, in some embodiments, the request may include one or more different types of reference signals (e.g., CSI-RS for CSI, CSI-RS for beam management, or PRS) for deriving the desired channel response. In some embodiments, the request may include a preferred TRP (e.g., cell ID) for the reference signal.

[0095] The UE may transmit the request directly to the LMF using the positioning protocol, or the UE may transmit the request to an associated base station (e.g., a serving gNB) which relays the request to the LMF.

[0096] When a request is received at the LMF, the LMF may determine the configuration of the RS to be transmitted to the UE. Based in part on the parameters included in the request, the LMF may select multiple TRPs (e.g., gNBs) for RS transmission and coordinate RS transmission from the various gNBs to the UE. The LMF may configure each of the gNBs for RS transmission separately. In some embodiments, the LMF may initiate RS configuration without a request.

[0097] At 825, the UE receives configuration for RS reception. In one embodiment, the LMF directly configures the UE using the positioning protocol via the serving gNB. In another embodiment, each of the gNBs is configured by the LMF, and each gNB configures its corresponding RS for the UE (this option requires the UE to be connected to each of the gNBs). In yet another embodiment, the serving gNB receives configuration for all gNBs used in the positioning method, and the serving gNB configures the UE for RS from all gNBs.

[0098] At 830, the UE estimates the channel response of the RS transmitted from each of the gNBs (e.g., gNB 1, 2, and 3). The channel response estimates may be used as input to the inference phase of the NN.

[0099] At 835, the UE performs the inference phase of the positioning NN and estimates the UE position. If the position estimate is for calibration purposes, the UE also estimates the error of the position estimate provided by the AI ​​model relative to the calibration position determined by other methods. In some embodiments, to determine the calibration position, the UE may use a positioning technology independent of RAT (e.g., GNSS) or a positioning technology dependent on RAT (e.g., TDOA). In other embodiments, the UE may receive the calibration position from a positioning reference point (e.g., an information kiosk).

[0100] Note that in these embodiments, the channel response is estimated based on the DL RS only at the UE, i.e., no complementary process is described for the gNB / LMF to estimate the channel response of the UL RS and provide feedback to the UE. However, this process can also be used if desired.

[0101] Fig. 9 A method 900 for AI-based UE positioning using a neural network (NN) is shown according to various exemplary embodiments, wherein the training and inference phases of the AI ​​process are performed at the LMF.

[0102] At 905, the UE receives a configuration for AI-based UE positioning from the LMF (e.g., via a serving gNB). The configuration may be received after the UE positioning capabilities are exchanged with the LMF.

[0103] At 910, the LMF identifies training data for training the NN.

[0104] At 915, the LMF trains the neural network using the identified training data.

[0105] At 920, the UE requests an AI-based position estimate. The inferred input for the NN includes a channel response estimate determined based on a DL RS received at the UE from one or more gNBs or a UL RS sent from the UE to one or more gNBs. The channel response may include, for example, L1-RSRP, beam direction, etc. Thus, in some embodiments, the request may include one or more different types of reference signals (e.g., CSI-RS for CSI, CSI-RS for beam management, PRS, SRS) for deriving the desired channel response. In some embodiments, the request may include a preferred TRP (e.g., cell ID) for the reference signal. The UE may transmit the request directly to the LMF using a positioning protocol, or the UE may transmit the request to an associated base station (e.g., a serving gNB) that relays the request to the LMF.

[0106] When receiving the request at the LMF, the LMF may determine the configuration of the RS to be transmitted or received to or from the UE. Based in part on the parameters included in the request, the LMF may select multiple TRPs (e.g., gNBs) for RS transmission / reception and coordinate RS transmission / reception to / from the corresponding gNBs of the UE. The LMF may configure each of the gNBs separately for RS transmission / reception.

[0107] At 925, the UE receives configuration for RS reception (DL RS from the positioning TRP / gNB) or RS transmission (UL RS to the positioning TRP / gNB). In one embodiment, the LMF directly configures the UE using the positioning protocol via the serving gNB. In another embodiment, each of the gNBs is configured by the LMF, and each gNB configures its corresponding RS for the UE (this option requires the UE to be connected to each of the gNBs). In yet another embodiment, the serving gNB receives configuration of all gNBs used in the positioning method, and the serving gNB configures the UE for RS to / from all gNBs.

[0108] At 930, the UE or positioning gNB estimates the channel response of the RS transmitted to / from each of the gNBs (e.g., gNB 1, 2, and 3). The channel response estimate may include, for example, RSRP, beam direction, etc.

[0109] At 935, the channel response estimate is provided from the UE / gNB to the LMF.

[0110] At 940, the LMF performs the inference phase of the positioning NN and estimates the UE position. If the position estimate is for calibration purposes, the LMF also estimates the error of the position estimate provided by the AI ​​model relative to the calibration position determined by other methods. In some embodiments, to determine the calibration position, the LMF may use a positioning technology independent of the RAT (e.g., GNSS) or a positioning technology dependent on the RAT (e.g., TDOA). In other embodiments, the LMF may receive the calibration position from the UE or via a positioning reference point (e.g., an information kiosk).

[0111] Fig.10 A method 1000 for AI-based UE positioning using a neural network (NN) according to various exemplary embodiments is shown, wherein a training phase of the AI ​​process is performed at the LMF and an inference phase of the AI ​​process is performed at the UE.

[0112] At 1005, the UE receives a configuration for AI-based UE positioning from the LMF (e.g., via a serving gNB). The configuration may be received after the UE positioning capabilities are exchanged with the LMF.

[0113] At 1010, the LMF identifies training data for training the NN.

[0114] At 1015, the LMF trains the neural network using the identified training data.

[0115] At 1020, the UE receives the trained NN from the LMF. The UE may optionally request the trained NN, or the NN provision may be LMF initiated. In one embodiment, the UE may download the trained NN offline from the positioning reference unit.

[0116] At 1025, the UE requests the NN inferred input. As described above at 820, the inferred input to the NN includes channel response estimates determined from DL RS received at the UE from one or more gNBs. The UE may transmit the request directly to the LMF using a positioning protocol, or the UE may transmit the request to an associated base station (e.g., a serving gNB) that relays the request to the LMF. Based on the request, the LMF may determine the configuration of RS to be sent to the UE from one or more positioning TRPs / gNBs.

[0117] At 1030 , the UE receives configuration for RS reception according to the implementation discussed above at 825 .

[0118] At 1035, the UE estimates the channel response of the RS sent from each of the gNBs (e.g., gNB 1, 2, and 3).

[0119] At 1040, the UE performs the inference phase of the positioning NN and estimates the UE position.

[0120] In this example, the UE performs only the inference phase and not the training phase. Therefore, if the trained NN is to be calibrated, the calibration is performed at the LMF. In this scenario, the LMF receives channel response feedback to perform the calibration inference phase, which can be feedback for the DL RS configured according to method 1000 or for the UL RS configured according to method 900. If the position estimate is for calibration purposes, the LMF also estimates the error of the position estimate provided by the AI ​​model relative to the calibration position determined by other methods.

[0121] Example

[0122] In a first embodiment, a method performed by a user equipment (UE) includes: receiving a configuration for an artificial intelligence (AI)-based UE positioning method using a neural network (NN) from a location management function (LMF); identifying a first training data set for training the NN; training the neural network with the first training data set; requesting NN inference input; receiving a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells; estimating a channel response for a DL RS received from the one or more positioning cells using the channel response as the NN inference input; and estimating the UE position.

[0123] In a second embodiment, according to the method of the first embodiment, the method further comprises sending a request for the first training data set to the LMF, and receiving the first training data set from the LMF.

[0124] In a third embodiment, the method according to the second embodiment, wherein the request for the first training data includes an identification of a type of DL RS used to estimate the channel response.

[0125] In a fourth embodiment, according to the method of the first embodiment, the method further includes receiving the first training data set from the LMF without a specific request from the UE.

[0126] In a fifth embodiment, according to the method of the first embodiment, the method further includes determining the first training data set from historical data stored at the UE.

[0127] In a sixth embodiment, according to the method of the first embodiment, the method also includes receiving the first training data set from a physical positioning reference point via a wired link, a Bluetooth connection, a near field communication (NFC) link, a WiGig link, or an ultra-wideband (UWB) link.

[0128] In a seventh embodiment, the method according to the first embodiment, wherein the NN infers that the input is one of periodic, aperiodic, or semi-persistent.

[0129] In an eighth embodiment, according to the method described in the first embodiment, the method also includes: identifying a calibration position to calibrate the trained NN; estimating the UE position based on the NN inference input used for calibration; determining whether the error between the UE position based on the NN inference input used for calibration and the calibration position is greater than a threshold; and when the error is greater than the threshold, retraining the neural network with second training data.

[0130] In a ninth embodiment, the method according to the eighth embodiment, wherein the calibration position is identified based on a positioning technology independent of a radio access technology (RAT) or a positioning technology dependent on a RAT.

[0131] In a tenth embodiment, the method of the ninth embodiment, wherein the calibration position is identified from a physical positioning reference point.

[0132] In an eleventh embodiment, the method according to the first embodiment, wherein the first training data is received from a first LMF instance, and the configuration for DL ​​RS reception is received from a second LMF instance.

[0133] In a twelfth embodiment, according to the method of the first embodiment, the request for the NN inference input includes a preferred cell for the DL RS reception and a preferred type for the DL RS.

[0134] In a thirteenth embodiment, the method according to the first embodiment, wherein the request for the NN to infer input is sent to the LMF using a positioning protocol.

[0135] In a fourteenth embodiment, the method according to the first embodiment, wherein the request for the NN inference input is sent to a serving cell.

[0136] In a fifteenth embodiment, according to the method of the first embodiment, the configuration for the DL RS reception is received from the LMF using a positioning protocol.

[0137] In a sixteenth embodiment, according to the method of the first embodiment, the configuration received for the DL RS is received from each positioning cell of the one or more positioning cells.

[0138] In a seventeenth embodiment, according to the method of the first embodiment, the configuration for the DL RS reception is received from the serving cell for each of the one or more positioning cells.

[0139] In an eighteenth embodiment, a processor of a user equipment (UE) is configured to perform any of the operations described in the first to seventeenth embodiments.

[0140] In a nineteenth embodiment, a user equipment (UE) includes: a transceiver configured to communicate with a network; and a processor communicatively coupled to the transceiver and configured to perform any of the operations described in accordance with the first to seventeenth embodiments.

[0141] In a twentieth embodiment, a method performed at a location management function (LMF) includes: sending a configuration for an artificial intelligence (AI)-based UE positioning method using a neural network (NN) to a user equipment (UE), wherein the UE identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a request for NN inference input from the UE; and sending a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells to the UE, wherein the UE estimates a channel response for the DL RS received from the one or more positioning cells; using the channel response as the NN inference input and estimating the UE position.

[0142] In a twenty-first embodiment, according to the method of the twentieth embodiment, the method further includes receiving a request for the first training data set from the UE and sending the first training data set from the LMF.

[0143] In a twenty-second embodiment, the method according to the twenty-first embodiment, wherein the request for the first training data includes identification of a type of DL RS used to estimate the channel response.

[0144] In a twenty-third embodiment, according to the method of the twentieth embodiment, the method further includes sending the first training data set to the UE without a specific request from the UE.

[0145] In a twenty-fourth embodiment, the method according to the twentieth embodiment, wherein the request for the NN inference input includes a preferred cell for the DL RS reception and a preferred type for the DL RS.

[0146] In a twenty-fifth embodiment, the method according to the twentieth embodiment, wherein the request for the NN to infer input is received from the UE using a positioning protocol.

[0147] In a twenty-sixth embodiment, according to the method of the twentieth embodiment, the configuration received by the DL RS is sent to the UE using a positioning protocol.

[0148] In a twenty-seventh embodiment, the method according to the twentieth embodiment, wherein the LMF is included in a 5G New Radio (NR) core network (CN) or included in one of the devices providing a side link (SL) connection to the UE.

[0149] In a twenty-eighth embodiment, a location management function (LMF) is configured to perform any of the operations described according to the twentieth to twenty-seventh embodiments.

[0150] In a twenty-ninth embodiment, a method performed by a user equipment (UE) includes: receiving a configuration for an artificial intelligence (AI)-based UE positioning method using a neural network (NN) from a location management function (LMF), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells; estimating a channel response for the DL RS received from the one or more positioning cells; sending the channel response to the LMF as an NN inference input, wherein the LMF estimates the UE position; and receiving the UE position estimate from the LMF.

[0151] In a thirtieth embodiment, according to the method described in the twenty-ninth embodiment, the method also includes: identifying a calibration position to calibrate the trained NN; and sending the calibration to the LMF, wherein the LMF estimates the UE position based on the NN inferred input used for calibration; determining whether the error between the UE position based on the NN inferred input used for calibration and the calibration position is greater than a threshold; and when the error is greater than the threshold, retraining the neural network with second training data.

[0152] In a thirty-first embodiment, the method according to the thirtieth embodiment, wherein the calibration position is identified based on a positioning technology independent of a radio access technology (RAT) or a positioning technology dependent on a RAT.

[0153] In a thirty-second embodiment, the method according to the thirtieth embodiment, wherein the calibration position is identified from a physical positioning reference point.

[0154] In a thirty-third embodiment, according to the method of the twenty-ninth embodiment, the configuration for the DL RS reception is received from the LMF using a positioning protocol.

[0155] In a thirty-fourth embodiment, according to the method of the twenty-ninth embodiment, the configuration received for the DL RS is received from each of the one or more positioning cells.

[0156] In a thirty-fifth embodiment, according to the method of the twenty-ninth embodiment, the configuration for the DL RS reception is received from a serving cell for each of the one or more positioning cells.

[0157] In a thirty-sixth embodiment, a user equipment (UE) includes: a transceiver configured to communicate with a network; and a processor communicatively coupled to the transceiver and configured to perform any of the operations described in accordance with the twenty-ninth to thirty-fifth embodiments.

[0158] In a thirty-seventh embodiment, a method performed by a location management function (LMF) includes: training a neural network (NN) using a first training data set for an artificial intelligence (AI)-based user equipment (UE) positioning method; receiving a request for AI-based positioning from the UE; transmitting to the UE a NN inference input request including a configuration for receiving a downlink (DL) reference signal (RS) on one or more positioning cells; receiving from the UE a channel response for the DL RS received from the one or more positioning cells, wherein the channel response is the NN inference input; and estimating the UE position using the NN based on the NN inference input.

[0159] In a thirty-eighth embodiment, according to the method described in the thirty-seventh embodiment, the method further includes transmitting a UE position estimate to the UE.

[0160] In a thirty-ninth embodiment, according to the method described in the thirty-seventh embodiment, the method also includes: identifying a calibration position to calibrate the trained NN; and sending the calibration to the LMF, wherein the LMF estimates the UE position based on the NN inferred input used for calibration; determining whether the error between the UE position based on the NN inferred input used for calibration and the calibration position is greater than a threshold; and when the error is greater than the threshold, retraining the neural network with the second training data.

[0161] In a fortieth embodiment, the method according to the thirty-ninth embodiment, wherein the calibration position is identified based on a positioning technology independent of a radio access technology (RAT) or a positioning technology dependent on a RAT.

[0162] In a forty-first embodiment, a method according to the thirty-ninth embodiment, wherein the calibration position is identified from a physical positioning reference point.

[0163] In a 42nd embodiment, according to the method of the 37th embodiment, the configuration for the DL RS reception is received from the LMF using a positioning protocol.

[0164] In a forty-third embodiment, according to the method of the thirty-seventh embodiment, the configuration received for the DL RS is received from each of the one or more positioning cells.

[0165] In a forty-fourth embodiment, according to the method of the thirty-seventh embodiment, the configuration for the DL RS reception is received from a serving cell for each of the one or more positioning cells.

[0166] In a forty-fifth embodiment, a location management function (LMF) is configured to perform any of the operations of the thirty-seventh to forty-fourth embodiments.

[0167] In a forty-sixth embodiment, a method performed by a user equipment (UE) includes: receiving a configuration for an artificial intelligence (AI)-based UE positioning method using a neural network (NN) from a location management function (LMF), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for sending an uplink (UL) reference signal (RS) to one or more positioning cells; sending a UL RS to the one or more positioning cells; and receiving a UE position estimate from the LMF.

[0168] In a forty-seventh embodiment, according to the method described in the forty-sixth embodiment, the method also includes: identifying a calibration position to calibrate the trained NN; and sending the calibration to the LMF, wherein the LMF estimates the UE position based on the NN inferred input used for calibration; determining whether the error between the UE position based on the NN inferred input used for calibration and the calibration position is greater than a threshold; and when the error is greater than the threshold, retraining the neural network with second training data.

[0169] In a forty-eighth embodiment, the method according to the forty-seventh embodiment, wherein the calibration position is identified based on a positioning technology independent of a radio access technology (RAT) or a positioning technology dependent on a RAT.

[0170] In a forty-ninth embodiment, a method according to the forty-seventh embodiment, wherein the calibration position is identified from a physical positioning reference point.

[0171] In a fiftieth embodiment, according to the method of the forty-sixth embodiment, the configuration for the UL RS reception is received from each of the one or more positioning cells.

[0172] In a fifty-first embodiment, according to the method of the forty-sixth embodiment, the configuration for the UL RS reception is received from a serving cell for each of the one or more positioning cells.

[0173] In a fifty-second embodiment, a user equipment (UE) includes: a transceiver configured to communicate with a network; and a processor communicatively coupled to the transceiver and configured to perform any of the operations described in the forty-sixth to fifty-first embodiments.

[0174] Those skilled in the art will appreciate that the exemplary embodiments described above may be implemented with any suitable software configuration or hardware configuration or combination thereof. Exemplary hardware platforms for implementing the exemplary embodiments may include, for example, Intel x86-based platforms with compatible operating systems, Windows OS, Mac platforms and MAC OS, mobile devices with operating systems such as iOS, Android, etc. The exemplary embodiments of the above methods may be embodied as a program comprising lines of code stored on a non-transitory computer-readable storage medium, which, when compiled, may be executed on a processor or microprocessor.

[0175] Although this patent application describes various combinations of various embodiments each having different features, those skilled in the art will understand that any feature of an embodiment may be combined with features of other embodiments in any manner not publicly denied or with features that are not functionally or logically inconsistent with the operation or function of the device of the embodiments disclosed in this disclosure.

[0176] It is understood that the use of personally identifiable information should be subject to privacy policies and practices that are generally recognized to meet or exceed industry or government requirements for maintaining user privacy. Specifically, personally identifiable information data should be managed and processed to minimize the risk of unintentional or unauthorized access or use, and the nature of the authorized use should be clearly stated to users.

[0177] It will be apparent to those skilled in the art that various modifications may be made to the present disclosure without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure is intended to cover modifications and variations of the present disclosure, provided that these modifications and variations are within the scope of the appended claims and their equivalents.

Claims

1. A processor of a user equipment (UE), the processor being configured to perform operations comprising: receiving, from a location management function (LMF), a configuration for an artificial intelligence (AI) based UE positioning method using a neural network (NN), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells; estimating a channel response for a DL RS received from the one or more positioning cells; sending the channel response as NN inference input to the LMF, wherein the LMF estimates the UE position; and A UE position estimate is received from the LMF.

2. The processor of claim 1 , wherein the operations further comprise: identifying a calibration location to calibrate the trained NN; as well as The calibration is sent to the LMF, wherein the LMF estimates the UE position based on the NN inference input used for calibration, determines whether the error between the UE position based on the NN inference input used for calibration and the calibration position is greater than a threshold, and when the error is greater than the threshold, retrains the neural network with second training data. 3 . The processor of claim 2 , wherein the calibration position is identified based on a radio access technology (RAT) independent positioning technology or a RAT dependent positioning technology.

4. The processor of claim 2, wherein the calibration position is identified from a physical positioning reference point.

5. The processor of claim 1, wherein the configuration for the DL RS reception is received from the LMF using a positioning protocol. 6 . The processor of claim 1 , wherein the configuration received for the DL RS is received from each of the one or more positioning cells. 7 . The processor of claim 1 , wherein the configuration received for the DL RS is received from a serving cell for each of the one or more positioning cells.

8. A location management function (LMF) of a cellular core network, the location management function (LMF) being configured to perform operations including: training a neural network (NN) using a first training data set for an artificial intelligence (AI) based user equipment (UE) positioning method; receiving a request for AI-based positioning from a UE; transmitting a NN inference input request to the UE, the NN inference input request comprising a configuration for downlink (DL) reference signal (RS) reception on one or more positioning cells; receiving, from the UE, a channel response for a DL RS received from the one or more positioning cells, wherein the channel response is a NN inference input; as well as The NN is used to estimate a UE position based on the NN inference input.

9. The LMF of claim 8, wherein the operations further comprise: A UE position estimate is transmitted to the UE.

10. The LMF of claim 8, wherein the operations further comprise: identifying a calibration location to calibrate the trained NN; as well as The calibration is sent to the LMF, wherein the LMF estimates the UE position based on the NN inference input used for calibration, determines whether the error between the UE position based on the NN inference input used for calibration and the calibration position is greater than a threshold, and when the error is greater than the threshold, retrains the neural network with second training data.

11. The LMF of claim 10, wherein the calibration position is identified based on a radio access technology (RAT) independent positioning technology or a RAT dependent positioning technology.

12. The LMF of claim 10, wherein the calibration position is identified from a physical positioning reference point.

13. The LMF of claim 8, wherein the configuration for the DL RS reception is received from the LMF using a positioning protocol.

14. The LMF of claim 8, wherein the configuration for the DL RS reception is received from each of the one or more positioning cells.

15. The LMF of claim 8, wherein the configuration for the DL RS reception is received from a serving cell for each of the one or more positioning cells.

16. A processor of a user equipment (UE), the processor configured to perform operations comprising: receiving, from a location management function (LMF), a configuration for an artificial intelligence (AI) based UE positioning method using a neural network (NN), wherein the LMF identifies a first training data set for training the NN and trains the neural network with the first training data set; receiving a configuration for uplink (UL) reference signal (RS) transmission to one or more positioning cells; Sending a UL RS to the one or more positioning cells; as well as A UE position estimate is received from the LMF.

17. The processor of claim 16, wherein the operations further comprise: identifying a calibration location to calibrate the trained NN; as well as The calibration is sent to the LMF, wherein the LMF estimates the UE position based on the NN inference input used for calibration, determines whether the error between the UE position based on the NN inference input used for calibration and the calibration position is greater than a threshold, and when the error is greater than the threshold, retrains the neural network with second training data.

18. The processor of claim 17, wherein the calibration location is identified based on a radio access technology (RAT) independent positioning technology or a RAT dependent positioning technology.

19. The processor of claim 17, wherein the calibration position is identified from a physical positioning reference point.

20. The processor of claim 16, wherein the configuration for UL RS reception is received from each of the one or more positioning cells.

21. The processor of claim 16, wherein the configuration for the UL RS reception is received from a serving cell for each of the one or more positioning cells.

Images