Methods and systems for artificial intelligence assisted positioning and sensing
By implementing data compression and feature extraction on UE and network sides, the system addresses the challenge of large measurement data in AI/ML positioning and sensing, achieving accurate and efficient location and velocity estimation.
Patent Information
- Application Number
- PCT/CN2024/070955
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-01-05
- Publication Date
- 2025-07-10
AI Technical Summary
Existing positioning and sensing methods face challenges in achieving accurate estimation results, particularly with AI/ML algorithms, where large measurement data reports are difficult for user equipment to manage, and existing protocols are not applicable.
A system and method involving data compression and feature extraction on user equipment (UE) to reduce the size of measurement reports, using models on both UE and network sides for efficient data handling and accurate location or velocity determination.
The proposed solution enables accurate and efficient reporting of UE measurements, allowing for precise location and velocity estimation with reduced data size, suitable for both AI/ML positioning and sensing scenarios.
Smart Images

Figure CN2024070955_10072025_PF_FP_ABST
Abstract
Description
METHODS AND SYSTEMS FOR ARTIFICIAL INTELLIGENCE ASSISTED POSITIONING AND SENSINGTECHNICAL FIELD
[0001] The disclosure relates generally to wireless communications, including but not limited to systems and methods for artificial intelligence assisted positioning and / or sensing.BACKGROUND
[0002] The standardization organization Third Generation Partnership Project (3GPP) is currently in the process of specifying a new Radio Interface called 5G New Radio (5G NR) as well as a Next Generation Packet Core Network (NG-CN or NGC) . The 5G NR will have three main components: a 5G Access Network (5G-AN) , a 5G Core Network (5GC) , and a User Equipment (UE) . In order to facilitate the enablement of different data services and requirements, the elements of the 5GC, also called Network Functions, have been simplified with some of them being software based, and some being hardware based, so that they could be adapted according to need.SUMMARY
[0003] The example embodiments disclosed herein are directed to solving the issues relating to one or more of the problems presented in the prior art, as well as providing additional features that will become readily apparent by reference to the following detailed description when taken in conjunction with the accompany drawings. In accordance with various embodiments, example systems, methods, devices and computer program products are disclosed herein. It is understood, however, that these embodiments are presented by way of example and are not limiting, and it will be apparent to those of ordinary skill in the art who read the present disclosure that various modifications to the disclosed embodiments can be made while remaining within the scope of this disclosure.
[0004] Artificial Intelligence (AI) and / or Machine Learning (ML) algorithms can have promising application prospects in positioning and sensing. Using AI and / or ML (AI / ML) algorithms in positioning and sensing can obtain more accurate estimation results compared with the traditional methods. For example, in positioning scenarios, the AI / ML positioning algorithms can obtain an estimation result with an error of less than 1 meter, which can be difficult to achieve in traditional positioning methods. However, different from the traditional positioning or sensing methods, the measurement data for AI / ML algorithms can be large and difficult for user equipment (UE) to report. The contents of the UE measurement report in the existing specifications or protocols can be not applicable to AI / ML algorithms. Solutions for UE measurement reports in AI / ML positioning and sensing are proposed in this disclosure.
[0005] At least one aspect is directed to a system, method, apparatus, or a computer-readable medium of the following, which may involve a first wireless communication node (e.g., base station) , a second wireless communication node (e.g., a core network, or any network function or part thereof) interacting with a wireless communication device (e.g., user equipment) . A wireless communication node can refer to a BS, a transmit-receive point (TRP) , or any function / portion of a core network. The method can include receiving, by the wireless communication device from a first wireless communication node, a plurality of signals (e.g., positioning reference signals) each associated with (e.g., received at) a respective one of a plurality of time units (e.g., time slots) . The method can include obtaining, by the wireless communication device, a plurality of measurements of the plurality of signals, each corresponding to a respective one of the plurality of time units. The method can include performing, by the wireless communication device, data compression and / or feature extraction on the plurality of measurements, to obtain a data-compressed and / or feature-extracted output across the time units. The method can include sending, by the wireless communication device to the second wireless communication node, the data-compressed and / or feature-extracted output to determine a location or velocity (e.g., Doppler shift / value) .
[0006] For a sensing scenario of BS transmission and UE reception, a two-side model solution can be used for reducing the data size of the UE measurement report. The method of using two-side model to reduce the data size of the UE measurement report in sensing scenario of BS transmission and UE reception can include a Doppler feature extraction model (e.g., neural network) which can be deployed on the UE side (e.g. encoder or any other neural networks with the capability of dimension reduction) , or an estimation model which can be deployed on the network side and can infer the target position and velocity with the output of Doppler feature extraction model. The input of the Doppler feature extraction model can be a number of channel impulse response (CIR) , power delay profile (PDP) , and delay profile (DP) matrices which can be measured from a number of slots when the sensing reference signal is OFDM format or a number of time units when the sensing reference signal is of another format. The output of the Doppler feature extraction model can be a vector or small feature matrix that contains the Doppler information of the target.
[0007] In some embodiments, the plurality of measurements (e.g., in matrix form) comprises at least one of a channel impulse response (CIR) , a power delay profile (PDP) , or a delay profile (DP) . The wireless communication device can use (e.g., execute, run, activate) a first model to perform data compression or feature extraction. In some embodiments, data compression is the same as feature extraction; in other embodiments, data compression can be different from feature extraction (e.g., where a feature-extracted output can be compressed as a further step) . A size of the data-compressed or feature-extracted output can be smaller than a combined size of the plurality of measurements. The second wireless communication node may use a second model to determine the location or velocity. The data-compressed and / or feature-extracted output can include a compressed feature matrix. Each column or row of the compressed feature matrix can correspond to a respective path. Each element in a column or row of the compressed feature matrix can correspond to a velocity of a respective target in a respective path.
[0008] In some embodiments, the wireless communication device can combine the compressed feature matrix with a matrix corresponding to one of the plurality of measurements. A series of matrices of at least one of CIR, PDP and / or DP can be used as training data to train a model to perform the data compression or feature extraction. The training data can include a label (e.g., ground truth information) comprising a vector or matrix having Doppler information estimated using a difference operation or a defined Doppler estimation method.
[0009] In some implementations, the method can include sending, by the wireless communication device to the second wireless communication node, an indication of at least one of a size of measurement data, request or requirement for accuracy of measurement. The method can include combining or concatenating (e.g., splicing) matrices, or measurement description information. The method can include receiving, by the wireless communication device from the second wireless communication node responsive to the indication, at least one of an identifier of a model for the data compression or feature extraction, or delivery of the model for data compression or feature extraction. The method can include sending, by the wireless communication device to the second wireless communication node, a capability report comprising the identifier of the model for data compression and / or feature extraction. The method can include sending, by the wireless communication device to the second wireless communication node, a capability report and / or functionality report comprising an indication of the wireless communication device’s capability of generating CIR, PDP and / or DP for the measurements.
[0010] In certain implementations, the second wireless communication node can select, according to the indication, at least one part of a model to be used to determine the location or velocity. The second wireless communication node can train at least one (or another) part of the model, according to the indication (e.g., according to the requirement for accuracy of measurement) . The wireless communication device can obtain a first of the plurality of measurements of the plurality of signals. The wireless communication device can obtain a first plurality of measurements corresponding to a first one of the plurality of time units, and corresponding to NResource number of resources. The wireless communication device can combine the first plurality of measurements corresponding to the NResource number of resources into a first measurement (e.g., a merged / combined measurement) of the plurality of measurements.
[0011] In some embodiments, the wireless communication device can report to the second wireless communication node the first measurement. A size of the first measurement can be NResource times of a matrix size of each of the plurality of measurements. The method can include sending, by the wireless communication device to the second wireless communication node, a capability report comprising an indication of at least one of a number of resources that the wireless communication device can measure or the matrix size of each of the plurality of measurements. In some embodiments, the matrix size can be expressed as {Nslot*NResource*NTRP*Nport*Nt, Breal} for sensing and / or {NResource*NTRP*Nport*Nt, Breal} for positioning.
[0012] Combining the first plurality of measurements into the first measurement can include providing the first measurement to comprise a reference matrix for one of the resources, and a list of difference information between the reference matrix and each matrix of others of the resources.
[0013] In certain embodiments, the method can include obtaining, by the wireless communication device, a first plurality of measurements corresponding to a first time unit of the plurality of time units and corresponding to a first resource, and a second plurality of measurements corresponding to the first time unit and corresponding to a second resource. The method can include sending, by the wireless communication device, the first plurality of measurements to the second wireless communication node, for a first model to determine a first estimation. The method can include sending, by the wireless communication device, the second plurality of measurements to the second wireless communication node, for a second model to determine a second estimation. The location or velocity can be determined using the first estimation and the second estimation.
[0014] In some embodiments, the method can include training data for training the first model collected for the first resource. The training data can include a label determined for the first resource. The location or velocity can be determined using a third model that receives as input the first estimation and the second estimation. The location or velocity can be determined by applying respective weights on first estimation and the second estimation. The method can include receiving, by the wireless communication device from the second wireless communication node, assistance data comprising an indication of at least one of the resources, or models corresponding to the resources. The second wireless communication node (e.g., network function or NW / network) can send a “nr-SelectedAI-Model-IndexList” to the wireless communication device, and the information element (IE) used to send / carry this to the wireless communication device can be referred to as assistance data or comprising assistance data. The method can include sending, by the wireless communication device to the second wireless communication node, an indicator of line-of-sight (LOS) or non line-of-sight (NLOS) , for use to select a model to determine the location or velocity.
[0015] In some embodiments, the indicator can be included in a matrix of the plurality of measurements. The indicator can include one or more vector elements each representing a probability of LOS propagation path between a first wireless communication node (e.g., a TRP) and the wireless communication device. The one or more vector elements can form / represent additional columns of the matrix. A model can be used to detect a measurement with the indicator. The training data for training the model can include a number of matrices that incorporate the indicator. The training data can include training labels each comprising vector elements that each represent an indication of an LOS or NLOS scenario. A respective model can be trained for each LOS or NLOS scenario. The model can determine, for the measurement with the indicator, an index of the respective model to select. The location or velocity can be determined using the respective model that is selected.
[0016] In some embodiments, the method can include performing, by the wireless communication device, Fourier transform on matrices of the measurements, to obtain Doppler information for each propagation path. The method can include determining, by the wireless communication device from the Doppler information, one or more Doppler feature matrices. The method can include sending, by the wireless communication device to the second wireless communication node, a reference matrix and the one or more Doppler feature matrices. Each column or row of the reference matrix can correspond to a respective path. Each of the one or more Doppler feature matrices can correspond to a column or row of the reference matrix. Each element in one of the one or more Doppler feature matrices can correspond to a velocity (e.g., Doppler shift / value) of a respective target (e.g., target object) in a respective path (e.g., path of a reference signal) .
[0017] In certain embodiments, the method can include applying, by the wireless communication device, a filter to eliminate delay elements in the reference matrix that fail to meet a defined threshold. The defined threshold can include a defined number of delays. The defined threshold can be applied on each column or each element of the reference matrix. A size of the output can be Nep times of a matrix size of each of the Doppler feature matrix. Nep can be a number of effective propagation paths in the reference matrix. The method can include applying, by the wireless communication device, a filter to eliminate Doppler elements in the one or more Doppler feature matrices that fail to meet a defined threshold. The defined threshold can include a defined number of frequency shifts or velocity. The defined threshold can be applied on each column or each element of the one or more Doppler feature matrices. A size of the feature-extracted output can be NDoppler times of a matrix size of each of the plurality of measurements, where NDoppler can be the number of effective Doppler values in a Doppler feature matrix.
[0018] In certain implementations, the method can include sending, by a first wireless communication node to a wireless communication device, a plurality of signals each associated with a respective one of a plurality of time units, for the wireless communication device to obtain a plurality of measurements of the plurality of signals, each corresponding to a respective one of the plurality of time units, and to perform data compression or feature extraction on the plurality of measurements, to obtain a data-compressed or feature-extracted output across the time units. The method can include receiving, by the second wireless communication node from the wireless communication device, the feature-extracted output to determine a location or velocity.
[0019] In some embodiments, each column in the Doppler feature matrix can correspond to each path (or Nt index) in the CIR, PDP, and DP matrix and each element in the Doppler feature matrix can correspond to a target Doppler (velocity) which can be obtained from the Doppler feature extraction model. The input of the estimation model deployed on the network side can be the output of the Doppler feature extraction model. The output of the estimation model deployed on the network side can be the position and / or velocity of the targets. The Doppler feature extraction model and the estimation model can have life cycle management stages including data collection, model training, model inference, and model monitoring.
[0020] In certain embodiments, the training data of the Doppler feature extraction model can be a number of the CIR, PDP, and / or DP matrices which are measured from a number of slots when the sensing reference signal is OFDM format or a number of time units when the sensing reference signal is another format. The training label of the Doppler feature extraction model can be a vector or small feature matrix containing Doppler information which can be obtained according to difference operation or FT (Fourier Transform) .
[0021] In certain implementations, the UE can report the corresponding Doppler feature extraction model ID that the UE can support to the network. The UE can provide its measurement data size, Doppler estimation accuracy request, matrix concatenation method, and other measurement description information to the network. The network can assign a Doppler extraction model with a model ID to the UE. In the model inference stage, after obtaining the Doppler feature matrix, the UE can splice the Doppler feature matrix with one original CIR, PDP, and / or DP matrix obtained from one slot or time unit. The UE can report its capability of generating CIR, PDP, and / or DP within condition information in a functionality report, or the UE can report its capability of generating the CIR, PDP, and / or DP in the UE capability report. In artificial intelligence (AI) or machine learning (ML) sensing and in artificial intelligence or machine learning positioning, the UE can report its capability of generating the CIR, PDP, and / or DP in the dimensions of Nslot, NTRP, Nport, and Nt. For both AI / ML positioning and sensing, the network can select model parts such as convolution layers related to the input data size based on the UE capability of generating the CIR, PDP, and DP. The network can train other parts of the model such as full connection layers, based on the position and velocity estimation accuracy request reported by the UE.
[0022] In some implementations, the UE can send its positioning and sensing accuracy request according to an AI-accuracy request which is included in a provided location information IE. The AI-accuracy request can include a horizontal accuracy request, vertical accuracy request and velocity accuracy request. For both AI positioning and sensing, the UE can regard the information of reference signal resource as an additional dimension of the input matrix. The UE can measure the CIR, PDP, and / or DP matrix per resource and can merge the measurement results from different resources into one complete matrix. The UE can report the CIR, PDP, and DP matrix with a size of Nslot*NResource*NTRP*NPort*Nt as the model input in the sensing scenario. The UE can report the CIR, PDP, and / or DP matrix with a size of NResource*NTRP*NPort*Nt in the positioning scenario. The CIR, PDP, and / or DP matrix can be merged and may have the same input size (e.g., the same NTRP, NPort and Nt) .
[0023] In certain embodiments, the UE can perform difference or dimension reduction operations on the CIR, PDP, and / or DP matrices from different resources. The UE can report a reference CIR, PDP, and / or DP matrix with an additional list. The reference matrix can be a CIR, PDP, and / or DP matrix which can be obtained from a specific reference signal resource. The additional list can be a matrix that includes the different information between other resources and the reference resource and can have the same NTRP and Nport dimension with the aboriginal CIR, PDP, and / or DP matrix. The UE can report a nr-DL-RS-ReferenceResourceID, nr-AI-CIR / PDP / DP and nr-AdditionalList in the NR-DL-AI-MeasElement IE. For both AI / ML positioning and sensing, the UE can measure and generate the CIR, PDP, and / or DP matrix per resource and the model deployed on the network can be trained per resource. The final positioning or sensing estimation result can be the fusion / combination of results from several models and each model corresponding to a specific resource. The training data (e.g., CIR, PDP, and / or DP) can be generated on the UE side and can be collected per resource. The training labels can be collected per resource.
[0024] In certain implementations, the network can input the output of models which are to be fused into the fused model. The output of the fused model can be the final positioning or sensing estimation result. The network can assign a weight to each model, which can be generated based on the accuracy requirements reported by the UE. The final estimation result can be obtained by weighting the outputs of each model. The network can provide the matching relationship between models and resources in the nr-SelectedAI-Model-IndexList which is included in the NR-AI-ProvideAssistanceData IE. For AI / ML positioning and sensing, the network can select a proper model according to the scenario information provided by the LOS / NLOS indicator. The UE can receive the PRS or sensing reference signal, and generate a CIR, PDP, and / or DP matrix and a vector of LOS / NLOS indicator per resource respectively. Each element in the LOS / NLOS vector can be a soft value and can represent the probability of LOS propagation path from a TRP to the UE.
[0025] In some implementations, the UE can append the LOS / NLOS vectors as an additional column of the Nt dimension. A LOS / NLOS distinguishing model can be deployed on the network side to distinguish the scenarios in which the UE is located based on the CIR, PDP, and / or DP matrix after the appending LOS / NLOS indicator reported by the UE. The training data for the LOS / NLOS distinguishing model can be a number of CIR, PDP, and / or DP matrices with an additional column of LOS / NLOS indicator. The training labels for the LOS / NLOS distinguishing model can be a number of vectors in which each element can represent an indicator of the environment. In the model training stage, the network can train different models using CIR, PDP, and / or DP matrices measured from different environments. The final positioning or sensing estimation result can be the output of the model after the environment selection according to the LOS / NLOS indicator. The UE can report the value of LOS / NLOS indicator according to a traditional measurement report. For a sensing scenario of BS transmission and UE reception, the UE can report the CIR, PDP, and / or DP matrix and Doppler measurement data without using models. The UE can perform Fourier Transform (FT) operations on the CIR, PDP, and / or DP matrices measured from multiple slots or time units and obtain Doppler information from each propagation path (e.g., in each Nt dimension index of the CIR, PDP, and / or DP matrix, there can be a corresponding Doppler feature matrix) . The size of Doppler feature matrix can be NTRP*Nport*NDoppler which can be related to the slots or time units used for measuring Doppler information. When CIR, PDP, and / or DP matrices are relatively sparse, the number of Doppler feature matrices obtained from multiple slots or time units can be less. The whole data size that the UE can report may be the sum of reference CIR, PDP, and / or DP matrix and the Doppler feature matrices (e.g., NTRP*Nport*Nt0+Nep*NTRP*Nport*NDoppler, wherein Nep represents the number of effective propagation paths in CIR, PDP, and / or DP matrices) . The UE can perform a threshold selection in the NDoppler dimension.
[0026] In certain embodiments, the Doppler selection threshold can be a certain number of frequency shifts or velocities. The Doppler threshold selection can be performed on each column or one element in the NDoppler dimension. After applying the threshold filter, the Doppler feature matrix can become sparse, and the whole data size that the UE can report may be represented as NDoppleri*NTRP*Nport*Nt0 wherein NDoppleri can be the number of Doppler values over the threshold in NDoppler dimension.
[0027] A non-transitory computer readable medium storing instructions, which when executed by at least one processor, can cause the at least one processor to perform the method of any one of the methods. An apparatus can include at least one processor configured to implement the method of any one of the methods.
[0028] In some implementations, a non-transitory computer-readable medium may store instructions that when executed by at least one processor may cause the at least one processor to perform any one or more of the methods disclosed herein. An apparatus may comprise at least one processor configured to perform any one or more of the methods disclosed herein.BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Various example embodiments of the present solution are described in detail below with reference to the following figures or drawings. The drawings are provided for purposes of illustration only and merely depict example embodiments of the present solution to facilitate the reader's understanding of the present solution. Therefore, the drawings should not be considered limiting of the breadth, scope, or applicability of the present solution. It should be noted that for clarity and ease of illustration, these drawings are not necessarily drawn to scale.
[0030] FIG. 1 illustrates an example cellular communication network in which techniques disclosed herein may be implemented, in accordance with an embodiment of the present disclosure;
[0031] FIG. 2 illustrates a block diagram of an example base station and a user equipment device, in accordance with some embodiments of the present disclosure;
[0032] FIG. 3 illustrates an example implementation of a sensing system based on BS transmission and UE reception, in accordance with some embodiments of the present disclosure;
[0033] FIG. 4 illustrates an example implementation with models on UE side and network (NW) side, in accordance with some embodiments of the present disclosure;
[0034] FIG. 5 illustrates the size relationship between the input and output of a Doppler feature extraction matrix, in accordance with some embodiments of the present disclosure;
[0035] FIG. 6 illustrates the size relationship between the input and output of a Doppler feature extraction matrix, in accordance with some embodiments of the present disclosure;
[0036] FIG. 7 illustrates a UE reporting processing capability in sensing scenarios, in accordance with some embodiments of the present disclosure;
[0037] FIG. 8 illustrates a UE reporting processing capability in positioning scenarios, in accordance with some embodiments of the present disclosure;
[0038] FIG. 9 illustrates the network selecting and / or training parts of a model according to the UE capability report and accuracy request, in accordance with some embodiments of the present disclosure;
[0039] FIG. 10 illustrates an example implementation of a UE merging a CIR, PDP, and / or DP matrices and reporting to the network (e.g., with a size of NResource*NTRP*NPort*Nt) , in accordance with some embodiments of the present disclosure;
[0040] FIG. 11 illustrates an example implementation of a UE reporting a CIR, PDP, and / or DP matrices from different resources with dimension reduction, in accordance with some embodiments of the present disclosure;
[0041] FIG. 12 illustrates an example implementation of models deployed on the network side, in accordance with some embodiments of the present disclosure;
[0042] FIG. 13 illustrates an example implementation of using a model to combine positioning or sensing estimation results from different resources, in accordance with some embodiments of the present disclosure;
[0043] FIG. 14 illustrates an example implementation of using weighting values to combine positioning or sensing estimation results from different resources, in accordance with some embodiments of the present disclosure;
[0044] FIG. 15 illustrates an example implementation of a UE incorporating an LOS / NLOS indicator as additional information to a CIR, PDP, and / or DP matrix, in accordance with some embodiments of the present disclosure;
[0045] FIG. 16 illustrates an example of how the UE can reduce the data size of the measurement report (e.g., without using an AI / ML model) , in accordance with some embodiments of the present disclosure;
[0046] FIG. 17 illustrates an example approach of using Doppler threshold selection, in accordance with some embodiments of the present disclosure;
[0047] FIG. 18 illustrates an example of a relationship between the CIR, PDP, and / or DP matrix and the Doppler feature matrix, in accordance with some embodiments of the present disclosure;
[0048] FIG. 19 illustrates a flow diagram of an example method for performing data compression or feature extraction on a plurality of measurements at the UE side, in accordance with an embodiment of the present disclosure.DETAILED DESCRIPTION
[0049] 1. Mobile Communication Technology and Environment
[0050] FIG. 1 illustrates an example wireless communication network, and / or system, 100 in which techniques disclosed herein may be implemented, in accordance with an embodiment of the present disclosure. In the following discussion, the wireless communication network 100 may be any wireless network, such as a cellular network or a narrowband Internet of things (NB-IoT) network, and is herein referred to as “network 100. ” Such an example network 100 includes a base station 102 (hereinafter “BS 102” ; also referred to as wireless communication node) and a user equipment device 104 (hereinafter “UE 104” ; also referred to as wireless communication device) that can communicate with each other via a communication link 110 (e.g., a wireless communication channel) , and a cluster of cells 126, 130, 132, 134, 136, 138 and 140 overlaying a geographical area 101. In FIG. 1, the BS 102 and UE 104 are contained within a respective geographic boundary of cell 126. Each of the other cells 130, 132, 134, 136, 138 and 140 may include at least one base station operating at its allocated bandwidth to provide adequate radio coverage to its intended users.
[0051] For example, the BS 102 may operate at an allocated channel transmission bandwidth to provide adequate coverage to the UE 104. The BS 102 and the UE 104 may communicate via a downlink radio frame 118, and an uplink radio frame 124 respectively. Each radio frame 118 / 124 may be further divided into sub-frames 120 / 127 which may include data symbols 122 / 128. In the present disclosure, the BS 102 and UE 104 are described herein as non-limiting examples of “communication nodes, ” generally, which can practice the methods disclosed herein. Such communication nodes may be capable of wireless and / or wired communications, in accordance with various embodiments of the present solution.
[0052] FIG. 2 illustrates a block diagram of an example wireless communication system 200 for transmitting and receiving wireless communication signals (e.g., OFDM / OFDMA signals) in accordance with some embodiments of the present solution. The system 200 may include components and elements configured to support known or conventional operating features that need not be described in detail herein. In one illustrative embodiment, system 200 can be used to communicate (e.g., transmit and receive) data symbols in a wireless communication environment such as the wireless communication environment 100 of FIG. 1, as described above.
[0053] System 200 generally includes a base station 202 (hereinafter “BS 202” ) and a user equipment device 204 (hereinafter “UE 204” ) . The BS 202 includes a BS (base station) transceiver module 210, a BS antenna 212, a BS processor module 214, a BS memory module 216, and a network communication module 218, each module being coupled and interconnected with one another as necessary via a data communication bus 220. The UE 204 includes a UE (user equipment) transceiver module 230, a UE antenna 232, a UE memory module 234, and a UE processor module 236, each module being coupled and interconnected with one another as necessary via a data communication bus 240. The BS 202 communicates with the UE 204 via a communication channel 250, which can be any wireless channel or other medium suitable for transmission of data as described herein.
[0054] As would be understood by persons of ordinary skill in the art, system 200 may further include any number of modules other than the modules shown in FIG. 2. Those skilled in the art will understand that the various illustrative blocks, modules, circuits, and processing logic described in connection with the embodiments disclosed herein may be implemented in hardware, computer-readable software, firmware, or any practical combination thereof. To clearly illustrate this interchangeability and compatibility of hardware, firmware, and software, various illustrative components, blocks, modules, circuits, and steps are described generally in terms of their functionality. Whether such functionality is implemented as hardware, firmware, or software can depend upon the particular application and design constraints imposed on the overall system. Those familiar with the concepts described herein may implement such functionality in a suitable manner for each particular application, but such implementation decisions should not be interpreted as limiting the scope of the present disclosure.
[0055] In accordance with some embodiments, the UE transceiver 230 may be referred to herein as an "uplink" transceiver 230 that includes a radio frequency (RF) transmitter and a RF receiver each comprising circuitry that is coupled to the antenna 232. A duplex switch (not shown) may alternatively couple the uplink transmitter or receiver to the uplink antenna in time duplex fashion. Similarly, in accordance with some embodiments, the BS transceiver 210 may be referred to herein as a "downlink" transceiver 210 that includes a RF transmitter and a RF receiver each comprising circuity that is coupled to the antenna 212. A downlink duplex switch may alternatively couple the downlink transmitter or receiver to the downlink antenna 212 in time duplex fashion. The operations of the two transceiver modules 210 and 230 may be coordinated in time such that the uplink receiver circuitry is coupled to the uplink antenna 232 for reception of transmissions over the wireless transmission link 250 at the same time that the downlink transmitter is coupled to the downlink antenna 212. Conversely, the operations of the two transceivers 210 and 230 may be coordinated in time such that the downlink receiver is coupled to the downlink antenna 212 for reception of transmissions over the wireless transmission link 250 at the same time that the uplink transmitter is coupled to the uplink antenna 232. In some embodiments, there is close time synchronization with a minimal guard time between changes in duplex direction.
[0056] The UE transceiver 230 and the base station transceiver 210 are configured to communicate via the wireless data communication link 250, and cooperate with a suitably configured RF antenna arrangement 212 / 232 that can support a particular wireless communication protocol and modulation scheme. In some illustrative embodiments, the UE transceiver 210 and the base station transceiver 210 are configured to support industry standards such as the Long Term Evolution (LTE) and emerging 5G standards, and the like. It is understood, however, that the present disclosure is not necessarily limited in application to a particular standard and associated protocols. Rather, the UE transceiver 230 and the base station transceiver 210 may be configured to support alternate, or additional, wireless data communication protocols, including future standards or variations thereof.
[0057] In accordance with various embodiments, the BS 202 may be an evolved node B (eNB) , a serving eNB, a target eNB, a femto station, or a pico station, for example. In some embodiments, the UE 204 may be embodied in various types of user devices such as a mobile phone, a smart phone, a personal digital assistant (PDA) , tablet, laptop computer, wearable computing device, etc. The processor modules 214 and 236 may be implemented, or realized, with a general purpose processor, a content addressable memory, a digital signal processor, an application specific integrated circuit, a field programmable gate array, any suitable programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof, designed to perform the functions described herein. In this manner, a processor may be realized as a microprocessor, a controller, a microcontroller, a state machine, or the like. A processor may also be implemented as a combination of computing devices, e.g., a combination of a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a digital signal processor core, or any other such configuration.
[0058] Furthermore, the steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in firmware, in a software module executed by processor modules 214 and 236, respectively, or in any practical combination thereof. The memory modules 216 and 234 may be realized as RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. In this regard, memory modules 216 and 234 may be coupled to the processor modules 210 and 230, respectively, such that the processors modules 210 and 230 can read information from, and write information to, memory modules 216 and 234, respectively. The memory modules 216 and 234 may also be integrated into their respective processor modules 210 and 230. In some embodiments, the memory modules 216 and 234 may each include a cache memory for storing temporary variables or other intermediate information during execution of instructions to be executed by processor modules 210 and 230, respectively. Memory modules 216 and 234 may also each include non-volatile memory for storing instructions to be executed by the processor modules 210 and 230, respectively.
[0059] The network communication module 218 generally represents the hardware, software, firmware, processing logic, and / or other components of the base station 202 that enable bi-directional communication between base station transceiver 210 and other network components and communication nodes configured to communication with the base station 202. For example, network communication module 218 may be configured to support internet or WiMAX traffic. In a typical deployment, without limitation, network communication module 218 provides an 802.3 Ethernet interface such that base station transceiver 210 can communicate with a conventional Ethernet based computer network. In this manner, the network communication module 218 may include a physical interface for connection to the computer network (e.g., Mobile Switching Center (MSC) ) . The terms “configured for, ” “configured to” and conjugations thereof, as used herein with respect to a specified operation or function, refer to a device, component, circuit, structure, machine, signal, etc., that is physically constructed, programmed, formatted and / or arranged to perform the specified operation or function.
[0060] The Open Systems Interconnection (OSI) Model (referred to herein as, “open system interconnection model” ) is a conceptual and logical layout that defines network communication used by systems (e.g., wireless communication device, wireless communication node) open to interconnection and communication with other systems. The model is broken into seven subcomponents, or layers, each of which represents a conceptual collection of services provided to the layers above and below it. The OSI Model also defines a logical network and effectively describes computer packet transfer by using different layer protocols. The OSI Model may also be referred to as the seven-layer OSI Model or the seven-layer model. In some embodiments, a first layer may be a physical layer. In some embodiments, a second layer may be a Medium Access Control (MAC) layer. In some embodiments, a third layer may be a Radio Link Control (RLC) layer. In some embodiments, a fourth layer may be a Packet Data Convergence Protocol (PDCP) layer. In some embodiments, a fifth layer may be a Radio Resource Control (RRC) layer. In some embodiments, a sixth layer may be a Non Access Stratum (NAS) layer or an Internet Protocol (IP) layer, and the seventh layer being the other layer.
[0061] Various example embodiments of the present solution are described below with reference to the accompanying figures to enable a person of ordinary skill in the art to make and use the present solution. As would be apparent to those of ordinary skill in the art, after reading the present disclosure, various changes or modifications to the examples described herein can be made without departing from the scope of the present solution. Thus, the present solution is not limited to the example embodiments and applications described and illustrated herein. Additionally, the specific order or hierarchy of steps in the methods disclosed herein are merely example approaches. Based upon design preferences, the specific order or hierarchy of steps of the disclosed methods or processes can be re-arranged while remaining within the scope of the present solution. Thus, those of ordinary skill in the art will understand that the methods and techniques disclosed herein present various steps or acts in a sample order, and the present solution is not limited to the specific order or hierarchy presented unless expressly stated otherwise.
[0062] 2. Systems and Methods for Artificial Intelligence Assisted Positioning and / or Sensing
[0063] In some implementations of an AI / ML physical layer framework, there can be at least two sub-use cases for positioning, namely AI / ML direct positioning and AI / ML assisted positioning. In AI / ML direct positioning, the output of a model can be the final UE location. In AI / ML assisted positioning, the output of the model can be new measurements or enhanced measurements of certain positioning methods (e.g., based on time delay, angle, or carrier phase) . From the perspective of model deployment, the two sub-use cases can be described as following:
[0064] Case 1: UE-based positioning with UE-side model, direct AI / ML or AI / ML assisted positioning
[0065] Case 2a: UE-assisted / LMF-based positioning with UE-side model, AI / ML assisted positioning
[0066] Case 2b: UE-assisted / LMF-based positioning with LMF-side model, direct AI / ML positioning
[0067] Case 3a: NG-RAN node assisted positioning with gNB-side model, AI / ML assisted positioning
[0068] Case 3b: NG-RAN node assisted positioning with LMF-side model, direct AI / ML positioning
[0069] In each case mentioned above, the life cycle of the model can include several stages (e.g., data collection, model training, model transfer, model inference, model monitoring and / or model update) . At each stage, there can be different operations performed on the model.
[0070] In some embodiments, three types of AI / ML model inputs may be defined for example: channel impulse response (CIR) , power delay profile (PDP) , and delay profile (DP) . Each type of model input can have three dimensions for instance, namely NTRP, Nport, and Nt, which can be different from other types of measurements for positioning and sensing (e.g., that are based on Relative Signal Time Difference (RSTD) , Reference Signal Received Power (RSRP) , or Rx-Tx time difference) . The data size of CIR, PDP and / or DP can exceed thousands of bits and may have an impact on the accuracy of positioning and sensing results.
[0071] Example Implementation 1
[0072] FIG. 3 illustrates an example implementation of a sensing system based on base station (BS) 305 transmission and UE 315 reception. The sensing operations of estimating positioning and / or velocity can be deployed in the network side 320. In sensing scenarios, a target 310 velocity (e.g., Doppler shift / value) can be estimated while estimating the target 310 position. The target 310 can include an unmanned aerial vehicle (drone) or aircraft, as a non-limiting example.
[0073] The estimation accuracy of the Doppler value can be related to the number of pulse trains used for measuring. For example, in orthogonal frequency division multiplex (OFDM) signals, the more slots used to measure the Doppler, the more accurate estimation results can be obtained. However, for sensing measurements using BS transmission and UE reception, excessive measurement slots can impose a burden on the measurement reports of the UE.
[0074] Three types of model inputs, CIR, PDP and DP, can be used in an AI physical layer framework. Taking CIR as an example, the CIR can be a complex matrix with dimensions of NTRP *Nport *Nt. When using the CIR as the model input, the data size of measurement can reach more than 4096 bits per sample. For a sensing system with a model deployed in the network (NW) side, using BS transmission and UE reception (similar to sub-use Case 2b in AI / ML Positioning) , it can be important to reduce the size of measurement data reported by the UE.
[0075] One solution can be to deploy models on both on the UE side and the NW side. The model deployed on the UE side can include a Doppler feature extraction model. The Doppler feature extraction model can include an encoder or other neural network with dimension reduction abilities. The model deployed on the network side can include a trained neural network which can infer target position and / or velocity result (s) using the output of the UE side model / Doppler feature extraction model. The model deployed on the network side can include a sensing result estimation model.
[0076] FIG. 4 can illustrate an example implementation of models on UE and NW sides, and can include a Doppler feature extraction model on the UE side and a sensing result estimation model deployed on NW side. For example, as shown FIG. 4, the UE can receive a sensing signal scattered by targets, and can obtain a CIR, PDP, and / or DP matrix per slot (or other time unit) . After accumulating a certain number of CIR, PDP, and / or DP matrices, the UE can input these matrices into the Doppler feature extraction model, and the Doppler feature extraction model can compress the matrices into a relatively smaller Doppler feature matrix with the size of NTRP*Nport*Nt3 for instance, wherein t0>>t3 (e.g., indicating sparsity of non-zero / null measurement values compared to the total number of measurement elements in each matrix; where Nport dimension have been omitted in FIG. 4) . The number of CIR, PDP, and / or DP matrices for accumulation and the size of the Doppler feature matrix can be related to the Doppler estimation accuracy and a sparsity condition of the CIR matrix.
[0077] FIGs. 5 and 6 can illustrate the size relationship (s) between the input and output of the Doppler feature extraction matrix for sparse and non-sparse CIR matrices. The input of the Doppler feature extraction model can be of the same size (e.g., NTRP*Nport*Nt0) and the output size can be of NTRP*Nport*Nt3 wherein t3 can be related to (or indicative of) the sparsity of Nt dimension in CIR, PDP, and / or DP matrices. For example, when the CIR, PDP, and / or DP matrices are relatively sparse such as FIG. 5, the size of the Doppler feature matrix can be obtained through cumulative compression and can have a smaller size (e.g., as compared to the combined size of the CIR, PDP, and / or DP matrices) . Each column in the Doppler feature matrix can correspond to each (e.g., reference signal transmission / reflection / traversal) path (or Nt index) in the CIR, PDP, and / or DP matrix. Each element in the Doppler feature matrix can correspond to the target Doppler value (target velocity) . The target Doppler value / shift (target velocity) can be obtained from the Doppler feature extraction model.
[0078] After obtaining the Doppler feature matrix, the UE can splice (e.g., combine) the Doppler feature matrix with one original CIR, PDP, and / or DP matrix obtained from one slot measurement. The dimension of the spliced matrix can be NTRP*Nport* (Nt0+Nt3) . The spliced matrix can include the information of target position and / or Doppler value. The splicing method can be determined by the UE. The size of the matrix obtained after splicing can be related to the splicing method.
[0079] The UE can transfer the spliced matrix to the network. The model deployed on network side can infer / estimate / calculate the target position and / or target velocity estimation result (s) .
[0080] In some sensing methods, the UE is to report a number of CIR, PDP, and / or DP matrices in order to obtain an accurate Doppler estimation; the measurement data size that the UE is to report may be Nslot*NTRP*Nport*Nt0. In the solution disclosed herein, the measurement data size that the UE is to report can be of NTRP*Nport* (Nt0+Nt3) for instance, which can be smaller in size than the other methods. With the deployment of a feature extraction model on the UE side, the solution disclosed herein can be applied for a common UE and / or a reduced capability UE.
[0081] The specific phases for life cycle management (LCM) for the solution disclosed herein can include data collection, model training, model inference, and model / or monitoring.
[0082] The training data for Doppler feature extraction model (UE side model) can include a series of CIR, PDP, and / or DP matrices. The training labels for the Doppler feature extraction model can include a vector or small feature matrix containing Doppler information. The Doppler information can be obtained according to a difference operation or other Doppler estimation methods. The training data for the estimation model (network side model) can include the output of the UE side model. The training label for the estimation model can include ground truth value (s) of the target position and / or velocity which can be obtained according to a Global Position System (GPS) and / or Doppler Velocimeter (DVL) .
[0083] The estimation model (network side model) can be trained in the network. The Doppler feature extraction model (UE side model) can be trained / established / configured in the network or the UE. When the Doppler feature extraction model is trained / established / configured in the network, the Doppler feature extraction model can be transferred from the network to the UE.
[0084] The UE can report a corresponding Doppler feature extraction model ID to the network. The UE can provide to the network the measurement data size, Doppler estimation accuracy request, matrix concatenation method, and / or any other measurement descriptive information, for selecting / identifying / assigning a suitable Doppler feature extraction model. The network can identify and / or assign a Doppler extraction model with a globally unique model ID, to the UE, according to any one or more of the reported / provided information. After model identification and delivery, the UE can notify the network of its functionality / capability (which can include the Doppler feature extraction model ID) according to the UE Capability Report.
[0085] The UE can receive the sensing signal scattered by the targets, and can obtain / generate CIR, PDP, and / or DP matrices. When measurements are completed, the UE can input the CIR, PDP, and / or DP matrices into the Doppler feature extraction model, and can splice the output of the Doppler feature extraction model with at least one of the original CIR, PDP, and / or DP matrices. The UE can report a size of the spliced matrix to the network. The network can select a sensing result estimation model that is suitable for the size of UE’s spliced matrix. In the absence of ground truth labels in actual sensing scenarios, the UE can determine and / or report position and / or velocity estimation result (s) to the network, to enable the network to evaluate the model performance.
[0086] Example Implementation 2
[0087] In an AI physical layer framework, the input size of CIR, PDP, and / or DP measurement / matrix can be expressed for instance as NTRP*Nport*Nt. The NTRP can be related to the number of the BS and / or TRPs located / utilized in the scenario. The Nport can be the number of ports in each TRP. The Nt can be the number of time domain samples. The Doppler feature extraction model in Example 1 can include the above three dimensions, and one additional dimension can represent the number of slots to accumulate. The input size can have an impact on model performance (e.g., the positioning accuracy is related to the dimension of NTRP and Nt) . It can be essential to consider the potential impact of the input size for AI / ML sensing and / or positioning.
[0088] In an actual channel estimation process, the UE can process the received sensing reference signal (Sensing RS) or positioning reference signal (PRS) , and can generate CIR, PDP, and / or DP measurement / matrix of an appropriate size based on its computing power (e.g., memory, Central Processing Unit or Graphic Processing Unit) . For example, the data size of channel estimation result that a UE can support may be M bits. Therefore, the UE may be able to generate the CIR with the size of NTRP1*Nport1*Nt1@BCIR, PDP with the size of NTRP2*Nport2*Nt2@BPDP, or DP with the size of NTRP3*Nport3*Nt3@BDP, for example. Wherein BCIR, BPDP and BDP can represent the number of bits corresponding to each time domain samples in the CIR, PDP, and / or DP measurement / matrix, respectively. When the UE can report the capability of generating CIR, PDP, and / or DP, the network can select a proper model according to the UE Capability Report.
[0089] FIG. 7 illustrated a UE reporting a capability of the UE (e.g., processing capability) in sensing scenarios. FIG. 7 can illustrate a solution in which the UE can report its processing capability in a sensing scenario. For example, the UE can report its capability of generating CIR, PDP, and / or DP within condition information in a Functionality Report (illustrated using solid lines) , or the UE can report its capability of generating CIR, PDP, and / or DP in an UE Capability Report (illustrated using dashed line) . Wherein the specific information of UE transferred to the network (e.g., via an information element) can be as follows:
[0090] Simultaneously, on the network side of FIG. 7, there may be three or more models such as Model 1, Model 2, and Model 3. For Model 1, the input data format can be Nslot1*NTRP1*Nport1*Nt1@BCIR and can be for the CIR matrix. For Model 2, the input data format can be Nslot2*NTRP2*Nport2*Nt2@BPDP and can be for the PDP matrix. For Model 3, the input data format can be Nslot3*NTRP3*Nport3*Nt3@BDP and can be for the DP matrix. The actual output dimensions of Models 1, 2, and / or 3 can be the same or different. The UE can decide on a method of splicing with the original CIR, PDP, and / or DP.
[0091] FIG. 8 can illustrate an UE reporting processing capability in positioning scenarios. For UE-assisted / (Location Management Function (LMF) ) -based positioning with LMF-side model (Case 2b) , the UE can report its capability of generating CIR, PDP, and / or DP. FIG. 8 is an illustration in which the UE can report its capability of generating CIR, PDP, and / or DP within condition information in a Functionality Report (e.g., shown using solid line) or in a UE Capability Report (e.g., shown using dashed line) . In a positioning scenario, the model (s) can be deployed in the network side only. The output of Models 1, 2 and / or 3 can be the position of a target and may have the same dimension.
[0092] The specific information of UE that are transferred to the network (e.g., via an information element) can be described via the following example:
[0093] Example Implementation 3
[0094] In Example 2, it has been proposed that the data size of CIR, PDP, and / or DP can affect sensing and / or positioning accuracy, and thereby can affect the model selection. The impacts can be focused on the model inference stage of LCM. However, in the model training stage, different training samples can affect the performance of the model. For example, using 20,000 samples of PDP matrices to train the model can achieve almost the same effect as using 10,000 samples of CIR. More training samples can be used to moderate the performance degradation caused by smaller input size (e.g., using 40,000 samples of PDP matrices with 6*1*256@16 bits can achieve similar performance to using 25000 samples of PDP matrices with 9*1*256@16 bits) .
[0095] When the UE can report its capability of measurement generation and the accuracy request of the target position and / or velocity, the network can select the proper model and can use a suitable number of samples to train the selected model. However, for AI / ML sensing and / or positioning, the model training can be offline. Hence, when the network can use the measurement information and positioning accuracy requirements reported by UE to train the model, it can involve a training delay which can bring significant difficulties to implementation.
[0096] A solution can be that the network can select model parts related to the input data size (such as convolution layers) based on the UE capability to generate the CIR, PDP, and / or DP. The network can train other parts of the model, such as full connection layers, based on the position and / or velocity estimation accuracy requirements reported by the UE.
[0097] FIG. 9 can illustrate a network selecting and training a model according to the UE Capability Report and / or Accuracy Request. For example, when the UE can enter a specific sensing or positioning area, the UE can report its reference signal processing capability to the network (e.g., the NTRP, Nport, Nt and Bbit ) for the CIR, PDP, and / or DP. After receiving the UE processing capability, the network can for example select a suitable convolution layer from Conv1, Conv2 and Conv3. The parts of the model to be selected can be related to the specific structure of the model.
[0098] Simultaneously or in parallel, the UE can report its positioning or sensing accuracy request (e.g., the upper bound of positioning or velocity estimation error) . After receiving the positioning and sensing accuracy request, the network can train parts of the model that may have a few parameters and can be easier to train (e.g., full connection layer) . The number of samples used to train the model can be suitable for the accuracy request and can be decided by the network. The parts of the model to be selected can be related to the specific structure of the model.
[0099] After the selection and training is completed, the network can integrate various parts into a complete model. The accuracy requirement information can be included in Provide Location Information IE as following:
[0100] Example Implementation 4
[0101] For a PRS or sensing reference signal in the form of OFDM, it may be important to consider the impact of resources on model inputs. For example, during positioning, the resource of PRS can be related to a TRP transmission beam. The CIR, PDP, and / or DP measured from each resource can include different channel information and can be different. The measurement size of one data sample can be defined for example as: (measurement data size of one PRS / SRS resource) * (number of PRS / SRS resources for model input) . The measurement data can be related to the resource.
[0102] As disclosed herein, one solution can be that the UE can regard the information about a corresponding resource as an additional dimension of the input matrix. The UE can measure a CIR, PDP, and / or DP matrix per resource and can merge the measurement results from different resources into one complete matrix. The UE can report the CIR, PDP, and / or DP matrix with a size of Nslot*NResource*NTRP*NPort*Nt for example, as the model input in a sensing scenario. The UE can report the CIR, PDP, and / or DP matrix with a size of NResource*NTRP*NPort*Nt for example, in a positioning scenario.
[0103] FIG. 10 illustrates an example implementation of the UE merging the CIR, PDP, and / or DP matrices and reporting to the network with a size of for example NResource*NTRP*NPort*Nt. For example, the UE can receive the PRS or the sensing reference signal from different resources and can merge the generated CIR, PDP, and / or DP matrices for these resources into one matrix. The merged matrix can include the position and / or Doppler information of / across all resources. The CIR, PDP, and / or DP matrix can be merged and may have the same input size (the same NTRP, NPort and Nt) . When merging is completed, the UE can notify the network of the number of resources it can support according to the RS Processing Capability report, and can include the Support-AI / ML-SensingRS-Inputsize IE or Support-AI / ML-PRS-Inputsize IE.
[0104] Adding the NResource dimension may impose a burden on the UE measurement report due to the measurement data size of over thousands of bits. Disclosed herein, the UE can perform difference or any other dimension reduction operations on the CIR, PDP, and / or DP matrices from different resources.
[0105] FIG. 11 illustrates an example implementation of the UE reporting the CIR, PDP, and / or DP matrices from different resources with dimension reduction. The UE can compress the CIR, PDP, and / or DP matrices from different resources through difference and / or dimension reduction operations. The UE can report a reference CIR, PDP, and / or DP matrix with an additional list. The reference matrix can be a CIR, PDP, and / or DP matrix from a particular resource. The additional list can be a matrix that includes the difference information between any one resource and the reference resource. The additional list can have the same NTRP and / or Nport dimension (s) as that of as the original CIR, PDP, and / or DP matrix. After receiving the reference CIR, PDP, and / or DP matrix and the additional list, the network can restore / recover / infer the CIR, PDP, and / or DP matrix from / for any particular resource.
[0106] The UE can report an additional list to include the information from / about different resources. The size of UE measurement data can be preserved / unchanged. As disclosed herein, the UE can measure the CIR, PDP, and / or DP matrix for each resource. The model on the network can be trained for each resource. The positioning or sensing estimation can be a combined result of several models and each model can correspond to a specific resource.
[0107] FIG. 12 illustrates the UE measuring CIR / DP / PDP per resource. The models deployed on the network side can be trained per resource (e.g., each model trained for a respective resource) . Several models can be included / established on the network side. Each model can correspond to a respective PRS resource or sensing reference signal resource.
[0108] The specific stages for LCM for FIG. 12 can include data collection, model training, and / or model inference.
[0109] In the data collection stage, the training data (e.g., the CIR, PDP, and / or DP) can be generated on the UE side and can be collected per resource. The training labels can be collected per resource.
[0110] In the model training stage, the network can train different models using the CIR, PDP, and / or DP matrix measured from the different resources (e.g., training model 1 using the CIR, PDP, and / or DP matrix measured from resource 1; training model 2 using the CIR, PDP, and / or DP matrix measured from resource 2; training model 3 using the CIR, PDP, and / or DP matrix measured from resource 3; etc. ) .
[0111] In the model inference stage, UE can report its measurement result of the CIR, PDP, and / or DP matrix to the network. In each measurement report process, the data size of reporting can for example be NTRP*Nport*Nt. When the UE measurement report is received by the network, the network can input the CIR, PDP, and / or DP matrix to the corresponding model which can be trained with the training data obtained from the same PRS resource or sensing reference signal resource.
[0112] FIG. 13 illustrates an example implementation of using a model to combine positioning or sensing estimation results from different resources. The network can combine the output of each model into a final / combined / merged result. For example, the network can input the output of model 1, 2, and / or 3 into a next level model. The next level model can be model 0. Model 0 can include a trained model whose input can be the output of model 1, 2, and / or 3 and whose output can be the final positioning or sensing result.
[0113] FIG. 14 illustrates an example of using weighting value to combine positioning or sensing estimation results from different resources. The network can assign a weight to each model, which can be generated based on the accuracy requirements reported by the UE mentioned in Example 3. The estimation result can be obtained by weighting the outputs of each model.
[0114] During model monitoring, the network can compare the final results with the position or velocity estimation results reported by the UE. The network can compare the output results of each model with traditional / other method results.
[0115] The matching / correspondence / pairing relationship between models and resources can be included / indicated in the assistance data (e.g., NR-AI-ProvideAssistanceData) .
[0116] Example 5
[0117] The UE can report an indicator of a Line-of-Sight (LOS) or Non-Line-of-Sight (NLOS) within the measurement report such as a DL-TDOA method or DL-AOD method. By using the LOS / NLOS indicator, the UE or network can eliminate the propagation path (s) representing the NLOS path (s) to obtain more accurate time delay or angle measurement results. For AI / ML positioning or sensing, the UE may indicate / report whether the UE can support reporting the LOS / NLOS indicator while reporting the CIR, PDP, and / or DP matrix. For AI / ML assisted positioning, the UE can (e.g., additionally) report the measurements of other / traditional methods (e.g., for comparison) . It may be important to incorporate the use of such a LOS / NLOS indicator for AI / ML positioning or sensing.
[0118] For AI / ML positioning and sensing, a benefit of the UE reporting LOS / NLOS indicator can be that the network can select a suitable model according to the scenario information provided by the LOS / NLOS indicator. For example, when most of the reported indicators are NLOS, the network can choose a model which has been trained in a heavy NLOS environment.
[0119] The LOS / NLOS indicator can be reported per TRP, per resource, or both per TRP and per resource. Given that the model input (e.g., the CIR, PDP, and / or DP) can have the dimension of NResource and NTRP, the UE can add the LOS / NLOS indicator as additional information to the CIR, PDP, and / or DP matrix.
[0120] FIG. 15 illustrates an example of the UE adding the LOS / NLOS indicator as additional information to the CIR, PDP, and / or DP matrix. The UE can receive the PRS or sensing reference signal, and can generate a CIR, PDP, and / or DP matrix and a vector of LOS / NLOS indicator per resource. Each element in the LOS / NLOS vector can include a soft value and can represent a respective probability of LOS for a propagation path between each different TRP (s) and the UE.
[0121] After obtaining the CIR, PDP, and / or DP matrices and LOS / NLOS indicator vectors, the UE can append the LOS / NLOS vectors as an additional column of Nt dimension. The UE can report the data size of the appende CIR, PDP, and / or DP matrix that the UE can support, to the network. The UE can determine the method of appending the LOS / NLOS indicator vectors to the CIR, PDP, and / or DP matrices.
[0122] A Model 0 can be deployed on the network side to distinguish the scenarios where the UE can be located based on the new CIR, PDP, and / or DP matrix reported by the UE. The stages for LCM of Model 0 can include data collection, model training, model inference, and / or model monitoring.
[0123] In the data collection stage, the training data for the LOS / NLOS Model 0 can include a number of CIR, PDP, and / or DP matrices with an additional column of LOS / NLOS indicators. The matrices can be generated on the UE side, and can be collected corresponding to the environment. The training labels for the LOS / NLOS distinguish model can include a number of vectors in which each element can represent an indicator of the environment (e.g., “1” can represent the LOS scenario, “2” can represent the NLOS scenario, “3” can represent the heavy NLOS scenario, etc. ) .
[0124] In the model training stage, the network can train different models using the CIR, PDP, and / or DP matrices measured from the different environments (e.g., training model 1 using the CIR, PDP, and / or DP matrices measured from the LOS environment and training model 2 using CIR, PDP, and / or DP matrices measured from the NLOS environment) .
[0125] In the model inference stage, the UE can report its measurement result of CIR, PDP, and / or DP matrix to the network. In each measurement report process, the data size of reporting can for example be NTRP*Nport* (Nt +1) . After receiving the UE measurement report, the network can input the CIR, PDP, and / or DP matrix into a LOS / NLOS distinguishing model (e.g., model 0) . The LOS / NLOS distinguishing model / model 0 can output the index of the model which corresponds to the specific LOS / NLOS environment. The network can select a suitable model form the output of the LOS / NLOS distinguishing model (model 0) . The final positioning or sensing estimation result can be the output of the selected model.
[0126] In model monitoring stage, the LOS / NLOS distinguish model and the estimation model can be monitored. To monitor the LOS / NLOS distinguish model, UE can report the hard value of the LOS / NLOS indicator according to other / traditional measurement approaches.
[0127] It should be understood that one or more features from the above / following example implementations are not exclusive to the specific example implementations, but can be combined in any manner (e.g., in any priority and / or order, concurrently or otherwise) .
[0128] Example Implementation 6
[0129] FIG. 16 illustrates an example of how the UE can reduce the data size of a measurement report (e.g., without using AI / ML model) . After obtaining several CIR, PDP, and / or DP matrices, the UE can perform Fourier Transform (FT) operations (e.g., fast Fourier transform (FFT) ) on the CIR, PDP, and / or DP matrices measured from multiple slots or time gaps. According to FT operations, the UE can obtain the Doppler information from each propagation path. Each Nt dimension index of the CIR, PDP, and / or DP matrix can include a corresponding Doppler feature matrix. The UE can report a reference CIR, PDP, and / or DP matrix and several Doppler feature matrices. The size of the Doppler feature matrix can be NTRP*Nport*NDoppler wherein NDoppler can be related to the number of slots or time units used for measuring.
[0130] FIG. 17 illustrates an example of a Doppler threshold selection. When the CIR, PDP, and / or DP matrix is / are relatively sparse, the number of Doppler feature matrix obtained from multiple slots or time gaps can be less. Three or more effective paths can be included in the CIR, PDP, and / or DP matrix; three Doppler feature matrices can be included and each can correspond to a propagation path or Nt dimension index. The data size that the UE can report is the sum of a reference CIR, PDP, and / or DP matrix and the Doppler feature matrices (NTRP*Nport*Nt0+Nep*NTRP*Nport*NDoppler) wherein Nep can represent the number of effective propagation paths in the CIR, PDP, and / or DP matrices.
[0131] For the UE to reduce the data size of measurement report in sensing scenario, the UE can calculate the CIR, PDP, and / or DP matrix from multiple slots or time gaps and can obtain a Doppler vector in each propagation path (e.g., Nt dimension index) . The UE can perform a filter on the Doppler feature matrix to select some elements which are over a threshold (e.g., eliminate elements which do not meet the threshold) . FIG. 18 illustrates an example of a relationship between the CIR, PDP, and / or DP matrix and the Doppler feature matrix. The UE can perform / apply threshold selection on the NDoppler dimension. The UE can remove the elements which are above or below the threshold. For example, in FIG. 17, the three Doppler feature matrices can correspond to the propagation path (s) of the reference CIR, PDP, and / or DP matrix in FIG. 16. The Doppler feature matrices can have the same size (e.g., NTRP*Nport*NDoppler) (Nport dimension has been omitted in FIG. 18) . The UE can perform the threshold selection. The Doppler threshold can include a certain number of frequency shifts or velocity and can be performed on each column or on one element of the feature matrix. After threshold filter, the Doppler feature can be / become sparse. The size of Doppler feature matrices after selection can for example be NTRP*NDoppler1, NTRP *NDoppler2 and / or NTRP *NDoppler3. The data size of the UE measurement report can for example be NDoppleri*NTRP*Nport*Nt0 wherein NDoppleri is equal to the number of effective Doppler values after selection.
[0132] FIG. 19 illustrates a flow diagram of a method 1900 of interoperation / communication among a wireless communication device, a first wireless communication node and / or a second wireless communication node. The method 1900 may be implemented using any one or more of the components and devices detailed herein in conjunction with FIGs. 1–18. In overview, the method 1900 may be performed by a wireless communication device (e.g., a UE) , a first wireless communication node (e.g., BS, gNB) and / or a second wireless communication node (e.g., core network, LMF, Sensing Function (SF) ) , in some embodiments. Additional, fewer, or different operations may be performed in the method 1900 depending on the embodiment. At least one aspect of the operations is directed to a system, method, apparatus, or a computer-readable medium.
[0133] With regards to (1905) , and in some embodiments, a first wireless communication node (e.g., base station (BS) ) can send to a wireless communication device (e.g., a user equipment (UE) ) a plurality of signals each associated with a respective one of a plurality of time units The wireless communication device can receive the plurality of signals, and / or obtain a plurality of measurements on the plurality of signals. The wireless communication device can perform data compression and / or feature extraction on the plurality of measurements, to obtain a data-compressed or feature-extracted output across the time unit. The wireless communication device can send the data-compressed or feature-extracted output to the second wireless communication node (e.g., of the core network) , to determine a location or velocity (e.g., of a target object) . The first communication node (e.g., BS, gNB) can send signals to the wireless communication device (e.g., UE) . The wireless communication device (e.g., UE) can receive the signal and can obtain the compressed data. The UE can send data (e.g., compressed data or other information) to the second wireless communication node (e.g., core network) .
[0134] With regards to (1910) , the wireless communication device can receive from the first wireless communication node a plurality of signals each associated with a respective one of a plurality of time units (e.g., time slots) . The first wireless communication node can send to the wireless communication node the plurality of signals each associated with a respective one of a plurality of time units.
[0135] With regards to (1915) , the wireless communication device can obtain a plurality of measurements of the plurality of signals, each corresponding to a respective one of the plurality of time units. The wireless communication device can use its antenna and / or receiver system to capture and / or measure the signals transmitted by the first wireless communication node at specific time intervals corresponding to the time units. The measurements can include data such as a channel impulse response (CIR) , a power delay profile (PDP) , a delay profile (DP) , signal strength, signal quality, timing information, and other parameters that are relevant to the communication method.
[0136] In some embodiments involving multiple resources at each time unit, the wireless communication device can obtain a plurality of measurements by obtaining a first plurality of measurements corresponding to a first one of the plurality of time units, and corresponding to NResource number of resources. The wireless communication device can combine the first plurality of measurements corresponding to the NResource number of resources into a first measurement of the plurality of measurements. The method can include obtaining, by the wireless communication device, a first plurality of measurements corresponding to a first time unit of the plurality of time units and corresponding to a first resource, and a second plurality of measurements corresponding to the first time unit and corresponding to a second resource, for example.
[0137] With regards to (1920) , and in some implementations, the wireless communication device can perform data compression or feature extraction on the / each plurality of measurements, to obtain a data-compressed or feature-extracted output across / over the time units. The wireless communication device may utilize certain algorithms (e.g., Fourier transform processing) and / or neural network / model processing, to reduce the size of the measurement data, which can be advantageous for efficient storage and / or transmission. The data compression and feature extraction processes can be achieved through various techniques such as quantization, encoding, or by identifying and removing redundancies within the measurements. The feature extraction processes can be applied to capture characteristics of the signals over time, which can then be used for further analysis or decision-making processes (e.g., at the network side) .
[0138] The wireless communication device can use a first model to perform the data compression or feature extraction. The model can be based on statistical methods, machine learning, or signal processing techniques that are optimized for the type of data and the desired outcomes. For data compression, the model can implement methods including lossless or lossy compression algorithms, which can minimize the amount of data without significant loss of information. For feature extraction, the model can include the use of techniques such as principal component analysis (PCA) , Fourier transforms, and / or deep learning-based feature detectors to identify and isolate the most important / relevant / significant information from the data / measurements. The size of the data-compressed or feature-extracted output can be smaller than a combined size of the plurality of measurements.
[0139] In some embodiments, the method can include performing, by the wireless communication device, Fourier transform on matrices of the measurements, to obtain Doppler information for each propagation path. The wireless communication device can determine from the Doppler information, one or more Doppler feature matrices. The wireless communication device can send to the second wireless communication node a reference matrix and the one or more Doppler feature matrices.
[0140] In some implementations, the second wireless communication node can use a second model to determine the location or velocity (e.g., of a target) . The second model can determine the location or velocity using any type of form of data-compressed or feature-extracted output (e.g., a reference matrix and the one or more Doppler feature matrices) . In some embodiments, the data-compressed or feature-extracted output can include a compressed feature matrix. Each column or row of the compressed feature matrix can correspond to a respective path (of traversal by a reference signal) and / or each element in a column or row of the matrix can correspond to a velocity of a respective target in a respective path. The wireless communication device can combine the compressed feature matrix with a matrix corresponding to one of the plurality of measurements. By combining the compressed feature matrix with another measurement matrix, the device can enrich / supplement the data context (e.g., with location information) , allowing the model to consider a broader range of variables and potentially improve the completeness, robustness and / or reliability of the location and / or velocity estimations. In some embodiments, the combination can facilitate the identification of patterns or anomalies across different signal paths and / or time units, which can be critical for dynamic environments where the wireless conditions change rapidly.
[0141] In certain embodiments, a series of matrices of at least one of CIR, PDP or DP can be used as training data to train a model to perform the data compression or feature extraction (e.g., at the UE / wireless communication device side) . The training data can include a label comprising a vector or matrix having Doppler information estimated using a difference operation or a defined Doppler estimation method (e.g., legacy methods) . The training data l can be collected on a per-resource basis (e.g., to train models corresponding to the specific resources on the network side) .
[0142] In certain implementations, the method can include applying, by the wireless communication device, a filter to eliminate Doppler elements in the one or more Doppler feature matrices that fail to meet a defined threshold. The defined threshold can include a defined value of frequency shift or velocity. The defined threshold can be applied on each column or each element of the one or more Doppler feature matrices; The size of the feature-extracted output can be NDoppler times of a matrix size (e.g., the same NTRP, NPort and Nt) of each of the plurality of measurements, where NDopplers can be an effective Doppler value.
[0143] In some embodiments, the method can include applying, by the wireless communication device, a filter to eliminate delay elements in the reference matrix that fail to meet a defined threshold. The defined threshold can include a defined number of delays. The defined threshold can be applied on each column or each element of the reference matrix. A size of the output can be Nep times of a matrix size of each of the Doppler feature matrix, where Nep can be a number of effective propagation paths in the reference matrix.
[0144] With regards to (1925) , and in some implementations, the wireless communication device can send to the second wireless communication node, the data-compressed or feature-extracted output to determine a location or velocity (e.g., Doppler) . In some embodiments, the wireless communication device can send an indication of at least one of: a size of measurement data, request or requirement for accuracy of measurement, a method for combining or concatenating matrices, or measurement description information. The wireless communication device can receive from the second wireless communication node responsive to the indication, at least one of: an identifier of a model for the data compression or feature extraction, or delivery of the model for data compression or feature extraction. Regarding the size of measurement data, the method can include reporting, by the wireless communication device to the second wireless communication node, information about the first measurement, wherein a size of the first measurement is Nresource times of a matrix size (e.g., the same NTRP, NPort and Nt) of each of the plurality of measurements. The wireless communication device can send to the second wireless communication node a capability report and / or a functionality report that can include the identifier of the model for data compression or feature extraction.
[0145] The capability report and / or functionality report can include an indication of the wireless communication device’s capability of generating CIR, PDP, and / or DP for the measurements. The second wireless communication node can train at least one part of the network-side model, according to the indication. The capability report may detail the efficiency of the data compression algorithms used, indicative of how the device maximizes data transmission while minimizing bandwidth usage. The functionality report can provide insights into the adaptability of the feature extraction process, indicative of how the wireless communication device can adjust its methods in response to varying signal conditions or requirements set by the second wireless communication node. The capability report and / or functionality report can assist the second wireless communication node in optimizing the overall communication system, ensuring that it operates effectively within the specific constraints (e.g., scenarios) of the network architecture and / or available resources. The capability report can include an indication of at least one of: a number of resources (e.g., NResource) that the wireless communication device can measure; or the matrix size of each of the plurality of measurements. In certain embodiments, the method can include combining / compressing / processing the first plurality of measurements, by providing the first measurement as a reference matrix for one of the resources, and a list of difference information between the reference matrix and each matrix of others of the resources.
[0146] In certain embodiments, the method can include sending, by the wireless communication device to the second wireless communication node, an indicator of line-of-sight (LOS) or non line-of-sight (NLOS) , for use to select a model (on the network or wireless communication node side) to determine the location or velocity. The indicator can be included in (e.g., incorporated or combined into) a matrix of the plurality of measurements. The indicator can include one or more vector elements each representing a probability of LOS propagation path between a respective transmit-receive point (TRP) and the wireless communication device. One or more vector elements can form additional columns of the matrix. A model (e.g., Model 0 at the network or wireless communication node side) can be used to detect a measurement with the indicator. The training data for training the model can include a number of matrices that incorporates the indicator, and can include training labels each comprising vector elements that each represents an indication of an LOS or NLOS scenario. A respective model (at the network or wireless communication node side) can be trained for each LOS or NLOS scenario. Such a model can determine for the measurement with the indicator, an index of the respective model to select. The location or velocity can be determined using the respective model that is selected.
[0147] With regards to (1930) , and in some implementations, the second wireless communication node can receive from the wireless communication device, the data-compressed or feature-extracted output to determine a location or velocity (e.g., Doppler) . In some embodiments, the second wireless communication node can receive from the wireless communication device information including but not limited to a reference matrix and / or one or more Doppler feature matrices. The second wireless communication node (e.g., core network) can receive the UE measurement and can estimate the location and velocity of one or more targets.
[0148] In some embodiments, the second wireless communication node can send the wireless communication device assistance data comprising an indication of at least one of: resources (e.g., allocated / configured for the plurality of signals) , or models corresponding to the resources. In some implementations, the wireless communication device obtains a first plurality of measurements corresponding to a first time unit of the plurality of time units and corresponding to a first resource, and a second plurality of measurements corresponding to the first time unit and corresponding to a second resource. The second wireless communication node can receive the first plurality of measurements for a first model (e.g., decoder or neural network) to determine a first estimation. The second wireless communication node can receive the second plurality of measurements for a second model to determine a second estimation; wherein the location or velocity can be determined using the first estimation and the second estimation. The location or velocity can be determined using a third model that receives as input at least the first estimation and the second estimation, and / or by applying respective weights on the first estimation and the second estimation.
[0149] While various embodiments of the present solution have been described above, it should be understood that they have been presented by way of example only, and not by way of limitation. Likewise, the various diagrams may depict an example architectural or configuration, which are provided to enable persons of ordinary skill in the art to understand example features and functions of the present solution. Such persons would understand, however, that the solution is not restricted to the illustrated example architectures or configurations, but can be implemented using a variety of alternative architectures and configurations. Additionally, as would be understood by persons of ordinary skill in the art, one or more features of one embodiment can be combined with one or more features of another embodiment described herein. Thus, the breadth and scope of the present disclosure should not be limited by any of the above-described illustrative embodiments.
[0150] It is also understood that any reference to an element herein using a designation such as "first, " "second, " and so forth does not generally limit the quantity or order of those elements. Rather, these designations can be used herein as a convenient means of distinguishing between two or more elements or instances of an element. Thus, a reference to first and second elements does not mean that only two elements can be employed, or that the first element must precede the second element in some manner.
[0151] Additionally, a person having ordinary skill in the art would understand that information and signals can be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits and symbols, for example, which may be referenced in the above description can be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.
[0152] A person of ordinary skill in the art would further appreciate that any of the various illustrative logical blocks, modules, processors, means, circuits, methods and functions described in connection with the aspects disclosed herein can be implemented by electronic hardware (e.g., a digital implementation, an analog implementation, or a combination of the two) , firmware, various forms of program or design code incorporating instructions (which can be referred to herein, for convenience, as "software" or a "software module) , or any combination of these techniques. To clearly illustrate this interchangeability of hardware, firmware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware, firmware or software, or a combination of these techniques, depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in various ways for each particular application, but such implementation decisions do not cause a departure from the scope of the present disclosure.
[0153] Furthermore, a person of ordinary skill in the art would understand that various illustrative logical blocks, modules, devices, components and circuits described herein can be implemented within or performed by an integrated circuit (IC) that can include a general purpose processor, a digital signal processor (DSP) , an application specific integrated circuit (ASIC) , a field programmable gate array (FPGA) or other programmable logic device, or any combination thereof. The logical blocks, modules, and circuits can further include antennas and / or transceivers to communicate with various components within the network or within the device. A general purpose processor can be a microprocessor, but in the alternative, the processor can be any conventional processor, controller, or state machine. A processor can also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other suitable configuration to perform the functions described herein.
[0154] If implemented in software, the functions can be stored as one or more instructions or code on a computer-readable medium. Thus, the steps of a method or algorithm disclosed herein can be implemented as software stored on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that can be enabled to transfer a computer program or code from one place to another. A storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer.
[0155] In this document, the term "module" as used herein, refers to software, firmware, hardware, and any combination of these elements for performing the associated functions described herein. Additionally, for purpose of discussion, the various modules are described as discrete modules; however, as would be apparent to one of ordinary skill in the art, two or more modules may be combined to form a single module that performs the associated functions according to embodiments of the present solution.
[0156] Additionally, memory or other storage, as well as communication components, may be employed in embodiments of the present solution. It will be appreciated that, for clarity purposes, the above description has described embodiments of the present solution with reference to different functional units and processors. However, it will be apparent that any suitable distribution of functionality between different functional units, processing logic elements or domains may be used without detracting from the present solution. For example, functionality illustrated to be performed by separate processing logic elements, or controllers, may be performed by the same processing logic element, or controller. Hence, references to specific functional units are only references to a suitable means for providing the described functionality, rather than indicative of a strict logical or physical structure or organization.
[0157] Various modifications to the embodiments described in this disclosure will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the scope of this disclosure. Thus, the disclosure is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the novel features and principles disclosed herein, as recited in the claims below.
Claims
1.A method comprising:receiving, by a wireless communication device from a first wireless communication node, a plurality of signals each associated with a respective one of a plurality of time units;obtaining, by the wireless communication device, a plurality of measurements of the plurality of signals, each corresponding to a respective one of the plurality of time units;performing, by the wireless communication device, data compression or feature extraction on the plurality of measurements, to obtain a data-compressed or feature-extracted output across the time units; andsending, by the wireless communication device to the second wireless communication node, the data-compressed or feature-extracted output to determine a location or velocity.2.The method of claim 1, wherein the plurality of measurements comprises at least one of: a channel impulse response (CIR) , a power delay profile (PDP) , or a delay profile (DP) .3.The method of claim 1, wherein at least one of:the wireless communication device uses a first model to perform the data compression or feature extraction;a size of the data-compressed or feature-extracted output is smaller than a combined size of the plurality of measurements;the second wireless communication node uses a second model to determine the location or velocity.4.The method of claim 1, wherein at least one of:the data-compressed or feature-extracted output comprises a compressed feature matrix;each column or row of the matrix corresponds to a respective path; oreach element in a column or row of the matrix corresponds to a velocity of a respective target in a respective path.5.The method of claim 4, comprising:combining, by the wireless communication device, the compressed feature matrix with a matrix corresponding to one of the plurality of measurements.6.The method of claim 2, wherein at least one of:a series of matrices of at least one of CIR, PDP or DP is used as training data to train a model to perform the data compression or feature extraction; orthe training data includes a label comprising a vector or matrix having Doppler information estimated using a difference operation or a defined Doppler estimation method.7.The method of claim 1, comprising at least one of:sending, by the wireless communication device to the second wireless communication node, an indication of at least one of: a size of measurement data, request or requirement for accuracy of measurement, a method for combining or concatenating matrices, or other measurement description information;receiving, by the wireless communication device from the second wireless communication node responsive to the indication, at least one of: an identifier of a model for the data compression or feature extraction, or delivery of the model for data compression or feature extraction; orsending, by the wireless communication device to the second wireless communication node, a capability report comprising the identifier of the model for data compression or feature extraction.8.The method of claim 2, comprising:sending, by the wireless communication device to the second wireless communication node, a capability report or functionality report comprising an indication of the wireless communication device’s capability of generating CIR, PDP or DP for the measurements.9.The method of claim 7, wherein at least one of:the second wireless communication node selects, according to the indication, at least one part of a model to be used to determine the location or velocity; orthe second wireless communication node trains at least one part of the model, according to the indication.10.The method of claim 1, comprising obtaining, by the wireless communication device, a first of the plurality of measurements of the plurality of signals, by:obtaining, by the wireless communication device, a first plurality of measurements corresponding to a first one of the plurality of time units, and corresponding to NResource number of resources; andcombining, by the wireless communication device, the first plurality of measurements corresponding to the NResource number of resources into a first measurement of the plurality of measurements.11.The method of claim 10, comprising:reporting, by the wireless communication device to the second wireless communication node, the first measurement,wherein a size of the first measurement is NResource times of a matrix size of each of the plurality of measurements.12.The method of claim 10, comprising:sending, by the wireless communication device to the second wireless communication node, a capability report comprising an indication of at least one of:a number of resources that the wireless communication device can measure; orthe matrix size of each of the plurality of measurements.13.The method of claim 10, wherein combining the first plurality of measurements into the first measurement comprises:providing the first measurement as: a reference matrix for one of the resources, and a list of difference information between the reference matrix and each matrix of others of the resources.14.The method of claim 1, comprising:obtaining, by the wireless communication device, a first plurality of measurements corresponding to a first time unit of the plurality of time units and corresponding to a first resource, and a second plurality of measurements corresponding to the first time unit and corresponding to a second resource; andsending, by the wireless communication device, the first plurality of measurements to the second wireless communication node, for a first model to determine a first estimation; andsending, by the wireless communication device, the second plurality of measurements to the second wireless communication node, for a second model to determine a second estimation,wherein the location or velocity is determined using the first estimation and the second estimation.15.The method of claim 14, wherein at least one of:training data for training the first model is collected for the first resource; orthe training data includes a label determined for the first resource.16.The method of claim 14, wherein the location or velocity is determined using a third model that receives as input the first estimation and the second estimation.17.The method of claim 14, wherein the location or velocity is determined by applying respective weights on first estimation and the second estimation.18.The method of claim 14, comprising:receiving, by the wireless communication device from the second wireless communication node, assistance data comprising an indication of at least one of: the resources, or models corresponding to the resources.19.The method of claim 1, comprising:sending, by the wireless communication device to the second wireless communication node, an indicator of line-of-sight (LOS) or non line-of-sight (NLOS) , for use to select a model to determine the location or velocity.20.The method of claim 19, wherein at least one of:the indicator is included in a matrix of the plurality of measurements;the indicator includes one or more vector elements each representing a probability of LOS propagation path between a first wireless communication node and the wireless communication device; orthe one or more vector elements form additional columns of the matrix.21.The method of claim 20, wherein at least one of:a model is used to detect a measurement with the indicator;training data for training the model comprises a number of matrices that incorporates the indicator;the training data includes training labels each comprising vector elements that each represents an indication of an LOS or NLOS scenario;a respective model is trained for each LOS or NLOS scenario;the model determines, for the measurement with the indicator, an index of the respective model to select; orthe location or velocity is determined using the respective model that is selected.22.The method of claim 1, comprising:performing, by the wireless communication device, Fourier transform on matrices of the measurements, to obtain Doppler information for each propagation path;determining, by the wireless communication device from the Doppler information, one or more Doppler feature matrices; andsending, by the wireless communication device to the second wireless communication node, a reference matrix and the one or more Doppler feature matrices.23.The method of claim 22, wherein at least one of:each column or row of the reference matrix corresponds to a respective path;each of the one or more Doppler feature matrices corresponds to a column or row of the reference matrix; oreach element in one of the one or more Doppler feature matrices corresponds to a velocity of a respective target in a respective path.24.The method of claim 22, comprising:applying, by the wireless communication device, a filter to eliminate delay elements in the reference matrix that fail to meet a defined threshold.25.The method of claim 24, wherein at least one of:the defined threshold comprises a defined number of delays;the defined threshold is applied on each column or each element of the reference matrix; ora size of the output is Nep times of a matrix size of each of the Doppler feature matrix, where Nep is a number of effective propagation paths in the reference matrix.26.The method of claim 22, comprising:applying, by the wireless communication device, a filter to eliminate Doppler elements in the one or more Doppler feature matrices that fail to meet a defined threshold.27.The method of claim 26, wherein at least one of:the defined threshold comprises a defined number of frequency shifts or velocity;the defined threshold is applied on each column or each element of the one or more Doppler feature matrices; ora size of the feature-extracted output is NDoppler times of a matrix size of each of the plurality of measurements, where NDoppler is the number of effective Doppler value in the Doppler feature matrix.28.A method comprising:sending, by a first wireless communication node to a wireless communication device, a plurality of signals each associated with a respective one of a plurality of time units, for the wireless communication device to obtain a plurality of measurements of the plurality of signals, each corresponding to a respective one of the plurality of time units, and to perform data compression or feature extraction on the plurality of measurements, to obtain a data-compressed or feature-extracted output across the time units; andreceiving, by the second wireless communication node from the wireless communication device, the feature-extracted output to determine a location or velocity.29.A non-transitory computer readable medium storing instructions, which when executed by at least one processor, cause the at least one processor to perform the method of any one of claims 1-28.30.An apparatus comprising:at least one processor configured to implement the method of any one of claims 1-28.
Citation Information
Patent Citations
Method, device and system for positioning terminal equipment
CN110022523A
High performance positioning for wireless communication using artificial neural networks
EP4231041A2
Hybrid model-learning solution for CSI reporting
WO2023140772A1
User equipment (UE) positioning with frequency domain windowing
WO2024005920A1