A terminal positioning method, an electronic device, and a storage medium

Abstract

The application provides a terminal positioning method, an electronic device and a storage medium, and relates to the technical field of terminals. The method comprises the following steps: a base station first acquires a target uplink reference signal sent by a target terminal or a PRU, and extracts corresponding target channel impulse response information. Then, the target channel impulse response information is input into a channel measurement mode identification model to obtain a target channel measurement mode. Then, it is determined whether the target channel measurement mode is the same as a channel measurement mode issued by an LMF. If the target channel measurement mode is the same as the channel measurement mode issued by the LMF, channel measurement is performed by using the target channel measurement mode to obtain a channel measurement result. If the target channel measurement mode is different from the channel measurement mode issued by the LMF, a channel measurement mode switching request is sent to the LMF, and channel measurement is performed by using a measurement mode fed back by the LMF in response to the switching request to obtain a channel measurement result. Then, the channel measurement result is sent to the LMF for positioning of the target terminal, and a more accurate positioning result is obtained, so that an effective balance between positioning accuracy improvement and network resource consumption is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of terminals, and in particular to a terminal positioning method, an electronic device, and a storage medium. BACKGROUND

[0002] With the continuous development of science and technology, terminal devices such as mobile phones, tablet computers, and smart watches are increasingly applied in people's life and work, bringing great convenience to people. In the 5th Generation Mobile Communication Technology (5G) network, the Internet of Things and intelligence have put forward higher requirements for services based on terminal device positions. Therefore, it is particularly important to achieve accurate positioning of terminal devices.

[0003] Currently, traditional positioning technologies such as Downlink Time Difference of Arrival (DL-TDOA) and Downlink Angle of Arrival (DL-AOA) will produce measurement error accumulation due to multipath effects in complex electromagnetic propagation scenarios such as dense urban areas, thereby reducing the positioning accuracy of terminal devices. In order to overcome this problem, a positioning enhancement scheme based on Artificial Intelligence (AI) and Machine Learning (ML) is usually used to utilize the powerful nonlinear modeling capability of AI and ML to achieve more accurate positioning of terminal devices. However, in this AI / ML-based positioning enhancement scheme, when a cross-network element deployment mode is adopted, i.e., when the channel measurement and positioning calculation model are deployed on different nodes of the network, a large amount of Channel State Information (CSI) needs to be transmitted across domains to assist positioning, thereby greatly increasing the signaling overhead in the 5G network. SUMMARY

[0004] To solve the above problems, the present application provides a terminal positioning method, an electronic device and a storage medium. In order to solve the problem of large signaling overhead in the AI / ML-based terminal positioning process, the appropriate channel measurement method is adaptively selected to better balance the signaling overhead and positioning accuracy. That is, when the positioning accuracy meets the requirements, switch to a channel measurement method that reduces signaling overhead, such as a path-based channel measurement (Path-based) method, etc. Conversely, when the positioning accuracy is low (i.e. does not meet the requirements), switch to a channel measurement method that improves positioning accuracy, such as a sample-based channel measurement (Sample-based) method, etc., so as to achieve effective balance between positioning accuracy improvement and network resource consumption.

[0005] In the first aspect, the present application provides a terminal positioning method, which comprises: when it is necessary to position any target terminal, the next generation base station (Next Generation NodeB, gNB) can first obtain the target uplink reference signal sent by the target terminal or PRU, and extract the corresponding target channel impulse response information. Then input the target channel impulse response information into the pre-constructed channel measurement method identification model to obtain the target channel measurement method (such as Path-based measurement method or Sample-based measurement method). Then judge whether the target channel measurement method is the same as the channel measurement method issued by the location management function (Location Management Function, LMF) network element, if the same, the target channel measurement method can be used to measure the current channel environment to obtain the channel measurement result, and send the channel measurement result to the LMF network element, otherwise, if not the same, send the channel measurement method switching request to the LMF network element, and use the measurement method fed back by the LMF network element in response to the switching request to measure the current channel environment to obtain the channel measurement result; then send the obtained channel measurement result to the LMF network element, so that the LMF network element can position the target terminal according to the channel measurement result to obtain the positioning result.

[0006] It can be seen that in the terminal positioning method, the embodiment pre-trains a base model (such as a convolutional neural network (CNN) model) based on a model-agnostic meta-learning (MAML) algorithm, and uses sample uplink reference signals (specifically, sample channel impulse response information corresponding thereto) labeled in advance in different channel measurement manners (specifically, Path-based measurement manner and Sample-based measurement manner) to obtain a channel measurement manner identification model, so that the target channel measurement manner finally output by the model can reduce signaling overhead while ensuring positioning accuracy, so that when the channel measurement result measured by using the target channel measurement manner is used to position the target terminal, the effective balance between positioning accuracy improvement and network resource consumption can be achieved, and thus the positioning effect is improved.

[0007] In a possible implementation, the channel measurement manner identification model is constructed in the following manner: a first sample uplink reference signal in a general scenario is obtained; based on the MAML algorithm, the first sample uplink reference signal and a first loss function are used to optimize initialization parameters of an initial scene identification model, to obtain an optimized scene identification model; a second sample uplink reference signal is obtained; sample channel impulse response information corresponding to the second sample uplink reference signal is extracted; the sample channel impulse response information and a second loss function are used to train the optimized scene identification model for a channel measurement manner classification task, to obtain the channel measurement manner identification model, so that the identification accuracy of the channel measurement manner identification model for the target channel measurement manner can be improved, and the channel measurement manner identification model can reduce signaling overhead while ensuring positioning accuracy.

[0008] In a possible implementation, the initial scene identification model is a CNN model.

[0009] In a possible implementation, the first loss function and the second loss function are both cross-entropy loss functions, which are used to improve the training accuracy of the model parameters.

[0010] In a possible implementation, the optimized scene recognition model is trained for a channel measurement mode classification task by using sample channel impulse response information and a second loss function, to obtain a channel measurement mode recognition model, including: measuring a channel environment in which a second sample uplink reference signal is located by using a first channel measurement mode and a second channel measurement mode, to obtain a first channel measurement result and a second channel measurement result; sending the first channel measurement result and the second channel measurement result to an LMF network element, so that the LMF network element respectively performs positioning on a terminal according to the first channel measurement result and the second channel measurement result, to obtain a first positioning result and a second positioning result; determining a positioning error corresponding to the first channel measurement mode according to the first positioning result; and determining a positioning error corresponding to the second channel measurement mode according to the second positioning result; labeling the sample channel impulse response information for the channel measurement mode according to the positioning error corresponding to the first channel measurement mode and the positioning error corresponding to the second channel measurement mode, to obtain a channel measurement mode labeling result of the sample channel impulse response information; inputting the sample channel impulse response information into the optimized scene recognition model for channel measurement mode classification processing, to obtain a classification result of the channel measurement mode corresponding to the sample channel impulse response information; calculating a value of the second loss function by using the channel measurement mode labeling result of the sample channel impulse response information and the classification result of the channel measurement mode corresponding to the sample channel impulse response information, and training the optimized scene recognition model by using the value and a gradient descent method until a preset condition (such as a maximum set training number of times is reached or a change value of a model parameter is less than a set threshold value) is met, then stopping updating of the model parameter, to obtain the trained channel measurement mode recognition model, so that the recognition accuracy of the model can be improved.

[0011] In a possible implementation, the second sample uplink reference signal is sent by a positioning reference unit (PRU); and the positioning error corresponding to the first channel measurement mode is determined according to the first positioning result, and the positioning error corresponding to the second channel measurement mode is determined according to the second positioning result, including: obtaining physical position information of the PRU; calculating the positioning error corresponding to the first channel measurement mode according to a difference between the first positioning result and the physical position information of the PRU; and calculating the positioning error corresponding to the second channel measurement mode according to a difference between the second positioning result and the physical position information of the PRU, so that the calculation accuracy of the positioning errors corresponding to the two channel measurement modes can be improved.

[0012] In a possible implementation, the first channel measurement manner is a Path-based measurement manner; the second channel measurement manner is a Sample-based measurement manner; and the sample channel impulse response information is labeled with the channel measurement manner according to a positioning error corresponding to the first channel measurement manner and a positioning error corresponding to the second channel measurement manner, to obtain a channel measurement manner labeling result of the sample channel impulse response information, including: calculating a difference between the positioning error corresponding to the first channel measurement manner and the positioning error corresponding to the second channel measurement manner as a positioning comparison result of the first channel measurement manner and the second channel measurement manner; determining whether the positioning comparison result is not greater than a preset threshold; if yes, labeling the channel measurement manner of the sample channel impulse response information as the Path-based measurement manner as the channel measurement manner labeling result of the sample channel impulse response information; and if not, labeling the channel measurement manner of the sample channel impulse response information as the Sample-based measurement manner as the channel measurement manner labeling result of the sample channel impulse response information. In this way, the accuracy of the channel measurement manner labeling result of the sample channel impulse response information can be improved.

[0013] In a possible implementation, the preset threshold has a value of 0.1 meter or 0.5 meter.

[0014] In a possible implementation, the target channel measurement manner is the Path-based measurement manner or the Sample-based measurement manner.

[0015] In a second aspect, the present application provides an electronic device, including a memory and a processor, the memory stores a computer program, and the processor is configured to invoke and execute the computer program to implement the terminal positioning method in any one of the above first aspect.

[0016] In a third aspect, the present application provides a computer readable storage medium, which stores a computer program, and the program is run by a processor of an electronic device to implement the terminal positioning method in any one of the above first aspect.

[0017] In a fourth aspect, the present application provides a computer program product, which, when run on a computer, causes the computer to execute the terminal positioning method in any one of the above first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A signaling diagram for implementing terminal positioning based on AI / ML Case 3b is provided for the embodiments of the present application;

[0019] Figure 2 A signaling diagram for a terminal positioning method is provided for the embodiments of the present application.

[0020] Figure 3 A schematic diagram of a terminal device provided for an embodiment of the application;

[0021] Figure 4 A software structure block diagram of a terminal device provided for an embodiment of the application;

[0022] Figure 5 A signaling diagram of a channel measurement mode identification model construction process provided for an embodiment of the application;

[0023] Figure 6 A construction process schematic diagram of a channel measurement mode identification model provided for an embodiment of the application;

[0024] Figure 7 A signaling diagram of a Case 2b terminal positioning implementation based on AI / ML provided for an embodiment of the application;

[0025] Figure 8 A signaling diagram of another terminal positioning method provided for an embodiment of the application;

[0026] Figure 9 A software structure block diagram of another terminal device provided for an embodiment of the application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. The terms used in the following embodiments are only for the purpose of describing the specific embodiments and are not intended to be limiting to the present application. As used in the specification and the appended claims of the present application, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0028] In this specification, the phrase “one embodiment” or “some embodiments” etc. means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrase “in one embodiment,” “in some embodiments,” “in other embodiments,” “in additional embodiments,” etc. in various places in the specification are not necessarily all referring to the same embodiment, unless otherwise specifically noted. The terms “comprises,” “comprising,” “includes,” “including,” “has,” “having” and the like are meant to be open-ended terms that specifically permit the inclusion of unspecified elements, and the terms “consisting of” and the like are meant to be closed terms that do not permit the inclusion of unspecified elements.

[0029] The plurality referred to in the embodiments of the present application refers to greater than or equal to two. It should be noted that in the description of the embodiments of the present application, the words "first", "second", and the like are only used for the purpose of distinguishing the description, and cannot be understood as indicating or implying relative importance, nor can it be understood as indicating or implying order.

[0030] In order for those skilled in the art to more clearly understand the scheme of the present application, first, the AI / ML-based positioning scenario related to the technical scheme of the present application is described.

[0031] It should be noted that with the evolution of 5G technology, downlink time difference of arrival (DL-TDOA) and downlink angle of arrival (DL-AOA) positioning technologies have been gradually standardized and deployed. However, due to the complex physical environment in cities, the multipath effect of measurement signals caused by reflection of buildings and diffraction of obstacles is significantly aggravated in complex electromagnetic propagation scenarios such as dense urban areas, resulting in a serious measurement error accumulation problem for these two geometric relationship-based positioning algorithms, and the positioning accuracy shows a clear downward trend, which has significantly restricted the practical application of high-precision positioning services.

[0032] Therefore, in order to overcome the above problems, AI / ML-based positioning enhancement schemes are now more commonly used to utilize the powerful nonlinear modeling capabilities of AI and ML to provide a new technical path for multipath channel feature analysis, thereby achieving more accurate positioning of terminal devices. Currently, in the 3GPP standardization work description of AI / ML applications in 5G positioning technology, five cases are mainly studied, and for Case 2b and Case 3b, which are cross-network element deployment modes, i.e., the channel measurement and positioning calculation model are placed in different network nodes, the positioning calculation model is deployed on the LMF side, and the channel measurement is completed by the terminal (User Equipment, UE) or gNB. Therefore, the UE or gNB needs to report the channel measurement results to the LMF, and this cross-domain transmission of massive CSI data will significantly increase the signaling overhead of the network.

[0033] Specifically, as shown in the signaling diagram for implementing terminal positioning based on AI / ML of Case 3b, in the Case 3b positioning mode, when the positioning calculation model is deployed on the LMF side, the specific positioning process includes the following steps S101-S104: Figure 1

[0034] S101: The gNB acquires the uplink reference signal sent by the UE.

[0035] ​The specific content of the uplink reference signal is not limited, and can be, but is not limited to, a sounding reference signal (SRS) or a physical random access channel (PRACH) signal.

[0036] S102: The gNB performs channel measurement on the uplink reference signal sent by the UE to obtain channel measurement results.

[0037] The specific content of the channel measurement results is not limited, and can include but is not limited to a channel impulse response (CIR) in the time domain, a transmission time, a measurement parameter, a measurement method used, and the like.

[0038] S103: The gNB sends the channel measurement results to the LMF.

[0039] S104: The LMF performs positioning on the UE according to the channel measurement results by using a positioning calculation model.

[0040] It can be seen that in the Case 3b positioning mode, since the channel measurement is completed by the gNB, the gNB needs to send a large amount of channel measurement data (Measurement results) to the LMF, which greatly increases the signaling overhead of the network, and can cause excessive consumption of core network transmission resources, and even more likely to impact existing radio resource management. Therefore, an optimized terminal positioning method is needed to reduce the signaling overhead while ensuring the positioning accuracy, so as to achieve an effective balance between the positioning accuracy improvement and the network resource consumption.

[0041] To achieve effective balance between positioning accuracy improvement and network resource consumption, the present application provides a terminal positioning method. First, based on the Model-Agnostic Meta-Learning (MAML) algorithm, the uplink reference signal (specifically, the sample channel impulse response information corresponding thereto) labeled in advance in different channel measurement modes (specifically, Path-based measurement mode and Sample-based measurement mode) is used to train the basic model (such as CNN model), and a channel measurement mode identification model is obtained. The subsequent embodiments will introduce the model in detail. Then, after obtaining the uplink reference signal (hereinafter defined as target uplink reference signal) (such as sent by the UE or Positioning Reference Unit (PRU)), the target channel impulse response information corresponding to the target uplink reference signal can be extracted using existing or future information extraction methods. Next, the target channel impulse response information can be input into the pre-constructed channel measurement mode identification model to identify a more suitable channel measurement mode (hereinafter defined as target channel measurement mode, which can be Path-based measurement mode or Sample-based measurement mode). Then, it is determined whether the target channel measurement mode is the same as the channel measurement mode issued by the LMF network element; if so, the target channel measurement mode can be used to measure the current channel environment to obtain a channel measurement result; otherwise, if not, a channel measurement mode switching request can be sent to the LMF network element, and the measurement mode fed back by the LMF in response to the switching request can be used to measure the current channel environment to obtain a channel measurement result. The obtained channel measurement result is sent to the LMF network element, so that the LMF network element can accurately position the target terminal according to the channel measurement result to obtain a positioning result. In this way, by adaptively selecting a more suitable target channel measurement mode through the channel measurement mode identification model, the signaling overhead and positioning accuracy can be better balanced. For example, when the positioning accuracy meets the requirements, the channel measurement mode with reduced signaling overhead, such as Path-based measurement mode, can be switched to; otherwise, when the positioning accuracy is low (i.e., does not meet the requirements), the channel measurement mode with improved positioning accuracy, such as Sample-based measurement mode, can be switched to, thereby achieving effective balance between positioning accuracy improvement and network resource consumption.

[0042] Next, the specific implementation process of the terminal positioning method provided by the present application will be introduced in detail:

[0043] In some embodiments, to solve the problem that Figure 1To solve the problem of large network signaling overhead in the Case 3b positioning mode shown, the pre-constructed channel measurement mode recognition model is deployed to the gNB, and an optimized terminal positioning method is provided accordingly, which is applied to the gNB to achieve effective balance between positioning accuracy improvement and network resource consumption in the Case 3b positioning mode. As shown in Figure 2 The specific implementation process of the terminal positioning method can include the following steps S201-S208:

[0044] S201: The gNB obtains the target uplink reference signal (such as SRS) sent by the UE or PRU.

[0045] In this embodiment, any terminal device performing positioning by this embodiment is defined as a target terminal, and the uplink reference signal (such as SRS) sent by the target terminal (or PRU) is defined as a target uplink reference signal, but the type and structure of the target terminal and the type, content and sending mode of the target uplink reference signal are not limited in this embodiment. For example, as shown in Figure 2 The target uplink reference signal sent by the UE or PRU to the gNB can be an SRS signal, which is used for the base station (gNB) to estimate the uplink channel state information (CSI). Specifically, the gNB can use existing or future signal feature extraction methods to obtain channel impulse response information (CIR) by measuring the characteristics of the wireless channel where the target uplink reference signal (such as SRS) is located, which is defined as target channel impulse response information here.

[0046] The channel impulse response information (CIR) can be understood as the embodiment of the channel state information (CSI) in the time domain, which reflects the time-varying characteristics and multipath propagation characteristics of the channel.

[0047] S202: The gNB inputs the target channel impulse response information into the pre-constructed channel measurement mode recognition model to obtain the target channel measurement mode.

[0048] In this embodiment, after the gNB obtains the target uplink reference signal (such as SRS) sent by the UE or PRU through step S201, further, in order to adaptively select a more suitable channel measurement mode, the target channel impulse response information can be input into the pre-constructed channel measurement mode recognition model to identify a more suitable channel measurement mode (hereinafter defined as the target channel measurement mode) that can balance the signaling overhead and positioning accuracy, which is used to perform the subsequent step S203.

[0049] The specific content of the target channel measurement mode is not limited, and can be, but is not limited to, a Path-based measurement mode or a Sample-based measurement mode, etc. The Sample-based measurement mode is a channel measurement mode based on uniform sampling, which uniformly samples channel information first, and then extracts part of the sampling points; and the Path-based measurement mode is a sampling mode based on a path detection algorithm, which samples channel information when the algorithm detects a signal path. The Sample-based measurement mode needs to uniformly sample channel information, and the amount of measurement data is large, and there is currently no related protocol content, and a related standard needs to be additionally formulated; the Path-based measurement mode only samples after detecting a signal path, and the amount of measurement data is small, and can be compatible with the existing protocol without the need to additionally formulate a related standard. The positioning model based on the Sample-based measurement mode has a large improvement in positioning accuracy, and still has high positioning accuracy in a dense multipath environment (Non-Line-of-Sight Environment, NLoS) scene; the positioning model based on the Path-based measurement mode has little difference in accuracy from the positioning model based on the Sample-based measurement mode in an open environment (Line-of-Sight Environment, LoS) scene, but performs poorly in a dense multipath scene. Therefore, the Path-based measurement mode and the Sample-based measurement mode have different signaling overheads, the Path-based measurement mode has lower signaling overhead and the positioning accuracy in an open scene can meet the requirements, and the Sample-based measurement mode has higher signaling overhead, but its positioning accuracy is higher than that of the Path-based measurement mode.

[0050] Next, the construction process of the channel measurement mode identification model will be described in detail. The specific construction process can include the following steps A-D:

[0051] Step A: Obtain a first sample uplink reference signal in a general scene.

[0052] It should be noted that the trained measurement mode identification model is deployed on the gNB side in this embodiment. In order to ensure that the identification model can quickly adapt to the identification task in different physical environments, the model needs to be initialized and trained before deployment. Since it is difficult to collect a large amount of sample data in the actual network environment for model training, the MAML algorithm is used to optimize the initial parameters of the model in this application, so that the trained identification model can adapt to the channel measurement mode classification task and complete the migration of the channel measurement mode identification model through only a few gradient updates in the case of small samples.

[0053] Among them, model-agnostic meta-learning (MAML) is a powerful meta-learning framework designed to quickly adapt to new tasks with only a few gradient updates. The reason why the present application uses MAML for model training is that MAML is not only suitable for small sample classification, but also for regression and reinforcement learning tasks. In the small sample learning scenario, the purpose of model-agnostic meta-learning is to find a model initialization parameter that can achieve good performance through a small number of gradient updates for a series of different tasks (such as channel measurement method classification tasks).

[0054] Specifically, suppose there is a task distribution , where each task contains a data set , which is composed of a training set and a test set . The goal is to obtain a general model initialization parameter through training under the condition of a small amount of training data, so that the model can quickly adapt to new tasks. The optimization goal of the model is to learn an optimal initialization parameter , which can quickly adapt to the task through a small number of gradient updates. The optimization goal of MAML can be defined as follows:

[0055]

[0056] , where represents the loss function on task ; and represents the parameter obtained by updating the initial parameter using the training data of task .

[0057] And, MAML can specifically use the following two steps to optimize parameters:

[0058] 1) Task internal update: for each task , the model parameter will be updated to the task-specific parameter through gradient descent, and the specific calculation formula is as follows:

[0059]

[0060] , where represents the learning rate of the inner loop, and the specific value is not limited and can be determined according to actual conditions and experience values, such as taking the value of 0.01, etc. ; and represents the gradient descent algorithm function to find the minimum value of the function through iteration, that is, to minimize the loss function.

[0061] 2) Meta-optimization update: update the parameters on the test set of the task and minimize the overall loss by optimizing the initial parameters The overall meta-optimization objective is calculated as follows:

[0062]

[0063] where the outer gradient is backpropagated through the chain rule, and the specific calculation formula is as follows:

[0064]

[0065] In this way, when the preset training number (such as 1000 times) is reached, or the change of the model initialization parameter is less than the threshold value (such as 0.01 or the like), the initialization training of the model can be completed.

[0066] In the process of constructing the channel measurement mode recognition model, when the MAML algorithm is used to initialize and train the model parameters, the uplink reference signals (such as SRS signals in LoS / NLoS scenarios) in various general scenarios (such as stations, shopping malls, parks, etc.) can be collected first to construct an initial model training data set. For example, assuming that N (the specific value is not limited, which can be any positive integer greater than 0) uplink reference signals (such as SRS signals in LoS / NLoS scenarios) in various scenarios are collected as the first sample uplink reference signal, which is used to perform the following step B.

[0067] Step B: based on the MAML algorithm, the first sample uplink reference signal and the first loss function are used to optimize the initialization parameters of the initial scene recognition model, and an optimized scene recognition model is obtained.

[0068] In this embodiment, after obtaining the first sample uplink reference signal (such as SRS signal in LoS / NLoS scenario) in N general scenarios through step A, further, based on the MAML algorithm, the first sample uplink reference signal and the first loss function are used to optimize the initialization parameters of the initial scene recognition model, and an optimized scene recognition model is obtained, which is used to perform the following step C.

[0069] Specifically, the training set and test set in the i-th (the specific value is not limited, which can be any positive integer greater than 0 and not greater than N) scenario of the first sample uplink reference signal (such as SRS signal in LoS / NLoS scenario) in N general scenarios can be represented as​ and , and the division manner is not limited. Thus, when the MAML algorithm is used to optimize the model initialization parameters, the optimal initial parameters of the model are represented as , and the optimization target of the parameter is specifically calculated as follows:

[0070]

[0071] wherein, represents the first loss function of the model training in the i-th scene, and the specific calculation manner is not limited, for example, it can be set as a cross-entropy loss function, etc. represents the model parameter obtained by training the reference signal (such as the SRS signal in the LoS / NLoS scene) on the first sample collected in the i-th scene.

[0072] It should be noted that the specific component structure of the initial scene recognition model is not limited in the present application, and can be set according to actual conditions and experience values. In some embodiments, the initial scene recognition model can be set as a convolutional neural network (CNN) model, etc.

[0073] It can be understood that MAML is a model training method, which is effective for all models. The present application takes the CNN model as an example of the initial scene recognition model, and then uses the MAML algorithm to train the optimal initialization parameters of the CNN model. In the training process, the input of the model is the CIR data of the first sample uplink reference signal collected, and the output of the model is a binary classification result, i.e., the scene recognition result. In the training, internal task updating is needed first. For the i-th scene, the model parameter training is performed, the training set collected is used , the gradient descent method is used, the model parameter is updated according to the above formula (2), and the recognition model for the current scene is obtained, i.e., . The model training is performed according to the above steps for all N scenes, and the respective recognition models are obtained. After the internal task updating of the MAML algorithm is completed, the overall loss of the external meta-optimization target is calculated in combination with the scene test data set and the above formula (5), and then the gradient direction of the meta-parameter is calculated by using the above formula (4), and the parameter updating is performed. The optimal initial parameters of the CNN model are optimized according to the above process, and the training is stopped when the training number reaches the maximum set training number (such as 1000 times) or the parameter When the change value of the scene recognition model is less than a set threshold (such as 0.01), the training is stopped, and at this time, the optimal initialization parameter of the scene recognition model is obtained . That is, the optimized scene recognition model is obtained, and the initialization parameter obtained by the training can be highly adapted to a new task (such as a channel measurement mode classification task) under a small amount of training samples, and task migration under a small amount of training samples is realized.

[0074] Step C: Obtain a second sample uplink reference signal; and extract sample channel impulse response information corresponding to the second sample uplink reference signal.

[0075] Step D: Use the sample channel impulse response information and the second loss function to train the optimized scene recognition model for the channel measurement mode classification task, and obtain a channel measurement mode recognition model.

[0076] In this embodiment, after obtaining the optimized scene recognition model through step B, it can be deployed on the gNB side, and the gNB can complete the selection and switching of the measurement mode by itself. In order to be able to adaptively select a suitable channel measurement mode, on the basis of the optimized scene recognition model, further obtain an uplink reference signal such as an SRS signal sent by a PRU or a UE to a gNB as a second sample uplink reference signal. As shown in Figure 3 , the existing or future signal feature extraction method is used again, the characteristics of the wireless channel where the second sample uplink reference signal (such as SRS) is located are measured to obtain channel impulse response information, and the channel impulse response information is defined as sample channel impulse response information. Then, the optimized scene recognition model can be trained for the channel measurement mode classification task using the sample channel impulse response information and the second loss function, and a channel measurement mode recognition model is obtained.

[0077] Specifically, as shown in Figure 3 , one optional implementation is that the specific implementation process of using the sample channel impulse response information and the second loss function to train the optimized scene recognition model for the channel measurement mode classification task to obtain the channel measurement mode recognition model can include the following steps S301-S307:

[0078] S301: The gNB obtains a second sample uplink reference signal (such as SRS) sent by a UE or a PRU.

[0079] Further, the gNB can use the existing or future signal feature extraction method to measure the characteristics of the wireless channel where the second sample uplink reference signal (such as SRS) is located to obtain sample channel impulse response information corresponding to the second sample uplink reference signal.

[0080] S302: The gNB measures the channel environment where the second sample uplink reference signal is located by using the first channel measurement method (such as the Path-based measurement method) and the second channel measurement method (such as the Sample-based measurement method), respectively, to obtain the first channel measurement result and the second channel measurement result.

[0081] S303: The gNB sends the first channel measurement result and the second channel measurement result to the LMF network element.

[0082] S304: The LMF network element acquires the physical position information sent by the PRU.

[0083] The specific content and format of the physical position information sent by the PRU are not limited, and are only used to represent the actual positioning position information of the terminal device.

[0084] S305: The LMF network element uses the positioning calculation model to locate the UE to obtain the first positioning result and the second positioning result, and combines the physical position information of the PRU to calculate the positioning errors corresponding to the two channel measurement methods, so as to determine the channel measurement method for labeling the sample channel impulse response information.

[0085] After the LMF network element acquires the first channel measurement result and the second channel measurement result sent by the gNB and the physical position information sent by the PRU (herein referred to as ), the difference between the first positioning result and the physical position information of the PRU is calculated to obtain the positioning error corresponding to the first channel measurement method, and the difference between the second positioning result and the physical position information of the PRU is calculated to obtain the positioning error corresponding to the second channel measurement method.

[0086] Specifically, assuming that the first channel measurement method is the Path-based measurement method, the second channel measurement method is the Sample-based measurement method, and the LMF uses the CNN model as the positioning calculation model, since the data dimensions of the first channel measurement result and the second channel measurement result measured by using the Path-based measurement method and the Sample-based measurement method are different, two CNN models are pre-deployed at the LMF as positioning calculation models for positioning calculation, such as Figure 4are shown, and the two models are represented as M-s and M-p, respectively. Among them, the input of M-s is the second channel measurement result (including CIR, measurement time, etc.) corresponding to the Sample-based measurement mode, and the input of M-p is the first channel measurement result (including CIR, measurement time, etc.) corresponding to the Path-based measurement mode; the output results of M-s and M-p are both the current physical position information of the UE. In this way, after the LMF network element obtains the first channel measurement result and the second channel measurement result sent by the gNB, it can first calculate the UE position by using the corresponding positioning calculation model, that is, the position information of the UE can be obtained, and is defined as the first positioning result and the second positioning result, respectively, and the difference between the two positioning results can be used to determine the positioning error of the first channel measurement mode (i.e. the Path-based measurement mode) and the second channel measurement mode (i.e. the Sample-based measurement mode). and represent the first positioning result and the second positioning result, respectively.

[0087] Then, based on the physical position information (x, y) sent by the PRU, , the positioning error of the first channel measurement mode (i.e. the Path-based measurement mode) can be determined according to the first positioning result , and is defined as ; and the positioning error of the second channel measurement mode (i.e. the Sample-based measurement mode) can be determined according to the second positioning result , and is defined as .

[0088] Next, the difference (herein defined as d) between the positioning error of the first channel measurement mode (i.e. the Path-based measurement mode) and the positioning error of the second channel measurement mode (i.e. the Sample-based measurement mode) can be calculated, that is, , and the difference d is used as the positioning comparison result of the first channel measurement mode (i.e. the Path-based measurement mode) and the second channel measurement mode (i.e. the Sample-based measurement mode).

[0089] Further, it can be judged whether the value of the positioning comparison result d is not greater than a preset threshold (herein defined as , and the specific value of is not limited and can be set according to actual situation and experience value, for example, when the requirement for positioning accuracy is relatively high, the value of can be set as 0.1m, or for example, when the requirement for positioning accuracy is not high but the requirement for signaling overhead is high, the value of can be set as 0.5m, etc.); if not, that is, , it indicates that the positioning error of the first channel measurement mode (i.e. the Path-based measurement mode) Within acceptable limits, the channel measurement method of the sample channel impulse response information can be labeled as a path-based measurement method, and this can be used as the labeling result of the channel measurement method of the sample channel impulse response information; conversely, if it is greater than that, i.e. If this is not true, it indicates the positioning error corresponding to the first channel measurement method (i.e., the path-based measurement method). If the signal is too large, a second channel measurement method (i.e., sample-based measurement method) is needed for positioning. In this case, the channel measurement method of the sample channel impulse response information can be labeled as the sample-based measurement method, which is used as the labeling result of the channel measurement method of the sample channel impulse response information.

[0090] S306: The LMF sends a channel measurement method to the gNB to label the sample channel impulse response information (in practical applications, the channel measurement method can be represented by a label, such as s for sample-based measurement method and p for path-based measurement method).

[0091] S307: gNB uses the label of the channel measurement method (such as s or p) to label the CIR, forming training data, which is used to train the optimized scene recognition model to obtain the trained channel measurement method recognition model.

[0092] Specifically, gNB can first input sample channel impulse response information (CIR) into the optimized scene recognition model to classify channel measurement methods, obtaining the classification results of the channel measurement methods corresponding to the sample channel impulse response information. Then, using the channel measurement method labeling results (i.e., s or p) of the sample channel impulse response information and the classification results of the channel measurement methods corresponding to the sample channel impulse response information, the value of the second loss function is calculated, and the optimized scene recognition model is trained using the value and gradient descent method until a preset condition is met (such as reaching the maximum number of training iterations), at which point the update of the model parameters is stopped, and the trained channel measurement method recognition model is obtained.

[0093] During training, the second sample uplink reference signal (specifically, its corresponding sample channel impulse response information) can be represented as the training dataset D. The model's task is to classify the channel measurement methods in the current signal transmission environment and select the appropriate channel measurement method. Therefore, the specific calculation formula for model training is as follows:

[0094]

[0095] in, This represents the second loss function used during model training. The specific content is not limited, but in this embodiment, the cross-entropy loss function is used.

[0096] In the model training, the optimized scene recognition model can still be a CNN model, the input of the model is the sample channel impulse response information, and the output of the model is a binary classification result, which is the identification result of the gNB to the sample environment, that is, the channel measurement mode expected to be taken by the gNB. In the training process of the model, the second loss function used is the cross-entropy loss function. In the binary classification problem, the class label is generally marked as 0 / 1 two classes, so the specific calculation formula of the cross-entropy loss function of the binary classification problem is as follows:

[0097]

[0098] Where m represents the number of model training samples; represents the true label of the i-th sample; represents the probability that the i-th sample belongs to the first class.

[0099] Then, the model training loss is calculated according to the above formula, and the gradient descent method is used to further update the model parameters , and the specific update formula is as follows:

[0100]

[0101] In this way, when the maximum set training times (such as 1000 times) are reached or the change value of the parameter is less than the set threshold (such as 0.01), the model training is stopped, and at this time the trained channel measurement mode recognition model is obtained.

[0102] On this basis, after inputting the target channel impulse response information into the trained channel measurement mode recognition model, the target channel measurement mode (that is, the Path-based measurement mode or the Sample-based measurement mode) that can guarantee positioning accuracy while reducing signaling overhead can be identified, so as to execute the subsequent step S203.

[0103] For ease of understanding, the criteria followed by the channel measurement mode recognition model for selecting the target channel measurement mode are represented as the contents of Table 1 as follows:

[0104]

[0105] Table 1

[0106] S203: The gNB judges whether the target channel measurement mode is the same as the channel measurement mode issued by the LMF network element.

[0107] In this embodiment, after the gNB identifies the target channel measurement mode (i.e., the Path-based measurement mode or the Sample-based measurement mode) that can guarantee the positioning accuracy while reducing the signaling overhead through step S202, the gNB can further compare the target channel measurement mode with the default measurement mode issued by the LMF. If they are consistent, the gNB continues to perform the subsequent step S204. If they are inconsistent, the gNB continues to perform the subsequent step S205. The specific content of the default measurement mode issued by the LMF is not limited and can be pre-set according to actual conditions and experience values.

[0108] S204: The gNB measures the current channel environment by using the target channel measurement mode, obtains a channel measurement result, and sends the channel measurement result to the LMF network element.

[0109] S205: The gNB sends a switching request of the channel measurement mode to the LMF network element.

[0110] S206: The LMF network element responds to the switching request sent by the gNB and issues a more suitable channel measurement mode (which can be the same as the target channel measurement mode) to the gNB.

[0111] S207: The gNB measures the current channel environment by using the measurement mode fed back by the LMF network element in response to the switching request, obtains a channel measurement result, and sends the channel measurement result to the LMF network element.

[0112] S208: The LMF selects a corresponding positioning calculation model to position the target terminal according to the channel measurement result sent by the gNB through step S204 or step S207, and obtains a positioning result.

[0113] In this way, after the gNB obtains the target uplink reference signal sent by the UE or the PRU and extracts the corresponding target channel impulse response information, the gNB can adaptively select a more suitable target channel measurement mode by using the pre-constructed channel measurement mode identification model through the processing procedure of steps S202-S208. That is, when the positioning accuracy can meet the requirements, the gNB selects a channel measurement mode (such as the Path-based measurement mode) that can reduce the signaling overhead. When the positioning accuracy is low (i.e., cannot meet the requirements), the gNB selects a channel measurement mode (such as the Sample-based measurement mode) that can improve the positioning accuracy. Therefore, the gNB can guarantee the positioning accuracy while reducing the signaling overhead, so as to achieve an effective balance between the positioning accuracy improvement and the network resource consumption.

[0114] It should be noted that the target terminal mentioned in the embodiments of the present application can be, but is not limited to, a mobile phone, a tablet computer, a personal digital assistant (PDA), a desktop, laptop, notebook computer, an ultra-mobile personal computer (UMPC), a handheld computer, a netbook, a wearable device, and the like terminal device (which can also be referred to as an electronic device).

[0115] In order to make the person skilled in the art more clearly understand the terminal positioning method provided by the present application, the hardware architecture and software system architecture of the target terminal are described in detail below.

[0116] Referring to Figure 5 , a schematic diagram of a terminal device provided by an embodiment of the present application is shown.

[0117] As Figure 5 indicated, the terminal device 500 can include a processor 510, a mobile communication module 520, a wireless communication module 530, a sensor module 540, a display screen 550, an internal memory 560, a camera 570, an audio module 580, a loudspeaker 580A, a receiver 580B, a microphone 580C, a headset interface 580D, an antenna group 1, and an antenna group 2.

[0118] It can be understood that the structure shown in the embodiments of the present application does not constitute a specific limitation on the terminal device 500. In other embodiments of the present application, the terminal device 500 can include more or fewer components than shown, or combine certain components, or split certain components, or different component arrangements. The components shown can be implemented in hardware, software, or a combination of software and hardware.

[0119] The processor 510 can include one or more processing units, for example: the processor 510 can include an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), etc. Different processing units can be independent devices, or can be integrated in one or more processors. The controller can generate operation control signals according to instruction operation codes and timing signals, and complete the control of fetching instructions and executing instructions.

[0120] The processor 510 can also include a memory for storing instructions and data. In some embodiments, the memory in the processor 510 is a cache memory. The memory can hold instructions or data that the processor 510 has just used or is using repeatedly. If the processor 510 needs to use the instructions or data again, it can call them directly from the memory. This avoids repeated access and reduces the waiting time of the processor 510, thus improving the efficiency of the system.

[0121] In some embodiments, the processor 510 can include one or more interfaces. The interfaces can include an inter-integrated circuit (I2C) interface, an inter-integrated circuit sound (I2S) interface, a pulse code modulation (PCM) interface, a universal asynchronous receiver / transmitter (UART) interface, a mobile industry processor interface (MIPI), a general-purpose input / output (GPIO) interface, a subscriber identity module (SIM) interface, and / or a universal serial bus (USB) interface, etc.

[0122] The sensor module 540 can be used to obtain data signals related to various aspects of the terminal device 500, which can be used as a basis for implementing corresponding functions. In some embodiments, the sensor module 540 can include, but is not limited to, an image sensor, a gyroscope sensor, a barometric pressure sensor, a sound pressure sensor, an acceleration sensor, a temperature sensor, a pressure sensor, etc.

[0123] The display screen 550 is configured to display images, videos, and the like. The display screen 550 includes a display panel. The display panel can be a liquid crystal display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light-emitting diode (AMOLED), a flex light-emitting diode (FLED), a Miniled, a MicroLed, a Micro-oLed, a quantum dot light emitting diodes (QLED), or the like. In some embodiments, the terminal device 500 can include one or P display screens 550, where P is a positive integer greater than 1.

[0124] The internal memory 560 can be configured to store computer-executable program code including instructions. The internal memory 560 can include a program storage area and a data storage area. The program storage area can store an operating system, at least one application program required by a function (such as a sound collection function, an image shooting function, and the like), and the like. The data storage area can store data (such as audio data, image data, and the like) created during use of the terminal device 500, and the like. In addition, the internal memory 560 can include a high-speed random access memory, and can further include a non-volatile memory such as at least one magnetic disk storage device, a flash memory device, a universal flash storage (UFS), and the like. The processor 510 executes various function applications and data processing of the terminal device 500 by running instructions stored in the internal memory 560 and / or instructions stored in a memory disposed in the processor.

[0125] The camera 570 is configured to capture still images or videos. For example, a user can hold the terminal device 500 behind his / her head, and use the camera 570 installed in the terminal device 500 to shoot images or videos, and the like. In some embodiments, the terminal device 500 can include one or K cameras 570, where K is a positive integer greater than 1.

[0126] The terminal device 500 implements a display function through a GPU, the display screen 550, an application processor, and the like. The GPU is a microprocessor for image processing, and is connected to the display screen 550 and the application processor. The GPU is configured to perform mathematical and geometric calculations for graphics rendering. The processor 510 can include one or more GPUs that execute program instructions to generate or change display information.

[0127] The terminal device 500 can realize audio functions such as music playing, voice input and output, etc. through the audio module 580, the speaker 580A, the receiver 580B, the microphone 580C, the earphone interface 580D, and the application processor, etc.

[0128] The audio module 580 is configured to convert digital audio information into an analog audio signal output, and is also configured to convert an analog audio input into a digital audio signal. The audio module 580 can also be configured to encode and decode audio signals. In some embodiments, the audio module 580 can be disposed in the processor 510, or some functional modules of the audio module 580 can be disposed in the processor 510.

[0129] The speaker 580A, also referred to as a “loudspeaker” or a “micro speaker”, etc. mentioned in the present application, is configured to convert an audio electrical signal into a sound signal for audio playing. The terminal device 500 can listen to music or listen to a hands-free call through the speaker 580A.

[0130] The receiver 580B, also referred to as a “earpiece”, is configured to convert an audio electrical signal into a sound signal. When the terminal device 500 answers a call or a voice message, the receiver 580B can be held close to a human ear to listen to the voice.

[0131] The microphone 580C, also referred to as a “microphone” or a “sound transducer”, is configured to convert a sound signal into an electrical signal. When making a call or sending a voice message, a user can speak into the microphone 580C close to the human mouth to input the sound signal into the microphone 580C. The terminal device 500 can be provided with at least one microphone 580C. In other embodiments, the terminal device 500 can be provided with two microphones 580C, which can not only collect sound signals, but also realize noise reduction functions. In other embodiments, the terminal device 500 can be provided with three, four or more microphones 580C, which can not only collect sound signals and reduce noise, but also identify the source of the sound, realize directional recording functions, etc.

[0132] The earphone interface 580D is configured to connect a wired earphone, and does not limit the standard properties of the interface.

[0133] It can be understood that the interface connection relationship between the modules shown in the embodiments of the present application is only illustrative, and does not constitute a structural limitation on the terminal device 500.

[0134] The wireless communication function of the terminal device 500 can be realized through the antenna 1, the antenna 2, the mobile communication module 520, the wireless communication module 530, the modem processor, and the baseband processor, etc.

[0135] Antennas 1 and 2 are used for transmitting and receiving electromagnetic wave signals. Each antenna in terminal device 500 can be used to cover a single or multiple communication frequency bands. Different antennas can also be multiplexed to improve the utilization of antennas. For example: antenna 1 can be multiplexed as a diversity antenna for wireless local area networks. In some other embodiments, antennas can be used in combination with tuning switches.

[0136] Mobile communication module 520 can provide solutions for wireless communication including 2G / 3G / 4G / 5G, etc. applied on terminal device 500. Mobile communication module 520 can include at least one filter, switch, power amplifier, low noise amplifier (LNA), etc. Mobile communication module 520 can receive electromagnetic waves by antenna 1, and perform filtering, amplification, etc. on the received electromagnetic waves, and transmit the processed signals to the modem processor for demodulation. Mobile communication module 520 can also amplify the signals modulated by the modem processor, and convert the signals into electromagnetic waves radiated by antenna 1. In some embodiments, at least part of the functional modules of mobile communication module 520 can be arranged in processor 510. In some embodiments, at least part of the functional modules of mobile communication module 520 and at least part of the modules of processor 510 can be arranged in the same device.

[0137] Wireless communication module 530 can provide solutions for wireless communication including wireless local area networks (WLAN) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), infrared (IR) technology, etc. applied on terminal device 500. Wireless communication module 530 can be one or more devices integrated with at least one communication processing module. Wireless communication module 530 receives electromagnetic waves via antenna 2, performs frequency modulation and filtering processing on the electromagnetic wave signals, and transmits the processed signals to processor 510. Wireless communication module 530 can also receive signals to be transmitted from processor 510, perform frequency modulation and amplification on the signals, and convert the signals into electromagnetic waves radiated by antenna 2.

[0138] In addition, on top of the above-mentioned components, terminal device 500 runs an operating system. For example, iOS operating system, Android operating system, Windows operating system, etc. Application programs can be installed and run on the operating system.

[0139] Referring to Figure 6 which shows a software structure diagram of the terminal device provided in the embodiments of the present application.

[0140] The software system of the terminal device 500 can adopt a layered architecture, an event-driven architecture, a microkernel architecture, a microservice architecture, or a cloud architecture. The embodiments of the present application take the Android system with a layered architecture as an example to exemplarily illustrate the software structure of the terminal device 500.

[0141] The layered architecture divides the software into several layers, each of which has a clear role and division of labor. The layers communicate with each other through software interfaces. In some embodiments, the Android system is divided into four layers, from top to bottom, which are the application layer, the application framework layer, the Android runtime and system library, and the kernel layer.

[0142] The application layer can include a series of application packages. As shown in Figure 6 , the application packages can include map, music, navigation, camera, calendar, WLAN, Bluetooth, and the like.

[0143] The application framework layer provides the application programming interface (API) and programming framework for the applications of the application layer. The application framework layer includes some pre-defined functions. As shown in Figure 6 , the application framework layer can include the window manager, the notification manager, the phone manager, the resource manager, the message manager, and the like.

[0144] The window manager is used to manage the window program. The window manager can obtain the size of the display screen, determine whether there is a status bar, lock the screen, and intercept the screen, etc.

[0145] The phone manager is used to provide the communication function of the electronic device 500. For example, the management of the call state (including connection, hang-up, etc.).

[0146] The message manager is used to provide the message sending and receiving function of the electronic device 500. For example, the management of the message sending state (including fallback retransmission, non-retransmission, etc.).

[0147] The resource manager provides various resources for the applications, such as localized strings, icons, pictures, layout files, video files, and the like.

[0148] The notification manager enables applications to display notification information in the status bar, which can be used to communicate alert-type messages that can automatically disappear after a brief stay without user interaction. For example, the notification manager is used to notify download completion, message reminders, etc. The notification manager can also be a notification that appears in the form of a chart or a scroll bar text in the system top status bar, such as a notification of a background running application, and can also be a notification that appears in the form of a dialog window on the screen. For example, the notification manager can prompt text information in the status bar, issue a prompt sound, vibrate the electronic device, and flash the indicator light, etc.

[0149] The Android Runtime includes a core library and a virtual machine. The Android runtime is responsible for scheduling and managing the Android system.

[0150] The core library includes two parts: one part is a function function that the java language needs to call, and the other part is the core library of Android.

[0151] The application layer and the application framework layer run in the virtual machine. The virtual machine executes the java files of the application layer and the application framework layer into binary files. The virtual machine is used to perform the management of the object life cycle, the management of the stack, the management of the thread, the management of the security and the exception, and the garbage collection, etc.

[0152] The system library can include multiple functional modules. For example: a surface manager, media libraries, a three-dimensional graphics processing library (for example: OpenGL ES), a 2D graphics engine (for example: SGL), etc.

[0153] The surface manager is used to manage the display subsystem, and provides a fusion of 2D and 3D layers for multiple applications.

[0154] The media library supports multiple commonly used audio, video format playback and recording, and static image files, etc. The media library can support multiple audio and video encoding formats, such as: MPEG4, H.264, MP3, AAC, AMR, JPG, PNG, etc.

[0155] The three-dimensional graphics processing library is used to realize three-dimensional graphics drawing, image rendering, synthesis, and layer processing, etc.

[0156] The 2D graphics engine is a drawing engine for 2D drawing.

[0157] The kernel layer is a layer between hardware and software. The kernel layer at least includes a display driver, a camera driver, an audio driver, and a sensor driver.

[0158] In some other embodiments, the embodiments of the present application also provide another terminal positioning method. The method is used for positioning a terminal in a system as shown inFigure 7 The network signaling overhead of the Case 2b positioning mode is large. Figure 7 As shown in the Case 2b positioning mode, when the positioning calculation model is deployed on the LMF side, the channel measurement is completed by the UE (completed by the gNB in Case 3b), and the specific positioning process includes the following steps S701-S704:

[0159] S701: The UE acquires the downlink reference signal sent by the gNB.

[0160] The specific content of the downlink reference signal is not limited, which can be but is not limited to a base station signal such as a positioning reference signal (PRS).

[0161] S702: The UE performs channel measurement on the downlink reference signal to obtain channel measurement results (Measurement results).

[0162] The specific content of the channel measurement results is not limited, which can include but is not limited to CIR, transmission time, measurement parameters, and measurement methods.

[0163] S703: The UE sends the channel measurement results (Measurement results) to the LMF.

[0164] S704: The LMF performs positioning on the UE according to the channel measurement results (Measurement results) by using the positioning calculation model.

[0165] It can be seen that in the Case 2b positioning mode, since the channel measurement is completed by the UE, the UE needs to send a large amount of channel measurement data (Measurement results) to the LMF, which greatly increases the network signaling overhead and may cause excessive consumption of core network transmission resources, and even more likely to impact existing radio resource management. Therefore, the embodiment of the present application also provides a terminal positioning method as shown in Figure 8 The terminal positioning method is applied to the UE to achieve effective balance between positioning accuracy and network resource consumption in the Case 2b positioning mode. The pre-constructed channel measurement method recognition model is deployed to the UE, and the specific implementation process of the terminal positioning method can include the following steps S801-S808:

[0166] S801: The UE acquires the target downlink reference signal (such as PRS) sent by the gNB.

[0167] S802: The UE inputs the target channel impulse response information corresponding to the target downlink reference signal into the pre-constructed channel measurement method identification model to obtain a target channel measurement method.

[0168] S803: The UE determines whether the target channel measurement method is the same as the channel measurement method issued by the LMF network element, and if so, continues to perform the subsequent step S804, otherwise, if not, continues to perform the subsequent step S805.

[0169] S804: The UE measures the current channel environment using the target channel measurement method to obtain a channel measurement result, and sends the channel measurement result to the LMF network element.

[0170] S805: The UE sends a switching request for the channel measurement method to the LMF network element.

[0171] S806: The LMF network element issues a more suitable channel measurement method (which can be the same as the target channel measurement method) to the UE in response to the switching request sent by the UE.

[0172] S807: The UE measures the current channel environment using the measurement method fed back by the LMF network element in response to the switching request to obtain a channel measurement result, and sends the channel measurement result to the LMF network element.

[0173] S808: The LMF selects a corresponding positioning calculation model to position the target terminal according to the channel measurement result sent by the UE through step S804 or step S807 to obtain a positioning result.

[0174] In this way, after the UE obtains the target downlink reference signal sent by the gNB and extracts the target channel impulse response information corresponding thereto, it can adaptively select a more suitable target channel measurement method by performing the processing procedures of steps S802-S808 above, using the pre-constructed channel measurement method identification model, to obtain a channel measurement result for positioning the target terminal. That is, when the positioning accuracy can meet the requirements, a channel measurement method with reduced signaling overhead (such as the Path-based measurement method) is selected, and when the positioning accuracy is low (i.e., does not meet the requirements), a channel measurement method that improves the positioning accuracy (such as the Sample-based measurement method) is selected, so that the signaling overhead can be reduced while ensuring the positioning accuracy, thereby achieving an effective balance between positioning accuracy improvement and network resource consumption.

[0175] It should be noted that the target terminal mentioned in the embodiments of the present application can still be, but is not limited to, a mobile phone, a tablet computer, a personal digital assistant (PDA), a desktop, laptop, notebook computer, an ultra-mobile personal computer (UMPC), a handheld computer, a netbook, and a wearable device, etc. terminal device (also referred to as electronic device for short), and the specific structure diagram can still be as shown in Figure 5 , and the terminal positioning method provided by the embodiments can also be applied to the target terminal.

[0176] Among them, for the processor 510, after obtaining the target downlink reference signal sent by the gNB and extracting the corresponding target channel impulse response information by using the terminal device 500, a more suitable target channel measurement method is adaptively selected by using the channel measurement method identification model deployed in the terminal device 500, that is, when the positioning accuracy can meet the requirements, the channel measurement method with reduced signaling overhead (such as Path-based measurement method) is selected, and when the positioning accuracy is low (that is, it cannot meet the requirements), the channel measurement method with improved positioning accuracy (such as Sample-based measurement method) is selected for channel measurement to obtain the channel measurement result, which is used for the final positioning of the target terminal, so as to reduce the signaling overhead while ensuring the positioning accuracy.

[0177] In some embodiments, the internal memory 560 stores instructions for executing the terminal positioning method. The processor 510 can realize the functions by executing the instructions stored in the internal memory 560, and the specific functions are as follows: the terminal device 500 first obtains the target downlink reference signal sent by the gNB and extracts the corresponding target channel impulse response information. Then, the target channel impulse response information corresponding to the target downlink reference signal is input into the pre-constructed channel measurement method identification model to obtain the target channel measurement method. Then, it is judged whether the target channel measurement method is the same as the channel measurement method issued by the LMF network element, if the same, the current channel environment is measured by using the target channel measurement method to obtain the channel measurement result, and the channel measurement result is sent to the LMF network element, otherwise, if not the same, the switching request of the channel measurement method is sent to the LMF network element, and the current channel environment is measured by using the measurement method fed back by the LMF network element in response to the switching request to obtain the channel measurement result; then the channel measurement result is sent to the LMF network element, so that the LMF network element can position the target terminal according to the channel measurement result to obtain the positioning result, so as to reduce the signaling overhead while ensuring the positioning accuracy, so as to realize the effective balance between the positioning accuracy improvement and the network resource consumption.

[0178] In addition, on the basis of the above components, in order to realize the terminal positioning method applied to the UE as introduced in steps S801-S808, the terminal positioning algorithm is added in the software structure of the terminal device, such asFigure 9 As shown, the terminal positioning algorithm can be added in the application framework layer, so that when a user wants to position the terminal device 500, the user can click the map or the navigation app on the terminal device 500, and the click instruction is transmitted to the terminal positioning algorithm in the application framework layer (the framework layer) through the corresponding interface for subsequent information processing operations.

[0179] The terminal positioning algorithm is used to adaptively select a more suitable target channel measurement mode by using a channel measurement mode identification model, that is, when the positioning accuracy can meet the requirements, a channel measurement mode (such as a Path-based measurement mode) with reduced signaling overhead is selected, and when the positioning accuracy is low (that is, cannot meet the requirements), a channel measurement mode (such as a Sample-based measurement mode) with improved positioning accuracy is selected for channel measurement to obtain a channel measurement result, which is used for the final positioning of the target terminal, so that the signaling overhead can be reduced while the positioning accuracy is ensured, thereby achieving an effective balance between the positioning accuracy improvement and the network resource consumption.

[0180] In addition, the embodiments of the present application also provide an electronic device (that is, a terminal device), and the hardware structure and the software framework of the electronic device can be referred to the corresponding description of Figure 5 、 Figure 6 or Figure 9 . The electronic device includes a memory and a processor, the memory stores a computer program, and the processor is configured to invoke and execute the computer program to implement the terminal positioning method provided in the above description.

[0181] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program is run by the processor of the terminal device to implement the terminal positioning method provided in the above description.

[0182] The above description is only used to illustrate the technical solutions of the present application, and not to limit them; although the foregoing embodiments of the present application have been described in detail, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A terminal positioning method, characterized in that, The method includes: Acquire the target uplink reference signal; and extract the target channel impulse response information corresponding to the target uplink reference signal; The target channel impulse response information is input into a pre-constructed channel measurement mode identification model to obtain the target channel measurement mode; the channel measurement mode identification model is obtained by training based on the Model Independent Element Learning (MAML) algorithm using sample channel impulse response information pre-labeled with channel measurement modes. Determine whether the target channel measurement method is the same as the channel measurement method issued by the location management function (LMF) network element; If so, the current channel environment is measured using the target channel measurement method to obtain the channel measurement results; If not, a channel measurement mode switching request is sent to the LMF, and the current channel environment is measured using the measurement mode fed back by the LMF network element in response to the switching request, so as to obtain the channel measurement result; The channel measurement results are sent to the LMF network element so that the LMF network element can locate the target terminal based on the channel measurement results and obtain the location result.

2. The method according to claim 1, characterized in that, The channel measurement method identification model is constructed as follows: Obtain the first sample uplink reference signal in a general scenario; Based on the MAML algorithm, the initialization parameters of the initial scene recognition model are optimized using the uplink reference signal of the first sample and the first loss function to obtain the optimized scene recognition model. Acquire the second sample uplink reference signal; and extract the sample channel impulse response information corresponding to the second sample uplink reference signal; Using the sample channel impulse response information and the second loss function, the optimized scene recognition model is trained for the channel measurement method classification task to obtain the channel measurement method recognition model.

3. The method according to claim 2, characterized in that, The initial scene recognition model is a convolutional neural network (CNN) model.

4. The method according to claim 2, characterized in that, Both the first loss function and the second loss function are cross-entropy loss functions.

5. The method according to claim 2, characterized in that, The process of training the optimized scene recognition model for a channel measurement mode classification task using the sample channel impulse response information and the second loss function to obtain the channel measurement mode recognition model includes: Using the first channel measurement method and the second channel measurement method, the channel environment where the uplink reference signal of the second sample is located is measured, respectively, to obtain the first channel measurement result and the second channel measurement result; The first channel measurement result and the second channel measurement result are sent to the LMF network element so that the LMF network element can locate the terminal based on the first channel measurement result and the second channel measurement result, respectively, and obtain the first location result and the second location result. The positioning error corresponding to the first channel measurement method is determined based on the first positioning result; and the positioning error corresponding to the second channel measurement method is determined based on the second positioning result. Based on the positioning error corresponding to the first channel measurement method and the positioning error corresponding to the second channel measurement method, the channel measurement method is labeled on the sample channel impulse response information to obtain the channel measurement method labeling result of the sample channel impulse response information; The sample channel impulse response information is input into the optimized scene recognition model for classification of channel measurement methods, thereby obtaining the classification result of the channel measurement method corresponding to the sample channel impulse response information. Using the channel measurement method labeling results of the sample channel impulse response information and the classification results of the channel measurement methods corresponding to the sample channel impulse response information, the value of the second loss function is calculated, and the optimized scene recognition model is trained using the value and gradient descent method until a preset condition is met, at which point the update of the model parameters is stopped, and the trained channel measurement method recognition model is obtained.

6. The method according to claim 5, characterized in that, The second sample uplink reference signal is sent by the Positioning Reference Unit (PRU); the positioning error corresponding to the first channel measurement method is determined based on the first positioning result; And determining the positioning error corresponding to the second channel measurement method based on the second positioning result, including: Obtain the physical location information of the PRU; Based on the difference between the first positioning result and the physical location information of the PRU, the positioning error corresponding to the first channel measurement method is calculated; and based on the difference between the second positioning result and the physical location information of the PRU, the positioning error corresponding to the second channel measurement method is calculated.

7. The method according to claim 5, characterized in that, The first channel measurement method is a path-based channel measurement method; the second channel measurement method is a sample-based channel measurement method; the step of labeling the sample channel impulse response information with the channel measurement method based on the positioning error corresponding to the first channel measurement method and the positioning error corresponding to the second channel measurement method, to obtain the channel measurement method labeling result of the sample channel impulse response information, includes: The difference between the positioning error corresponding to the first channel measurement method and the positioning error corresponding to the second channel measurement method is calculated and used as the positioning comparison result between the first channel measurement method and the second channel measurement method. Determine whether the value of the positioning comparison result is not greater than a preset threshold; If so, the channel measurement method of the sample channel impulse response information is labeled as a path-based measurement method, and this is used as the labeling result of the channel measurement method of the sample channel impulse response information; If not, the channel measurement method of the sample channel impulse response information is labeled as a sample-based measurement method, and this is used as the labeling result of the channel measurement method of the sample channel impulse response information.

8. The method according to claim 7, characterized in that, The preset threshold value is either 0.1 meters or 0.5 meters.

9. The method according to any one of claims 1-8, characterized in that, The target channel measurement method is either a path-based measurement method or a sample-based measurement method.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor being used to invoke and execute the computer program to implement the method of any one of claims 1-9.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by an electronic device, implements the method of any one of claims 1-9.

Images