Network node and method performed thereby, electronic device, and storage medium

Abstract

Embodiments of the present disclosure provide a network node and a method performed by the network node, an electronic device and a storage medium, relating to the field of artificial intelligence. The method comprises: obtaining a channel impulse response of a non-line-of-sight channel between a target object and a signal receiving device; determining a plurality of signal paths of the non-line-of-sight channel based on the channel impulse response; determining a plurality of clusters based on the plurality of signal paths; selecting at least one cluster from the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters; and positioning the target object based on the at least one cluster. Optionally, the above-mentioned method performed by the electronic device can be performed using an artificial intelligence model.

Description

Technical Field

[0001] This disclosure relates to the field of communications, and more specifically to a network node, a method performed by the network node, an electronic device, and a computer-readable storage medium. Background Technology

[0002] With the continuous advancement of 5G mobile communication network construction and the advent of the Internet of Things (IoT) era, customers' demands for wireless communication capabilities are increasing. Among these demands, accurate location tracking of target objects is becoming a fundamental capability and essential requirement for various application industries and equipment functions. Application scenarios such as emergency rescue, connected vehicles, intelligent manufacturing, and smart logistics are placing ever higher expectations on positioning capabilities. Furthermore, positioning capability has become one of the core capabilities of 5G, and the standards for positioning accuracy are constantly improving to meet the needs of specific vertical industries in various application scenarios.

[0003] However, most of the currently widely used target object localization methods are only applicable to the localization of target objects in line-of-sight (LOS) scenes, and cannot be well applied to the localization of target objects in non-line-of-sight (NLOS) scenes. Research on localization technology for NLOS scenes is still insufficient, and the limited localization technologies currently available for NLOS scenes suffer from high localization complexity or low localization accuracy. Therefore, there is a need for localization technologies that reduce localization complexity and improve localization accuracy in NLOS scenes. Summary of the Invention

[0004] According to a first aspect of the present disclosure, a method performed by a network node is provided, comprising: acquiring a channel impulse response of a non-line-of-sight channel between a target object and a signal receiving device; determining a plurality of signal paths of the non-line-of-sight channel based on the channel impulse response; determining a plurality of clusters based on the plurality of signal paths; selecting at least one cluster from the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters; and locating the target object based on the at least one cluster.

[0005] Optionally, selecting at least one cluster from the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters includes: obtaining power distribution information and / or angle distribution information corresponding to the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters, wherein the power distribution information is used to represent the spatial distribution of power corresponding to the plurality of clusters, and the angle distribution information is used to represent the spatial distribution of angle of arrival corresponding to the plurality of clusters; and selecting the at least one cluster from the plurality of clusters based on the power distribution information and / or the angle distribution information.

[0006] Optionally, based on the power information and / or angle of arrival information corresponding to the plurality of clusters, obtaining power distribution information and / or angle distribution information corresponding to the plurality of clusters includes: constructing a spatial grid corresponding to the target object based on the angle of arrival information, wherein the angle of arrival information includes a horizontal angle of arrival and a vertical angle of arrival, and the coordinates of the horizontal and vertical directions of the spatial grid are the horizontal angle of arrival and the vertical angle of arrival, respectively; mapping the angle of arrival information to the spatial grid to obtain the angle distribution information, and / or mapping the power information to the spatial grid to obtain the power distribution information.

[0007] Optionally, constructing a spatial grid corresponding to the target object based on the angle of arrival information includes: estimating the horizontal and vertical angles of arrival corresponding to the strongest signal path in each of the plurality of clusters; determining the coordinate range of the spatial grid in the horizontal and vertical directions based on the maximum and minimum values ​​of the horizontal and vertical angles of arrival corresponding to the strongest signal path in each of the plurality of clusters; and determining the resolution of the spatial grid in the horizontal and vertical directions based on the coordinate range and the number of the plurality of clusters.

[0008] Optionally, mapping the power information to the spatial grid to obtain the power distribution information includes: mapping the power of the strongest signal path in each of the plurality of clusters to the spatial grid to obtain the power distribution information.

[0009] Optionally, selecting at least one cluster from the plurality of clusters based on the power distribution information and / or the angle distribution information includes: performing convolution on the angle distribution information in the spatial grid using a first convolution kernel to obtain an angle distribution density; performing convolution on the power distribution information in the spatial grid using a second convolution kernel to obtain a power distribution density; and selecting at least one cluster from the plurality of clusters based on the angle distribution density and the power distribution density.

[0010] Optionally, performing convolution on the power distribution information in the spatial grid using the second convolution kernel to obtain the power distribution density includes: adjusting the second convolution kernel based on the power distribution information in the convolution region of the second convolution kernel to obtain an adjusted second convolution kernel; and performing convolution on the power distribution information in the convolution region using the adjusted second convolution kernel to obtain the power distribution density.

[0011] Optionally, selecting at least one cluster from the plurality of clusters based on the angular distribution density and the power distribution density includes: obtaining an angular power spectral density based on the angular distribution density and the power distribution density; and selecting at least one cluster from the plurality of clusters based on the angular power spectral density.

[0012] Optionally, selecting at least one cluster from the plurality of clusters based on the angular power spectral density includes: performing data mapping on the angular power spectral density to obtain mapping data within a preset interval; determining a first threshold based on the mapping data; selecting at least one convolutional region in the spatial grid based on the first threshold; and selecting the cluster with the strongest power from each of the at least one convolutional region.

[0013] Optionally, the sizes of the first convolution kernel and the second convolution kernel are determined based on the scene in which the target object is located.

[0014] Optionally, the power corresponding to the plurality of clusters includes the cross-correlation power corresponding to each signal path included in the plurality of clusters, wherein the cross-correlation power is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window.

[0015] Optionally, determining multiple clusters based on the multiple signal paths includes: determining the multiple clusters based on the power information and time delay information corresponding to each signal path in the multiple signal paths, wherein the power information includes the cross-correlation power corresponding to each signal path, and wherein the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window.

[0016] Optionally, determining the plurality of clusters based on the power information and time delay information of each of the plurality of signal paths includes: detecting the power peak in the time domain based on the changing trend of the cross-correlation power of each of the plurality of signal paths and the time delay information; and determining the plurality of clusters based on the power peak.

[0017] Optionally, determining multiple signal paths of the non-line-of-sight channel based on the channel impulse response includes: determining the cross-correlation power of each signal path in the signal path set of the non-line-of-sight channel based on the channel impulse response, wherein the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window; and removing at least one signal path from the signal path set of the non-line-of-sight channel based on the cross-correlation power to determine multiple signal paths.

[0018] Optionally, determining multiple signal paths by removing at least one signal path from the set of signal paths in the non-line-of-sight channel based on the cross-correlation power includes: determining a cross-correlation power threshold based on the cross-correlation power; removing at least one signal path from the set of signal paths in the non-line-of-sight channel based on the cross-correlation power threshold, and using the remaining signal path after removing the at least one signal path as the multiple signal paths.

[0019] Optionally, the signal receiving device includes multiple antennas, and the cross-correlation power includes the cross-correlation power of each signal path corresponding to each of the multiple antennas. Determining the cross-correlation power threshold based on the cross-correlation power includes: averaging the cross-correlation power of each signal path corresponding to the multiple antennas and arranging them from high to low to obtain an average cross-correlation power queue; and determining the average cross-correlation power corresponding to the start position of the effective path in the average cross-correlation power queue based on the predetermined time window and the cross-correlation power of each signal path corresponding to each antenna, as the cross-correlation power threshold.

[0020] Optionally, locating the target object based on the at least one cluster includes: selecting at least one first signal path from each of the at least one cluster based on the power of the signal path; obtaining a location result using an artificial intelligence (AI) model based on relevant information of the signal paths selected from each cluster; determining whether to output a location result based on the difference between the current location result and the previous location result obtained using the AI ​​model; if so, obtaining the location information of the target object based on the current location result; otherwise, selecting at least one second signal path from each cluster based on the power of the signal path, and performing the step of obtaining a location result using an AI model based on relevant information of the signal paths selected from each cluster; wherein the second signal path includes the first signal path.

[0021] Optionally, after selecting at least one second signal path from each cluster, a localization result is obtained using an AI model based on the relevant information of the signal paths selected from each cluster. This includes: performing a first attention operation between the feature data corresponding to the relevant information of the signal paths in different clusters to obtain a first attention result; performing a second attention operation between the feature data corresponding to the relevant information of the signal paths in the same cluster to obtain a second attention result; and obtaining a localization result based on the first attention result and the second attention result.

[0022] Optionally, the relevant information includes angle information, power information, and time delay information corresponding to the signal path, obtained based on the channel impulse response corresponding to the signal path.

[0023] Optionally, obtaining the location information of the target object based on the current location result includes: performing a validity verification on the current location result; if the current location result is valid, then obtaining the location information based on the current location result; if the current location result is invalid, then obtaining the location information based on the previous location result of the target object.

[0024] According to a second aspect of the present disclosure, a method performed by a network node is provided, comprising: acquiring the channel impulse response of a non-line-of-sight channel between a target object and a signal receiving device; determining the cross-correlation power of each signal path in a set of signal paths of the non-line-of-sight channel based on the channel impulse response, wherein the cross-correlation power of each signal path is determined based on cross-correlation operations between the channel impulse responses of each signal path at different times within a predetermined time window; removing at least one signal path from the set of signal paths of the non-line-of-sight channel based on the cross-correlation power to determine a plurality of signal paths; and locating the target object based on the plurality of signal paths.

[0025] Optionally, determining multiple signal paths by removing at least one signal path from the set of signal paths in the non-line-of-sight channel based on the cross-correlation power includes: determining a cross-correlation power threshold based on the cross-correlation power; removing at least one signal path from the set of signal paths in the non-line-of-sight channel based on the cross-correlation power threshold, and using the remaining signal path after removing the at least one signal path as the multiple signal paths.

[0026] Optionally, the signal receiving device includes multiple antennas, and the cross-correlation power includes the cross-correlation power of each signal path corresponding to each of the multiple antennas. Determining a cross-correlation power threshold based on the cross-correlation power includes: averaging the cross-correlation power of each signal path corresponding to the multiple antennas and arranging them from high to low to obtain an average cross-correlation power queue; and determining the average cross-correlation power corresponding to the start position of the effective path in the average cross-correlation power queue based on the predetermined time window and the cross-correlation power of each signal path corresponding to each antenna, as the cross-correlation power threshold.

[0027] According to a third aspect of the present disclosure, a method performed by a network node is provided, comprising: acquiring a channel impulse response of a non-line-of-sight channel between a target object and a signal receiving device; determining a plurality of signal paths of the non-line-of-sight channel based on the channel impulse response; determining a plurality of clusters based on the plurality of signal paths; selecting at least one first signal path from each of the plurality of clusters based on the power of the signal paths; obtaining a positioning result using an artificial intelligence (AI) model based on relevant information of the signal paths selected from each cluster; determining whether to output a positioning result based on the difference between the current positioning result and the previous positioning result obtained using the AI ​​model; if so, obtaining location information of the target object based on the current positioning result; otherwise, selecting at least one second signal path from each cluster based on the power of the signal paths, and performing the step of obtaining a positioning result using an AI model based on relevant information of the signal paths selected from each cluster; wherein the second signal path includes the first signal path.

[0028] Optionally, after selecting at least one second signal path from each cluster, a localization result is obtained using an AI model based on the relevant information of the signal paths selected from each cluster. This includes: performing a first attention operation between the feature data corresponding to the relevant information of the signal paths in different clusters to obtain a first attention result; performing a second attention operation between the feature data corresponding to the relevant information of the signal paths in the same cluster to obtain a second attention result; and obtaining a localization result based on the first attention result and the second attention result.

[0029] Optionally, the relevant information includes angle information, power information, and time delay information corresponding to the signal path, obtained based on the channel impulse response corresponding to the signal path.

[0030] According to a fourth aspect of the present disclosure, a network node is provided, comprising: a transceiver; and a processor coupled to the transceiver and configured to perform any of the methods described above.

[0031] According to a fifth aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; at least one transceiver and / or at least one memory coupled to the at least one processor, wherein the at least one processor is configured to perform any of the methods described above.

[0032] According to a sixth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program or instructions are stored, which, when executed by at least one processor, cause the at least one processor to perform any of the methods described above.

[0033] According to the technical solution provided in the embodiments of this disclosure, after obtaining the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device, determining multiple signal paths of the non-line-of-sight channel based on the channel impulse response, and determining multiple clusters based on the multiple signal paths, at least one cluster is selected from the multiple clusters based on the power information and / or the angle of arrival information corresponding to the multiple clusters, and the target object is located based on the at least one cluster. This allows for the selection of more valuable clusters, and the target object is located only based on valuable clusters, thereby reducing the positioning complexity and improving the positioning accuracy.

[0034] According to the technical solution provided in the embodiments of this disclosure, after obtaining the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device, the cross-correlation power of each signal path in the signal path set of the non-line-of-sight channel is determined based on the channel impulse response (wherein, the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window). Then, based on the cross-correlation power, at least one signal path is removed from the signal path set of the non-line-of-sight channel to determine multiple signal paths. The target object is then located based on the multiple signal paths. Since the cross-correlation operation in the cross-correlation power determination process can eliminate signal noise, the target object is located based on the remaining valuable signal paths after removing invalid signal paths based on the cross-correlation power, which not only reduces the positioning complexity but also helps to improve the positioning accuracy.

[0035] According to the technical solution of this disclosure, after obtaining the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device, determining multiple signal paths of the non-line-of-sight channel between the target object and the signal receiving device based on the channel impulse response, and determining multiple clusters based on the multiple signal paths, at least one first signal path is selected from each of the multiple clusters based on the power of the signal path; based on the relevant information of the signal path selected from each cluster, an artificial intelligence (AI) model is used to obtain a positioning result; based on the difference between the current positioning result and the positioning result obtained by the AI ​​model last time, it is determined whether to output the positioning result. If so, the location information of the target object is obtained based on the current positioning result; otherwise, at least one second signal path (including the first signal path) is selected from each cluster based on the power of the signal path, and the step of obtaining the positioning result using the AI ​​model based on the relevant information of the signal path selected from each cluster is executed. In this way, the scale of data input to the artificial intelligence model can be reduced, the positioning complexity can be reduced, and since the positioning result is obtained based on the feature data corresponding to the continuously increasing number of signal paths, the positioning accuracy of the target object can be improved.

[0036] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0037] The accompanying drawings, which are incorporated in and form part of this specification, illustrate exemplary embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0038] Figure 1 This is a schematic diagram illustrating an NLOS channel environment.

[0039] Figure 2 This is a flowchart illustrating a method performed by a network node according to an embodiment of the present disclosure.

[0040] Figure 3 This is a schematic diagram illustrating the path for removing invalid signals according to an embodiment of the present disclosure.

[0041] Figure 4 This is a schematic diagram illustrating cluster detection and cluster selection according to embodiments of the present disclosure.

[0042] Figure 5 This is a schematic diagram illustrating an example of cluster detection according to an embodiment of the present disclosure.

[0043] Figure 6 This is a schematic diagram illustrating an example of constructing a spatial grid according to an embodiment of the present disclosure.

[0044] Figure 7 This is a schematic diagram illustrating a selection cluster according to an embodiment of the present disclosure.

[0045] Figure 8 This is a schematic diagram illustrating examples of using and not using a scaling factor when selecting clusters according to embodiments of the present disclosure.

[0046] Figure 9 This is a schematic diagram illustrating examples of data preprocessing and no data preprocessing when calculating the threshold for cluster selection according to embodiments of the present disclosure.

[0047] Figure 10 This is a schematic diagram illustrating an example of cluster selection according to an embodiment of the present disclosure.

[0048] Figure 11 This is a schematic diagram illustrating the preliminary screening of special clusters according to an embodiment of the present disclosure.

[0049] Figure 12 This is a schematic diagram illustrating a non-progressive AI network according to an embodiment of the present disclosure.

[0050] Figure 13 This is a schematic diagram illustrating a progressive AI network according to an embodiment of the present disclosure.

[0051] Figure 14 This is a schematic diagram illustrating feature data extraction according to an embodiment of the present disclosure.

[0052] Figure 15 This is a schematic diagram illustrating the reflection loss of scatterers made of different materials.

[0053] Figure 16 This is a schematic diagram illustrating examples of inter-cluster correlation and intra-cluster correlation according to embodiments of the present disclosure.

[0054] Figure 17 This is a schematic diagram illustrating how progressive input, inter-cluster attention, and intra-cluster attention reduce positioning complexity and improve positioning accuracy according to embodiments of the present disclosure.

[0055] Figure 18 This is a schematic diagram illustrating the performance of inter-cluster attention and intra-cluster attention in an attention layer included in a progressive localization network according to an embodiment of the present disclosure.

[0056] Figure 19 This is a schematic diagram illustrating the structure of a progressive positioning network according to an embodiment of the present disclosure.

[0057] Figure 20 This is a schematic diagram illustrating the operation of a progressive positioning network according to an embodiment of the present disclosure.

[0058] Figure 21 This is a schematic diagram illustrating the validity verification of the positioning results performed according to an embodiment of the present disclosure.

[0059] Figure 22 This is an exemplary architecture diagram illustrating a method performed by a network node according to an embodiment of the present disclosure.

[0060] Figure 23 This is an exemplary flowchart illustrating a method performed by a network node according to an embodiment of the present disclosure.

[0061] Figure 24 This is a flowchart illustrating a method performed by a network node according to an embodiment of the present disclosure.

[0062] Figure 25 This is a flowchart illustrating a method performed by a network node according to an embodiment of the present disclosure.

[0063] Figure 26 This is a block diagram illustrating a network node according to an embodiment of the present disclosure.

[0064] Figure 27 This is a schematic diagram illustrating the structure of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0065] The following description, with reference to the accompanying drawings, is provided to aid in a thorough understanding of the various embodiments of this disclosure as defined by the claims and their equivalents. This description includes various specific details to aid understanding but should be considered exemplary only. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the various embodiments described herein without departing from the scope and spirit of this disclosure. Furthermore, for clarity and brevity, descriptions of well-known functions and structures may be omitted.

[0066] The terms and wording used in the following description and claims are not limited to their dictionary meanings, but are merely used by the inventors to enable a clear and consistent understanding of this disclosure. Therefore, it will be apparent to those skilled in the art that the following description of various embodiments of this disclosure is for illustrative purposes only and not for limiting the purpose of this disclosure as defined in the appended claims and their equivalents.

[0067] It should be understood that the singular forms of “a,” “an,” and “the” can also include plural references unless the context clearly indicates otherwise. Thus, for example, the reference to “component surface” includes referring to one or more such surfaces. When we say that an element is “connected” or “coupled” to another element, the element can be directly connected or coupled to the other element, or it can mean that the element and the other element are connected through an intermediate element. Furthermore, the use of “connected” or “coupled” herein can include wireless connections or wireless couplings.

[0068] The terms “comprising” or “may include” refer to the presence of a corresponding disclosed function, operation, or component that may be used in the various embodiments of this disclosure, rather than limiting the presence of one or more additional functions, operations, or features. Furthermore, the terms “comprising” or “having” may be interpreted as indicating certain characteristics, numbers, steps, operations, constituent elements, components, or combinations thereof, but should not be construed as excluding the possibility of the presence of one or more other characteristics, numbers, steps, operations, constituent elements, components, or combinations thereof.

[0069] The term "or" as used in the various embodiments of this disclosure includes any of the listed terms and all combinations thereof. For example, "A or B" may include A, may include B, or may include both A and B. When describing multiple (two or more) items, if the relationship between the multiple items is not explicitly defined, the multiple items may refer to one, more, or all of the multiple items. For example, the description "parameter A includes A1, A2, A3" can be implemented as parameter A includes A1 or A2 or A3, or it can be implemented as parameter A includes at least two of the three items A1, A2, and A3.

[0070] Unless otherwise defined, all terms used in this disclosure (including technical or scientific terms) have the same meaning as understood by one of those skilled in the art to which this disclosure pertains. Common terms as defined in dictionaries are to be interpreted as having a meaning consistent with the context in the relevant technical field and should not be interpreted ideally or overly formally unless expressly defined in this disclosure.

[0071] At least some of the functions of the device or electronic device provided in this disclosure embodiment can be implemented by an AI model, such as implementing at least one module of a plurality of modules of the device or electronic device by an AI model. AI-related functions can be executed by non-volatile memory, volatile memory, and a processor.

[0072] The processor may include one or more processors. In this case, the one or more processors may be general-purpose processors, such as central processing unit (CPU), application processor (AP), etc., or pure graphics processing unit, such as graphics processing unit (GPU), vision processing unit (VPU), and / or AI-specific processors, such as neural processing unit (NPU).

[0073] The one or more processors control the processing of input data based on predefined operating rules or artificial intelligence (AI) models stored in non-volatile and volatile memory. These predefined operating rules or AI models are provided through training or learning.

[0074] Here, "providing through learning" refers to obtaining predefined operating rules or an AI model with desired characteristics by applying a learning algorithm to multiple learning datasets. This learning can be performed within the device or electronic device itself, in which the AI ​​is executed according to the embodiment, and / or can be implemented via a separate server / system.

[0075] AI models can contain multiple neural network layers. Each layer has multiple weight values, and each layer performs neural network computations by calculating the input data of that layer (such as the computation results of the previous layer and / or the input data of the AI ​​model) and the multiple weight values ​​of the current layer. Examples of neural networks include, but are not limited to, convolutional neural networks (CNNs), deep neural networks (DNNs), recurrent neural networks (RNNs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), bidirectional recurrent deep neural networks (BRDNNs), generative adversarial networks (GANs), and deep Q-networks.

[0076] A learning algorithm is a method of training a predetermined target device (e.g., a robot) using multiple learning data sets to enable, allow, or control the target device to make determinations or predictions. Examples of such learning algorithms include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

[0077] According to this disclosure, at least one step of a method performed in an electronic device can be implemented using an artificial intelligence model. The processor of the electronic device can perform preprocessing operations on data to transform it into a form suitable for use as input to an artificial intelligence model. The artificial intelligence model can be obtained through training. Here, "obtained through training" means obtaining a predefined operating rule or artificial intelligence model configured to perform desired features (or objectives) by training a basic artificial intelligence model with multiple training data using a training algorithm.

[0078] The following description of several optional embodiments illustrates the technical solutions of this disclosure and the technical effects produced by these solutions. It should be noted that the following embodiments can be referenced, learned from, or combined with each other. Identical terms, similar features, and similar implementation steps in different embodiments will not be repeated.

[0079] In a Line-of-Sight (LOS) scenario, the wireless signal propagates in a straight line between the transmitter and receiver without obstruction; that is, the signal path between the transmitter and receiver is a direct path. Since the direct path has the strongest signal power and the shortest transmission distance, it most directly describes the spatial relationship between the transmitter and receiver. However, in reality, due to the characteristics of wireless communication, especially in densely populated urban environments, spatial obstruction is almost inevitable, making Line-of-Sight (NLOS) scenarios more common. In NLOS scenarios, the wireless signal does not propagate in a straight line between the transmitter and receiver without obstruction. Instead, the signal received by the receiver is often a multipath signal superimposed after reflection, diffraction, and scattering through obstacles. Multipath channels bring a series of problems such as asynchronous time delays, signal attenuation, and link instability. Using traditional positioning techniques based on geometric and mathematical models to perform positioning in NLOS scenarios is inaccurate. Furthermore, current NLOS positioning research involves large amounts of input data, leading to high computational complexity, making it difficult to meet practical application requirements. For example, current AI-based NLOS positioning research requires a large amount of model input data, necessitating highly complex structures for each layer of the AI ​​network, resulting in high computational complexity and overhead, making it difficult to meet commercial deployment requirements. Especially with the deployment of 5G, the number of smart antennas is increasing. Inputting the sampling data from all antennas into the AI ​​network is not only impossible for real-time online learning, but the processing power requirements also far exceed the engineering capacity.

[0080] In response to this, this disclosure proposes a technical solution to reduce positioning complexity and improve positioning accuracy in NLOS scenarios.

[0081] Figure 1 A schematic diagram of an NLOS channel environment is shown. (e.g.) Figure 1As shown, wireless signals from a target object (e.g., user equipment) reach a signal receiving device (e.g., a base station) only after being reflected by multiple scatterers. In other words, the channel between the target object and the signal receiving device is an NLOS channel, which includes multiple signal paths. The set of paths reflected by the same scatterer is called a cluster. Paths within the same cluster have similar angles of arrival and delays. An NLOS channel can be understood as being composed of different clusters, each cluster including multiple paths.

[0082] Figure 2 This is a flowchart illustrating a method performed by a network node according to an embodiment of the present disclosure.

[0083] Reference Figure 2 In step S210, the channel impulse response (CIR) of the non-line-of-sight channel between the target object and the signal receiving device is obtained. CIR describes the multipath effect of a channel. In wireless communication, signals propagate along different paths with varying lengths and propagation times, resulting in time delays and phase shifts when the signal reaches the receiver. The channel impulse response describes the energy distribution of the signal at different points in time after passing through the channel. The CIR can be obtained in step S210 using any method known in the art, and this disclosure is not limited thereto.

[0084] In step S220, multiple signal paths of the non-line-of-sight channel are determined based on the channel impulse response. These multiple signal paths can be a set of all signal paths of the non-line-of-sight channel, or they can be multiple signal paths determined after removing at least one signal path from the set of signal paths of the non-line-of-sight channel. Optionally, according to an embodiment of this disclosure, step S220 may include: determining the cross-correlation power of each signal path in the set of signal paths of the non-line-of-sight channel based on the channel impulse response, wherein the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window; and removing at least one signal path from the set of signal paths of the non-line-of-sight channel based on the cross-correlation power to determine multiple signal paths. CIR signals have a large number of paths in large bandwidth and large antenna arrays; for example, with a 100MHz bandwidth and a 64RX antenna, the number of possible paths is 2048*64 = 131072. Therefore, the positioning complexity can be reduced by decreasing the number of paths used for target positioning.

[0085] According to an embodiment, the signal receiving device includes multiple antennas, and the cross-correlation power includes the cross-correlation power of each signal path corresponding to each of the multiple antennas. Determining multiple signal paths by removing at least one signal path from the set of signal paths in the non-line-of-sight channel based on the cross-correlation power may include: determining a cross-correlation power threshold based on the cross-correlation power; removing at least one signal path from the set of signal paths in the non-line-of-sight channel based on the cross-correlation power threshold, and using the remaining signal paths after removing the at least one signal path as the multiple signal paths. For example, the cross-correlation power of each signal path corresponding to the multiple antennas can be averaged and arranged from high to low to obtain an average cross-correlation power queue; the average cross-correlation power corresponding to the start position of the effective path in the average cross-correlation power queue can be determined based on the predetermined time window and the cross-correlation power of each signal path corresponding to each antenna, and used as the cross-correlation power threshold. However, the method of determining the cross-correlation power threshold based on the cross-correlation power is not limited to the above examples.

[0086] In one example, it can be like this: Figure 3 As shown in 301 and 302, the cross-correlation power P of each signal path i corresponding to each antenna is calculated using the following formula. i :

[0087] Where i = 1, ..., N, M is the size of the time window, and H is the channel impulse response (CIR). The signal part of the CIR is a steady-state random process, and the noise part of the CIR is a Gaussian random process, with different cross-correlation properties. If the time is very short, i.e., the value of M is relatively small, the channel environment can be considered almost constant. The cross-correlation result of Gaussian noise is 0, so the cross-correlation result of the channel impulse response is the power of the signal itself. Through the above cross-correlation operation, signal noise can be removed, making the cross-correlation power the power after removing noise interference. Therefore, based on the cross-correlation power, invalid signal paths can be removed more accurately, and more valuable signal paths can be obtained.

[0088] To obtain the cross-correlation power P of each signal path i for each antenna i Then, the Akaike information criterion (AIC) can be used, based on the cross-correlation power P. i The method of determining the cross-correlation power threshold for removing invalid signal paths is referred to in this disclosure as the PAIC method, but the method of determining the cross-correlation power threshold is not limited to this method.

[0089] For example, such as Figure 3 As shown in Figure 303, the average cross-correlation power of all antennas can be calculated. For example, the cross-correlation power P of each signal path i of antenna 0, antenna 1, ... antenna Nrx can be calculated.i Calculate the average. Next, sort the average cross-correlation power of all signal paths from high to low. For K = 1, 2, ..., M-1, iterate through each K and calculate the starting position of the effective path after arranging the average cross-correlation power from high to low. The calculation formula is as follows, where N′ is a configurable hyperparameter.

[0090]

[0091] The cross-correlation power value corresponding to the starting position of the effective signal path can be used as the cross-correlation power threshold. This threshold can be used to remove invalid signal paths, resulting in a set of effective signal paths, such as... Figure 3 As shown in 304, by utilizing a threshold determined based on cross-correlation power, worthless signal paths can be adaptively removed from the set of signal paths, while valuable signal paths are selected.

[0092] Return to reference Figure 2 After determining multiple signal paths in step S220, multiple clusters are determined based on the multiple signal paths in step S230. According to an embodiment, the multiple clusters can be determined based on the power information and time delay information corresponding to each of the multiple signal paths. For example, multiple clusters can be determined based on the power delay spectrum corresponding to the multiple signal paths. For example, the power information corresponding to each signal path may include, but is not limited to, the cross-correlation power of each signal path. As described above, the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window. The power delay spectrum can be constructed based on the cross-correlation power and time delay information. The cross-correlation power and time delay information can be obtained based on the channel impulse response corresponding to each signal path.

[0093] In this disclosure, determining multiple clusters based on the multiple signal paths is also referred to as "cluster detection." Cluster detection can be performed based on the similarity between clusters. Typically, each signal path within a cluster has its own corresponding horizontal and vertical angles of arrival, and their arrival times are similar, resulting in power peaks in the time domain. Therefore, cluster detection can be performed by detecting these power peaks. For example, as... Figure 4 As shown, cluster detection can be achieved by detecting power peaks in the time domain based on the power and time delay corresponding to each signal path.

[0094] According to an embodiment, optionally, the power peak value in the time domain can be detected based on the variation trend of the cross-correlation power of each of the plurality of signal paths and the time delay information, and the plurality of clusters can be determined based on the power peak value. For example, the power peak value can be detected by the first-order and second-order difference of the cross-correlation power, thereby determining the plurality of clusters.

[0095] Figure 5 This is a schematic diagram illustrating an example of cluster detection according to an embodiment of the present disclosure. Figure 5 As shown, firstly, ① the rate of change of the cross-correlation power of each path in the time domain is obtained by calculating the first difference of the cross-correlation power of each path, such as... Figure 5 As shown in 501 and 502; secondly, ② apply the sign function to the result of the first-order difference. If the first-order difference is greater than 0, return 1; if the first-order difference is less than 0, return -1, as shown in 501 and 502. Figure 5 As shown in Figure 503. The `sign` function can be used to obtain the trend of cross-correlation power, allowing us to focus only on the trend of signal change, not the magnitude of the rate of change. Next, ③ calculate the second-order difference for the data after calculating `sign` on the first-order difference result. If the result of the second-order difference is -2, it indicates that the next moment is a peak; if the result of the second-order difference is 2, it indicates that the next moment is a trough, as shown in Figure 503. Figure 5 As shown in 504. The range of a cluster can be initially determined by the troughs. Subsequently, ④ a threshold can be set to determine whether two adjacent peaks belong to the same cluster, such as... Figure 5 As shown in Figure 505, for example, if the power difference between two adjacent peaks is less than a threshold, then the two adjacent peaks are determined to belong to the same cluster; otherwise, they do not belong to the same cluster. The same operation can be performed on troughs. Finally, ⑤ outputs each cluster, as shown... Figure 5 As shown in Figure 506, the threshold can be determined based on the scene in which the target object is located, but is not limited thereto. For example, if the scene in which the target object is located is more complex (e.g., with many scatterers), the threshold can be a relatively small value; if the scene in which the target object is located is simpler, the threshold can be a relatively large value. The selection of the threshold can affect the number of clusters deleted.

[0096] Return to reference Figure 2 After determining multiple clusters based on the multiple signal paths, in step S240, at least one cluster is selected from the multiple clusters based on the power information and / or angle of arrival information corresponding to the multiple clusters. Optionally, the power information and / or the angle of arrival information may be obtained based on the channel impulse response corresponding to the signal paths included in the multiple clusters, but the method of obtaining the power information and / or the angle of arrival information is not limited to this.

[0097] Optionally, according to an embodiment, step S240 may include: obtaining power distribution information and / or angle distribution information corresponding to the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters, wherein the power distribution information is used to represent the spatial distribution of power corresponding to the plurality of clusters, and the angle distribution information is used to represent the spatial distribution of angle of arrival corresponding to the plurality of clusters; selecting the at least one cluster from the plurality of clusters based on the power distribution information and / or the angle distribution information. According to an embodiment, optionally, the power corresponding to the plurality of clusters includes the cross-correlation power corresponding to each signal path included in the plurality of clusters, wherein the cross-correlation power is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window. For example, the power may be the cross-correlation power mentioned above in the section on removing invalid signal paths, but is not limited thereto.

[0098] Optionally, according to embodiments of this disclosure, obtaining power distribution information and / or angle distribution information corresponding to the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters includes: constructing a spatial grid corresponding to the target object based on the angle of arrival information, wherein the angle of arrival information includes a horizontal angle of arrival and a vertical angle of arrival, and the coordinates of the horizontal and vertical directions of the spatial grid are the horizontal angle of arrival and the vertical angle of arrival, respectively; mapping the angle of arrival information to the spatial grid to obtain the angle distribution information, and / or mapping the power information to the spatial grid to obtain the power distribution information.

[0099] Optionally, according to embodiments of this disclosure, constructing a spatial grid corresponding to the target object based on the angle of arrival information includes: estimating the horizontal angle of arrival (AoA) and the vertical angle of arrival (ZoA) corresponding to the strongest signal path in each of the plurality of clusters; determining the coordinate range of the spatial grid in the horizontal and vertical directions based on the maximum and minimum values ​​of the horizontal and vertical angles of arrival corresponding to the strongest signal path in each of the plurality of clusters; and determining the resolution of the spatial grid in the horizontal and vertical directions based on the coordinate range and the number of the plurality of clusters.

[0100] Figure 6 This is a schematic diagram illustrating an example of constructing a spatial grid according to an embodiment of the present disclosure. Spatial grids of different resolutions can be constructed for different target objects, providing a basis for selecting clusters with high reliability and diversity. For example, a spatial grid can be constructed through the following operations:

[0101] (1) Estimate the AoA and ZoA of the strongest path in each cluster, and use them as the AoA and ZoA corresponding to that cluster. The AoA and ZoA of the strongest path in each cluster can be estimated in various ways, and this disclosure is not limited in this regard. For example, traditional AoA and ZoA estimation methods can be used, or estimation methods based on artificial intelligence models can be used.

[0102] (2) Figure 6 As shown, based on the estimated AoA and ZoA, spatial grids of different resolutions are established for the target object. The maximum and minimum values ​​of AoA and ZoA for all candidate clusters can be set as the upper and lower bounds of the spatial grid, respectively. The resolution of the grid size can depend on the scene or environment in which the target object is located. The scene or environment can be, for example, Uma (Urban Macrocell), UMi-Street Canyon (Urban Microcell Street Canyon), Indoor-office, RMa (Rural Macrocell), etc. Generally, the sparser the environment, the larger the resolution of the grid size; the more complex the environment, the smaller the grid size. For example, if the environment in which UE1 is located is more complex than the environment in which UE2 is located, then the resolution of the spatial grid corresponding to UE1 is smaller than the resolution of the spatial grid corresponding to UE2. For example, the construction process of the spatial angle grid can be as follows:

[0103] (a) Calculate the AoA of all candidate clusters (i.e., the multiple clusters determined in step S230).

[0104] The maximum and minimum values ​​of ZoA are determined, and these are set as the upper and lower bounds of the grid in the horizontal and vertical directions, respectively, thus obtaining the ranges of AoA and ZoA.

[0105] (b) For example, the representable range of AoA and ZoA can be divided by the number of all candidate clusters, such that each cluster is distributed across a single grid. This determines the resolution of the spatial grid in both the horizontal and vertical directions. The number of all candidate clusters is related to the scene in which the target device is located, and the number of candidate clusters can be controlled based on the scene in which the target device is located, thereby controlling the resolution of the spatial grid based on the scene in which the target object is located. However, the methods for determining the resolution of the spatial grid in both the horizontal and vertical directions are not limited to the above examples. For example, AoA and ZoA can be used...

[0106] The range of ZoA is divided by other values ​​determined based on the scene in which the target object is located, which can reflect the complexity of the scene in which the target object is located.

[0107] After constructing the spatial grid by determining its range and resolution, the power information and angle of arrival information corresponding to the multiple clusters determined in step S230 can be mapped onto the spatial grid. Through this mapping, the angle and power distribution corresponding to the multiple clusters can be obtained, such as... Figure 4 As shown, the above mapping process transforms each path in multiple clusters into the angle domain, thereby enabling cluster selection in the angle domain, allowing the selection of clusters with higher power and more diverse angles.

[0108] Figure 7 This is a schematic diagram illustrating an example of a selection cluster according to an embodiment of the present disclosure.

[0109] For example, mapping the power information to the spatial grid to obtain the power distribution information may include: mapping the power of the strongest signal path in each of the plurality of clusters to the spatial grid to obtain the power distribution information. Figure 7 As shown, each cluster can be represented by its most powerful signal path. By mapping the power of only the most powerful signal path in each cluster to a spatial grid, power distribution information can be obtained; that is, the distribution of the maximum power in each cluster within the spatial grid (e.g., ...). Figure 7 Power distribution in the data (1). Optionally, after mapping the power of the strongest signal path in each cluster to a spatial grid, the power mapped to the spatial grid can be further normalized to obtain normalized power distribution information, that is, the distribution of normalized power in the spatial grid (e.g., Figure 7 Power distribution in 2).

[0110] For example, mapping the arrival angle information to the spatial grid to obtain the angle distribution information may include: if the arrival angle corresponding to a cluster falls within a certain grid in the spatial grid, then the value of that grid is set to a preset value. For example, the preset value can be 1, but is not limited to this; otherwise, the value of that grid is set to empty or 0. This allows the acquisition of angle distribution information, that is, the distribution of the preset value in the spatial grid (e.g., Figure 7 (The angular distribution in the middle).

[0111] After obtaining power distribution information and / or angular distribution information corresponding to multiple clusters, at least one cluster can be selected from the multiple clusters based on the power distribution information and angular distribution information. Optionally, according to embodiments of this disclosure, selecting at least one cluster from the multiple clusters based on the power distribution information and / or the angular distribution information includes: performing convolution on the angular distribution information in the spatial grid using a first convolution kernel to obtain an angular distribution density; performing convolution on the power distribution information in the spatial grid using a second convolution kernel to obtain a power distribution density; and selecting at least one cluster from the multiple clusters based on the angular distribution density and the power distribution density. According to embodiments, the sizes of the first and second convolution kernels can be determined based on the scene in which the target object is located. Generally, the more complex the scene in which the target object is located, the smaller the convolution kernel; the simpler or emptier the scene in which the target object is located, the larger the convolution kernel. For example, furthermore, the size of the convolution kernel can depend on the scene and the balance between complexity and accuracy.

[0112] In this disclosure, the first convolution kernel is also referred to as an "angular convolution kernel," and the second convolution kernel is also referred to as a "power convolution kernel." For example, both the first and second convolution kernels can be convolution kernels of a predetermined size with preset values, such as a predetermined size of 3*3 and a preset value of 1, but are not limited to this.

[0113] By applying angular and power convolution kernels to a spatial grid, respectively, the angular distribution density and power distribution density can be calculated. Subsequently, at least one cluster can be selected from the plurality of clusters based on the angular and power distribution densities, such as... Figure 7 As shown in 701 and 702. According to an embodiment, selecting at least one cluster from the plurality of clusters based on the angular distribution density and the power distribution density may include: obtaining an angular power spectral density based on the angular distribution density and the power distribution density; and selecting at least one cluster from the plurality of clusters based on the angular power spectral density.

[0114] The following section introduces the calculation of angular distribution density, power distribution density, and angular power spectral density with examples.

[0115] For example, the angular distribution density can be calculated as follows: By traversing the spatial grid with an all-1 angular convolution kernel, the resulting feature map is the angular distribution density ρ(angle) of the cluster. For example, as... Figure 7 As shown in 701, the angle distribution is convolved in the spatial grid using an all-1 angle convolution kernel. That is, the all-1 convolution kernel is convolved with the values ​​(such as 1) in each convolution region of the spatial grid to obtain the angle distribution density ρ (angle).

[0116] Optionally, according to an embodiment, performing convolution on the power distribution information in the spatial grid using a second convolution kernel to obtain a power distribution density includes: adjusting the second convolution kernel based on the power distribution information in the convolution region of the second convolution kernel to obtain an adjusted second convolution kernel; and performing convolution on the power distribution information in the convolution region using the adjusted second convolution kernel to obtain a power distribution density. For example, the second convolution kernel can be adjusted based on the maximum power value in the convolution region of the second convolution kernel, for example, by determining a scaling factor based on the maximum value, and obtaining the adjusted second convolution kernel based on the scaling factor and the second convolution kernel. By setting an adaptive second convolution kernel according to the power distribution in the convolution region, it is helpful to select representative clusters more accurately.

[0117] For example, the power distribution density can be calculated as follows: A power convolution kernel is obtained by multiplying a convolution kernel of all 1s by a scaling factor (scale(i, j), including but not limited to the maximum power value of each convolutional region). This results in an adjusted power convolution kernel, enabling adaptive power convolution kernels where each convolutional region has a more suitable kernel. For example, by traversing a spatial grid using the adjusted power convolution kernel, the resulting feature map is the cluster's power distribution density ρ(power). Figure 7 Convolve the power distribution shown in Figure 2 to obtain the power distribution density ρ(power).

[0118] For example, multiplying the angle distribution density ρ(angle) by the power distribution density ρ(power) yields the angular power spectral density ρ. APSD , i.e., ρ APSD =ρ(angle)*ρ(power).

[0119] Using power convolution kernels adjusted by scaling factors can help select representative clusters more accurately. This is because whether a cluster is selected is affected by power and angle, for example, by the power and distribution density of the cluster, but power often has a higher priority. Figure 8 This is a schematic diagram illustrating examples of using and not using a scaling factor when selecting clusters according to embodiments of the present disclosure. Figure 8 In the example on the left, the power convolution kernel is not scaled. Figure 8 In the example on the right, the power convolution kernel is applied with a scaling factor s(i,j), where s(i,j) is short for scale(i,j), indicating the scaling factor at the convolution position in the i-th row and j-th column of the kernel. Figure 8As shown on the right, in the power values ​​of the spatial grid, the maximum value of the first region is 20, so scale(0,0) = max pooling(0,0) = 20. The maximum value of the second region is 2, so scale(0,1) = max pooling(0,1) = 2. After the power convolution kernel and power distribution are applied, the resulting power density values ​​are ρ(0,0) = 400, ρ(0,1) = 36, and ρ APSD (0,0)=400,ρ APSD (0,1) = 324. For example... Figure 8 As shown on the left, if a power convolution kernel without a scaling factor is used, for example, always using a fixed kernel of all 1s, the resulting power density values ​​after the power convolution kernel and power distribution are ρ(0,0) = 20 and ρ(0,1) = 18. Without scaling, after calculating the angular distribution density ρ(APSD), the cluster corresponding to the maximum power of 20 is discarded, potentially leading to incorrect clusters. Therefore, when selecting clusters based on power distribution density and angular distribution density, applying a scaling factor to the power convolution kernel can give power greater priority in cluster selection. Figure 8 As shown on the right, the correct clusters can be selected by using the power convolution kernel after applying the scaling factor.

[0120] After determining the angular power spectral density, at least one cluster can be selected from multiple clusters based on the angular power spectral density. For example, selecting at least one cluster from multiple clusters based on the angular power spectral density may include: performing data mapping on the angular power spectral density to obtain mapped data within a preset interval; determining a first threshold based on the mapped data; selecting at least one convolutional region in the spatial grid based on the first threshold; and selecting the cluster with the strongest power from each of the at least one convolutional region. For example, performing data mapping on the angular power spectral density to obtain mapped data within a preset interval includes: normalizing the angular power spectral density and applying a sigmoid function to the normalized result to obtain mapped data within the (0,1) interval, which can be used as a score for each convolutional region. The first threshold can be set based on this score. For example, the first threshold can be the average, median, or other values ​​of the scores for each convolutional region. For example, convolutional regions with scores exceeding the average of all scores can be selected, and then the cluster with the strongest power can be selected from each of the selected convolutional regions.

[0121] For example, such as Figure 7 As shown in Figure 702, the angular power spectral density ρ is obtained by multiplying the angular distribution density ρ(angle) by the power distribution density ρ(power). APSD Then, ρ can be processed in the following way. APSDNormalization (also known as standardization) yields ρ′ APSD To make it conform as closely as possible to the standard normal distribution N(0,1):

[0122] μ = mean(ρ) APSD )σ=std(ρ APSD )

[0123] ρ′ APSD =(ρ APSD -μ) / σ.

[0124] For example, such as Figure 9 The example shown on the right, ρ APSD The distribution is between 0.001 and 10000, with a mean u = 1361 and a standard deviation σ = 2929. Convert this to ρ′. APSD The distribution after that ranges from -0.4646 to 2.949.

[0125] Next, we can work on ρ′ APSD Apply the sigmoid function to map it to the (0,1) interval to obtain the final cluster score.

[0126] Score = sigmoid(ρ′) APSD )

[0127] The sigmoid function is an S-shaped curve that can map any data to the (0,1) interval, meaning it can map data of all orders of magnitude to the same order of magnitude. To preserve the distribution characteristics as much as possible, we can first standardize the dataset and then use the sigmoid function to obtain the score. The sigmoid function can also relatively raise the lower bound of the dataset and lower the upper bound, making the algebraic mean of the newly generated dataset closer to the median, thus making it more suitable as a threshold for cluster selection. Figure 9 As shown on the right, the scores range from 0.3859 to 0.9502. However, in cases like... Figure 9 In the example shown on the left, when calculating ρ APSD Subsequently, without further data mapping (e.g., normalization and application of Sigmoid), ρ was directly... APSD Using the average value as a threshold for cluster selection may ultimately lead to some representative clusters being missed.

[0128] Next, the threshold used for cluster selection will be... th Set as the algebraic mean of all scores, its calculation method is as follows: cluster th =mean(Score). For example... Figure 9As shown on the right, the calculated average score is 0.47252. Based on this threshold, four convolutional regions were selected, with corresponding scores of 0.9502, 0.4948, 0.4863, and 0.4777. For all selected convolutional regions, the cluster with the strongest power in each region was chosen as the representative of that region, thus completing the selection of representative clusters.

[0129] Figure 10 A schematic diagram illustrating an example of cluster selection according to an embodiment of this disclosure is shown. Figure 10 As shown, after identifying multiple clusters, the AoA and ZoA of the signal paths included in each cluster can be estimated based on the channel impulse responses corresponding to those clusters. Based on the estimated AoA or ZoA, some special clusters can be initially removed. For example, as... Figure 11 As shown, in some scenarios, if the vertical antenna distance of the signal receiving device is large, the ZoA of the edge reflector is often incorrectly estimated. This problem is caused by the antenna configuration. To avoid this, after calculating the AoA and ZoA of the path corresponding to such an edge reflector, the cluster corresponding to the edge reflector can be deleted based on its size. Deleting such a cluster helps improve the accuracy of subsequent target object localization. For example, as... Figure 11 As shown, the cluster corresponding to a ZoA of 78 degrees can be deleted. However, depending on the antenna configuration, such cluster deletion may not be necessary. After deleting a specific cluster, a spatial grid can be constructed as described in the text, and the angular distribution density and power distribution density can be obtained based on the constructed spatial grid, further obtaining the angular power spectral density. Finally, the cluster for target location is selected based on the angular power spectral density.

[0130] Return to reference Figure 2 After selecting at least one cluster in step S240, the target object is located based on the at least one cluster in step S250. For example, the target object can be located using an artificial intelligence model based on the at least one cluster.

[0131] For example, one approach is to input information about all signal paths included in each selected cluster into an artificial intelligence model to obtain the location information of the target object. Such an artificial intelligence model is called a "non-progressive AI network". Figure 12 This is a schematic diagram illustrating a non-progressive AI network according to an embodiment of the present disclosure. For example, information related to the signal path may include angle information, power information, and delay information corresponding to the signal path, obtained based on the channel impulse response corresponding to the signal path. Figure 12As shown, relevant information for each path is extracted based on the channel impulse response corresponding to each path, forming feature data for each cluster. This feature data is then input into the AI ​​network to output the target object's location information, such as the horizontal and vertical angles of arrival. However, this method is highly complex. For example, obtaining the angle, power, and delay of a path requires approximately 70k complexity. Assuming there are N clusters and J paths in each cluster, the overall complexity is N*J*70k. For non-progressive AI networks, the increased data processing volume due to the extraction of relevant information for each path and the input of this extracted information leads to higher localization complexity. To address this and reduce localization complexity, this disclosure proposes a progressive data input method.

[0132] Optionally, according to an embodiment, step S250 may include: selecting at least one first signal path from each of the at least one cluster based on the power of the signal path; obtaining a positioning result using an artificial intelligence (AI) model based on relevant information of the signal paths selected from each cluster; determining whether to output a positioning result based on the difference between the current positioning result and the positioning result obtained using the AI ​​model last time; if so, obtaining the location information of the target object based on the current positioning result; otherwise, selecting at least one second signal path from each cluster based on the power of the signal path, and performing the step of obtaining a positioning result using an AI model based on relevant information of the signal paths selected from each cluster, wherein the second signal path includes the first signal path. For example, selecting at least one first signal path from each of the at least one cluster based on the power of the signal path may include: selecting the strongest signal path from each of the at least one cluster based on the power of the signal path. For example, selecting at least one second signal path from each cluster based on the power of the signal path may include: selecting the strongest signal path and the second strongest signal path from each cluster based on the power of the signal path. For example, obtaining a positioning result using an artificial intelligence (AI) model based on relevant information of signal paths selected from various clusters may include: obtaining feature data corresponding to the relevant information of the signal paths selected from various clusters, and inputting the feature data into the AI ​​model to obtain the positioning result. As mentioned above, for example, the relevant information of the signal path may include angle information, power information, and time delay information corresponding to the signal path obtained based on the channel impulse response corresponding to the signal path. Accordingly, for example, the feature data corresponding to the relevant information may include angle features, power features, and time delay features. In the following text, for ease of description, the combination of angle features, power features, and time delay features is referred to as "APD features". For example, determining whether to output a positioning result based on the difference between the current positioning result and the previous positioning result obtained using the AI ​​model may include: if the difference between the current positioning result and the previous positioning result obtained using the AI ​​model meets a predetermined condition, then determining to output the positioning result. For example, the predetermined condition may be that the difference between the current positioning result and the previous positioning result is less than a threshold. In the above operation, since the relevant information of the signal path is progressively input into the AI ​​model, and the positioning result is output when the difference between the current positioning result and the previous positioning result meets the predetermined conditions, the input of information to the AI ​​model is stopped. Therefore, the amount of data input to the artificial intelligence model can be reduced, thereby effectively reducing the positioning complexity. At the same time, since the positioning result is obtained based on the increasingly more relevant information of the signal path, the positioning accuracy of the UE can be improved.In the following text, an artificial intelligence model operating in the manner described above may be referred to as an "incremental AI network" or an "incremental localization network (PPosNet)". For example, an incremental AI network may be an incremental transformer network, but it is not limited to this.

[0133] Figure 13 This is a schematic diagram illustrating a progressive AI network according to an embodiment of the present disclosure. Figure 13 As shown, firstly, relevant information of the first path (the most powerful path) in each cluster is extracted, such as angle information, power information, and latency information. This information is combined to obtain feature data corresponding to each first path. All feature data corresponding to the first path are input into the AI ​​network to obtain the first localization result, such as the horizontal and vertical angles of arrival corresponding to the target object. Then, relevant information of the second path (the second most powerful path) in each cluster is extracted, such as angle information, power information, and latency information. This information is combined to obtain feature data corresponding to each first path. The feature data corresponding to the second path in each cluster is input together with the feature data corresponding to the first path into the AI ​​network. In this case, the AI ​​network will obtain the second localization result based on the feature data corresponding to the second and first paths. This process is repeated, and a third time, the feature data corresponding to the most powerful path, the second most powerful path, and the third most powerful path of all clusters are input into the AI ​​model until an early stopping condition is met, such as the difference between two localization results being less than a threshold. Using a non-progressive AI network effectively reduces localization complexity by reducing both the amount of feature data extracted and the amount of data input to the AI ​​network. For example, if obtaining the angle, power, and delay features of a path requires approximately 70k complexity, and assuming there are M clusters, N paths in each cluster, and J paths in each cluster, then the overall complexity is N*M*70k (2≤M≤J). Compared to a non-progressive AI network, this saves (JM) / J in complexity.

[0134] According to an embodiment, the relevant information of the information path may include angle information, power information, and time delay information corresponding to the signal path, extracted based on the channel impulse response corresponding to the signal path. Accordingly, for example, the feature data may be features composed of the angle, power, and time delay corresponding to each signal path.

[0135] In other words, "APD features" can effectively describe the characteristics of a signal path. For example, angles can be AoA and ZoA, representing the direction of the incoming wave. Power can be cross-correlation power, indicating the scattering material (absorption and scattering characteristics). Time delay can represent the path distance. A combination of multiple APD features corresponds one-to-one with the location of the target object. For example, as... Figure 14 As shown, User1, User2, and User3 each correspond to multiple paths, and each path corresponds to one APD feature. Therefore, User1, User2, and User3 each correspond to multiple combinations of APD features. This demonstrates a one-to-one correspondence between the UE's APD feature set and its location. Thus, the UE's location can be estimated based on its APD feature set. Extracting APD features based on CIR can reduce the dimensionality of the input data, further reducing computational complexity and positioning complexity. CIR can include the CIR corresponding to each signal path of each antenna, for example, CIR[t][Ant0]...CIR[t][AntN]. rx The APD characteristics corresponding to each signal path can be obtained based on the CIR corresponding to each signal path. For example... Figure 13 As shown, after extracting the APD features, the APD features can be input into the AI ​​network in the progressive input method described above to obtain the position information of the target object, such as the horizontal and vertical directions of the target object.

[0136] As described above, after selecting at least one second signal path from each cluster, an AI model is used to obtain the localization result based on the relevant information of the signal paths selected from each cluster. Optionally, obtaining the localization result based on the relevant information of the signal paths selected from each cluster may include: performing a first attention operation between the feature data corresponding to the relevant information of the signal paths in different clusters to obtain a first attention result; performing a second attention operation between the feature data corresponding to the relevant information of the signal paths in the same cluster to obtain a second attention result; and obtaining the localization result based on the first attention result and the second attention result. By performing the first and second attention operations in the above manner, the accuracy of target object localization can be further improved. In the following text, the first attention operation is also referred to as "inter-cluster attention," and the second attention operation is also referred to as "intra-cluster attention." Intra-cluster attention refers to the attention of each path within a cluster, which can learn the reflection characteristics of the scatterer by learning the relationship between the power, angle of arrival, and delay corresponding to the same cluster. Inter-cluster attention learns the complete environmental characteristics by learning the relationship between the power, angle of arrival, and delay corresponding to different clusters. Therefore, by utilizing intra-cluster attention and inter-cluster attention, the characteristics of the scatterer in the space where the target object is located and the NLOS channel environment can be learned.

[0137] Figure 15 This is a schematic diagram illustrating the reflection loss of scatterers made of different materials. For example... Figure 15 As shown, different scatterers have different reflection characteristics. For example, glass curtain walls and concrete walls have different reflection losses. Furthermore, the incident angle (i.e., the input direction of the signal) also affects the reflection loss. Reflection loss contains positional information and can be used to locate the target object. Due to the different materials and structures of the scatterers, the relationship between power, time delay, and angle varies from scatterer to scatterer. Therefore, the reflection characteristics of scatterers made of different materials, such as reflection loss, can be learned through intra-cluster attention.

[0138] Figure 16 This is a schematic diagram illustrating examples of inter-cluster correlation and intra-cluster correlation according to embodiments of the present disclosure. For example... Figure 16 As shown, correlations exist not only within clusters (i.e., intra-cluster correlations) but also between different scatterers in space (i.e., inter-cluster correlations). Intra-cluster correlations can be learned through intra-cluster attention, and inter-cluster correlations can be learned through inter-cluster attention. For example, the input to the attention operation can be APD features, including angle, power, and delay. Different materials result in different reflectivity coefficients; the material of the reflecting surface can be learned through the power relationship, the distance to the reflecting surface through the delay relationship, and the angular relationship of the reflecting surface through the angle of arrival relationship. This learned knowledge can be used for more accurate localization of target objects.

[0139] Figure 17 This is a schematic diagram illustrating how progressive input, inter-cluster attention, and intra-cluster attention reduce positioning complexity and improve positioning accuracy according to embodiments of the present disclosure. Figure 17 As shown, since the paths in each cluster have similar AoA / ZoA, it's not necessary to use all paths in the cluster; instead, a subset of paths can be selected for subsequent target object localization. For example, feature data corresponding to paths can be progressively input based on their power levels. If the localization accuracy changes only slightly between consecutive inputs, the process stops early. Therefore, the features corresponding to the remaining paths in the cluster do not need to be input into the AI ​​network, reducing complexity through this progressive input method. Furthermore, by utilizing inter-cluster and intra-cluster attention, reflection and environmental features can be learned through the relationship between power, angle of arrival, and delay, thereby improving localization accuracy. Figure 17 As shown, feature data corresponding to the paths of the selected cluster can be input based on the power intensity of the paths. First, the feature data corresponding to the path with the highest power is selected as the initial input, which can obtain a rough positioning accuracy. Second, the feature data corresponding to the path with the second highest power is input, and then combined with the previously input feature data corresponding to the path with the highest power for positioning, which can improve the positioning accuracy.

[0140] Inter-cluster attention and intra-cluster attention can be applied using attention matrix multiplication. Figure 18 This is a schematic diagram illustrating the performance of inter-cluster attention and intra-cluster attention in an attention layer included in a progressive localization network according to embodiments of the present disclosure. Figure 18 As shown, the APD features corresponding to path0 (the most powerful path) selected from clusters 1 and 2 are subjected to inter-cluster attention, which includes: respectively with the weight matrix W Q′ W k′ and W V′ After multiplication, the data passes through a connection function and a normalized exponential function, and finally through a feedforward neural network (FFN). The APD features corresponding to path 1 (the second strongest power path) selected from clusters 1 and 2 are also subject to inter-cluster attention. Additionally, the APD features corresponding to path 0 of cluster 1 and path 1 of cluster 1 are subject to intra-cluster attention, and are respectively correlated with the weight matrix W. Q W k and W V After multiplication, the data passes through a connection function and a normalized exponential function, and finally through an FFN (Focused Filtering Network). Additionally, the APD features corresponding to path0 in cluster 2 and path1 in cluster 2 are also subject to intra-cluster attention. The output of the FFN is further normalized before the localization result is output.

[0141] Figure 19 This is a schematic diagram illustrating the structure of a progressive positioning network according to an embodiment of the present disclosure. Figure 20 This is a schematic diagram illustrating the operations performed by a progressive positioning network. For example... Figure 19 As shown, the progressive localization network may include an APD feature extraction module, a cluster self-attention encoding module, a localization regressor, and an early stopping decision module. Clusters obtained after cluster detection and selection are input into the APD feature extraction module to extract APD features, namely angle, power, and delay features. This module ensures that different signal paths in different clusters enter the same group based on the power principle, and estimates the angle of arrival for each signal path in the group. For each signal path in the group, the horizontal and vertical angles of arrival, along with the corresponding power and delay, are combined to obtain angle, power, and delay features. The extracted APD features are input into the cluster self-attention encoding module. The cluster self-attention encoding module performs cluster self-attention encoding, i.e., the inter-cluster attention and intra-cluster attention mentioned above. For example, as... Figure 20As shown, the first input bundles paths 0 from each cluster together as path group 0 and inputs it into the AI ​​network. The second input bundles paths 1 from each cluster together as path group 1 and inputs it along with path group 0 into the AI ​​network. Then, an early stop judgment is performed. If the positioning result meets the accuracy requirements, the iteration stops; otherwise, a third iteration is performed, inputting path groups 0 / 1 / 2 together into the AI ​​network, and so on. The newly added path groups can continuously supplement channel environment information, improving the UE's positioning accuracy. The positioning regressor performs position regression calculations, combining the APD features of all selected clusters to obtain the horizontal and vertical angles of arrival of the target object. For each iteration, the output features of the current iteration are used to continuously approach the true position of the target object. The early stop decision module executes an early stop mechanism, for example, it can stop if there are no more angle, power, or delay features to input. Or, if the output Y of the previous and next iterations is... n-1 (AoA n-1 ZoA n-1 ) and Y n (AoA n ZoA n The difference between them is less than a threshold, for example, |AoA n -AoA n-1 |< threshold α and |Y n -Y n-1 If | < threshold β, then stop input, where AoA n AoA n-1 These are the horizontal angles of arrival output at the nth and (n-1)th iterations, respectively, ZoA. n ZoA n-1 These are the vertical angles of arrival output for the nth and (n-1)th iterations, respectively. Thresholds α and β can be the same or different.

[0142] After obtaining the location result of the target object through a progressive positioning network, the location result can be validated. Optionally, according to an embodiment, obtaining the location information of the target object based on the current location result mentioned above may include: validating the current location result; if the current location result is valid, obtaining the location information based on the current location result; if the current location result is invalid, obtaining the location information based on the previous location result of the target object. Figure 21 This is a schematic diagram illustrating the validity verification of the positioning results performed according to an embodiment of the present disclosure.

[0143] like Figure 21As shown, based on the UE's previous positioning results (previously recorded AOA and ZOA) and combined with the UE's current positioning results, the possible location range of the UE is predicted. First, the UE's speed is estimated to calculate the maximum distance the UE can reach. For example, based on the CIR, the frequency offset FO is calculated, and the UE's speed is estimated using the following formula:

[0144] Where FC is the frequency of the transmitted signal, and c is the speed of radio waves, i.e., the speed of light.

[0145] After estimating the UE's speed, the predicted range of the UE's possible location is obtained by combining the UE's position at the previous moment. For example, with the UE's position at the previous moment as the center and the maximum distance as the radius of the circle, this circle represents the predicted range of the UE's possible location. The final location information of the UE is determined by combining the predicted range of location with the UE's positioning result predicted by the AI ​​network. If the current positioning result (AoA, ZoA) predicted by the AI ​​network is within the predicted range, then the current positioning result predicted by the AI ​​network can be used as the UE's final location information. If the current positioning result (AoA, ZoA) predicted by the AI ​​network exceeds the predicted range, then the positioning result (AoA, ZoA) from the previous moment is used as the UE's final location information.

[0146] Figure 22 This is an exemplary architecture diagram illustrating a method performed by a network node according to an embodiment of this disclosure. Figure 22 As shown, in some embodiments, the method executed by the network node can acquire the channel impulse response of the non-line-of-sight channel. After determining multiple signal paths based on the channel impulse response, invalid signal paths are removed from the multiple signal paths based on cross-correlation power. For the remaining paths after removing invalid paths, cluster detection and cluster selection are performed to select representative clusters. Based on the relevant information of the signal paths included in the selected clusters, the horizontal and vertical angles of arrival of the target object can be obtained using a progressive positioning network. Finally, the validity of the horizontal and vertical angles of arrival output by the progressive positioning network can be verified to obtain the final horizontal and vertical angles of arrival corresponding to the target object. By removing invalid signal paths, selecting representative clusters, and using a progressive positioning network, not only is the positioning complexity reduced, but the positioning accuracy is also improved. Although in Figure 22 The example shown is a method performed by a network node that removes invalid signal paths, selects representative clusters, and utilizes a progressive positioning network. However, the method for locating a target object according to this disclosure may optionally remove invalid signal paths, select representative clusters, and use a progressive positioning network. That is, this disclosure may use one of the above means alone to reduce the complexity of target positioning and improve positioning accuracy.

[0147] Figure 23 This is an exemplary flowchart illustrating a method performed by a network node according to embodiments of the present disclosure. In some embodiments, the method according to embodiments of the present disclosure can perform target object localization in both non-line-of-sight (LOS) and LOS scenarios. Optionally, the method according to embodiments of the present disclosure may include, in addition to the operations described above for performing target object localization in LOS scenarios (e.g., in combination with the above...), Figures 1 to 22 In addition to the operations described, it may also include operations for performing target object localization in a line-of-sight channel scenario. The method for performing target object localization in a line-of-sight channel scenario can be any localization method known to those skilled in the art, and this disclosure does not limit it.

[0148] For example, such as Figure 23 As shown, after acquiring the channel impulse response, a network node (e.g., a base station) can perform line-of-sight (LAS) and non-LAS channel scenario identification, that is, identify whether the current channel impulse response belongs to the LAS or non-LAS channel. If a non-LAS channel impulse response is identified, the target object localization operation described above for non-LAS channel scenarios is performed. For example, firstly, the cross-correlation power is determined based on the channel impulse response, invalid signal paths are removed based on the cross-correlation power, then, for example, cluster detection is performed based on the power delay spectrum, next, for example, clusters are selected based on the angular power spectral density, based on the selected clusters, the horizontal and vertical angles of arrival corresponding to the target object are predicted using a progressive localization network, and the progressive iteration is terminated by an early stopping decision. Finally, the validity of the angle of arrival is verified. For example, based on the target object's current horizontal and vertical angles of arrival, historical location, frequency offset, and channel impulse response, the corrected horizontal and vertical angles of arrival of the target device are output. If a LAS channel impulse response is identified, the angle of arrival estimation is performed based on the channel impulse response. For example, the horizontal and vertical angles of arrival of the target device can be estimated based on the channel impulse response corresponding to the most powerful signal path. Figure 23 The example method shown can perform positioning not only in non-line-of-sight channel scenarios, but also in line-of-sight channel scenarios.

[0149] As mentioned in the description above, this disclosure can selectively reduce target localization complexity and improve localization accuracy through any one or a combination of three techniques: removal of invalid signal paths, selection of representative clusters, and progressive localization networks. Therefore, optionally, in addition to providing [specific techniques], this disclosure can also [address specific requirements]. Figure 2 In addition to the methods shown, the following descriptions may also be provided. Figure 24 and Figure 25 The method shown.

[0150] Figure 24 This is a flowchart illustrating a method performed by a network node according to an embodiment of this disclosure. Figure 24 As shown, in step S2410, the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device is acquired. In step S2420, based on the channel impulse response, the cross-correlation power of each signal path in the signal path set of the non-line-of-sight channel is determined, wherein the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window. In step S2430, based on the cross-correlation power, at least one signal path is removed from the signal path set of the non-line-of-sight channel to determine multiple signal paths. Optionally, removing at least one signal path from the signal path set of the non-line-of-sight channel based on the cross-correlation power to determine multiple signal paths includes: determining a cross-correlation power threshold based on the cross-correlation power; removing at least one signal path from the signal path set of the non-line-of-sight channel based on the cross-correlation power threshold, and using the remaining signal path after removing the at least one signal path as the multiple signal paths. Optionally, according to an embodiment, the signal receiving device includes multiple antennas, and the cross-correlation power includes the cross-correlation power of each signal path corresponding to each of the multiple antennas. Determining a cross-correlation power threshold based on the cross-correlation power includes: averaging the cross-correlation power of each signal path corresponding to the multiple antennas and arranging them from high to low to obtain an average cross-correlation power queue; determining the average cross-correlation power corresponding to the start position of the effective path in the average cross-correlation power queue based on the predetermined time window and the cross-correlation power of each signal path corresponding to each antenna, and using this as the cross-correlation power threshold. In step S2440, the target object is located based on the multiple signal paths. Figure 24 The method shown primarily reduces positioning complexity by removing invalid signal paths, and according to Figure 24 The method shown, by removing invalid signal paths based on cross-correlation power and locating the target object based on the multiple signal paths remaining after removing invalid signal paths, not only reduces positioning complexity but also improves positioning accuracy. Figure 24 The previous description has already covered... Figure 24 The operations involved, such as how to determine the cross-correlation power and how to remove at least one path based on the cross-correlation power, have been described before and will not be repeated here. For relevant content, please refer to the description above.

[0151] Figure 25 This is a flowchart illustrating a method performed by a network node according to an embodiment of this disclosure. Figure 25As shown, in step S2510, the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device is acquired. In step S2520, multiple signal paths of the non-line-of-sight channel are determined based on the channel impulse response. In step S2530, multiple clusters are determined based on the multiple signal paths. In step S2540, at least one first signal path is selected from each of the multiple clusters based on the power of the signal paths. In step S2550, a positioning result is obtained using an artificial intelligence (AI) model based on the relevant information of the signal paths selected from each cluster. In step S2560, based on the difference between the current positioning result and the positioning result obtained using the AI ​​model last time, it is determined whether to output the positioning result. If yes, the location information of the target object is obtained based on the current positioning result; otherwise, at least one second signal path is selected from each cluster based on the power of the signal paths, and the step of obtaining the positioning result using the AI ​​model based on the relevant information of the signal paths selected from each cluster is executed (i.e., step S2550 is executed). Optionally, after selecting at least one second signal path from each cluster, obtaining the localization result using an AI model based on the relevant information of the signal paths selected from each cluster may include: performing a first attention operation between the feature data corresponding to the relevant information of the signal paths in different clusters to obtain a first attention result; performing a second attention operation between the feature data corresponding to the relevant information of the signal paths in the same cluster to obtain a second attention result; and obtaining the localization result based on the first attention result and the second attention result. The first attention is the inter-cluster attention described above, and the second attention is the intra-cluster attention described above. The artificial intelligence model is the progressive localization network described above. Optionally, the relevant information may include angle information, power information, and time delay information corresponding to the signal path obtained based on the channel impulse response corresponding to the signal path. Figure 25 The method shown, by employing a progressive data input approach, can reduce the scale of data input to the artificial intelligence model, thereby decreasing positioning complexity. Furthermore, since the positioning results are obtained based on feature data corresponding to an increasing number of signal paths, it can improve the positioning accuracy of the UE. Additionally, optionally, inter-cluster attention and intra-cluster attention can further enhance the positioning accuracy of the UE.

[0152] Figure 24 and Figure 25 For details regarding the operations or content already described above, please refer to the section above. Figures 1 to 23 The descriptions in the text will not be repeated here.

[0153] The methods and their effects performed by network nodes according to various embodiments of the present disclosure have been described above with reference to the accompanying drawings. Below, a brief description of the devices according to embodiments of the present disclosure will be provided.

[0154] Figure 26 This is a block diagram illustrating a network node according to an embodiment of the present disclosure. (Refer to...) Figure 26 Network node 2600 may include transceiver 2610 and processor 2620, wherein processor 2620 is coupled to transceiver 2610 and configured to perform any of the methods described above. For example, network node 2600 may be a base station or a network-side functional entity.

[0155] This disclosure also provides an electronic device including at least one processor, and optionally, at least one transceiver coupled to the at least one processor and / or at least one memory, wherein the at least one processor is configured to perform the steps of the method provided in any optional embodiment of this disclosure.

[0156] Figure 27 The diagram shows a structural schematic of an electronic device to which an embodiment of the present invention applies, such as... Figure 27 As shown, Figure 27 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, each of the processor 4001, memory 4003, and transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of this disclosure. Optionally, the electronic device may be a first network node, a second network node, or a third network node.

[0157] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with this disclosure. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0158] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 27 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0159] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.

[0160] The memory 4003 is used to store computer programs or executable instructions that execute the embodiments of this disclosure, and is controlled by the processor 4001 to execute them. The processor 4001 is used to execute the computer programs or executable instructions stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0161] This disclosure provides a computer-readable storage medium storing a computer program or instructions that, when executed by at least one processor, can perform or implement the steps and corresponding content of the aforementioned method embodiments.

[0162] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.

[0163] The terms “first,” “second,” “third,” “fourth,” “1,” “2,” etc. (if present) in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in a sequence other than that shown in the figures or text.

[0164] It should be understood that although arrows indicate various operation steps in the flowcharts of the embodiments of this disclosure, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of the embodiments of this disclosure, the implementation steps in each flowchart can be executed in other orders as required. Furthermore, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage can also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured as required, and the embodiments of this disclosure do not limit this.

[0165] The above text and accompanying drawings are provided as examples only to help the reader understand this disclosure. They are not intended and should not be construed as limiting the scope of this disclosure in any way. Although certain embodiments and examples have been provided, it will be apparent to those skilled in the art, based on the content disclosed herein, that changes can be made to the illustrated embodiments and examples, and other similar implementations based on the technical concept of this disclosure can be adopted without departing from the scope of this disclosure, and these modifications and modifications are also within the protection scope of the embodiments of this disclosure.

Claims

1. A method executed by a network node, comprising: Obtain the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device; Multiple signal paths of the non-line-of-sight channel are determined based on the channel impulse response; Multiple clusters are determined based on the multiple signal paths; At least one cluster is selected from the plurality of clusters based on the power information and / or angle of arrival information corresponding to the plurality of clusters; The target object is located based on the at least one cluster.

2. The method according to claim 1, wherein, Selecting at least one cluster from the plurality of clusters based on power information and / or angle of arrival information corresponding to the plurality of clusters, including: Based on the power information and / or angle of arrival information corresponding to the plurality of clusters, power distribution information and / or angle distribution information corresponding to the plurality of clusters are obtained, wherein the power distribution information is used to represent the spatial distribution of power corresponding to the plurality of clusters, and the angle distribution information is used to represent the spatial distribution of the angle of arrival corresponding to the plurality of clusters; At least one cluster is selected from the plurality of clusters based on the power distribution information and / or the angle distribution information.

3. The method according to claim 2, wherein, Based on the power information and / or angle of arrival information corresponding to the plurality of clusters, obtain the power distribution information and / or angle distribution information corresponding to the plurality of clusters, including: Based on the arrival angle information, a spatial grid corresponding to the target object is constructed, wherein the arrival angle information includes a horizontal arrival angle and a vertical arrival angle, and the coordinates of the horizontal and vertical directions of the spatial grid are the horizontal arrival angle and the vertical arrival angle, respectively. The arrival angle information is mapped to the spatial grid to obtain the angle distribution information, and / or the power information is mapped to the spatial grid to obtain the power distribution information.

4. The method according to claim 3, wherein, Constructing a spatial grid corresponding to the target object based on the arrival angle information includes: Estimate the horizontal and vertical angles of arrival for the most powerful signal path in each of the multiple clusters; The coordinate ranges of the spatial grid in the horizontal and vertical directions are determined based on the maximum and minimum values ​​of the horizontal and vertical angles of arrival corresponding to the strongest signal path in each of the multiple clusters. The resolution of the spatial grid in the horizontal and vertical directions is determined based on the coordinate range and the number of the plurality of clusters.

5. The method according to claim 3, wherein, Mapping the power information to the spatial grid to obtain the power distribution information includes: The power of the strongest signal path in each of the multiple clusters is mapped to the spatial grid to obtain the power distribution information.

6. The method according to claim 3, wherein, Selecting at least one cluster from the plurality of clusters based on the power distribution information and / or the angle distribution information includes: The angle distribution information is convolved in the spatial grid using the first convolution kernel to obtain the angle distribution density; The power distribution information is convolved in the spatial grid using a second convolution kernel to obtain the power distribution density. At least one cluster is selected from the plurality of clusters based on the angular distribution density and the power distribution density.

7. The method according to claim 6, wherein, The power distribution information is convolved in the spatial grid using a second convolution kernel to obtain the power distribution density, including: The second convolution kernel is adjusted based on the power distribution information in the convolution region of the second convolution kernel to obtain the adjusted second convolution kernel; The power distribution information in the convolution region is convolved using the adjusted second convolution kernel to obtain the power distribution density.

8. The method according to claim 6, wherein, Selecting at least one cluster from the plurality of clusters based on the angular distribution density and the power distribution density includes: The angular power spectral density is obtained based on the angular distribution density and the power distribution density; At least one cluster is selected from the plurality of clusters based on the angular power spectral density.

9. The method according to claim 8, wherein, Selecting at least one cluster from the plurality of clusters based on the angular power spectral density includes: Perform data mapping on the angular power spectral density to obtain mapped data within a preset range; A first threshold is determined based on the mapping data; At least one convolutional region in the spatial grid is selected based on a first threshold. Select the cluster with the strongest power from each of the at least one convolutional regions.

10. The method according to claim 6, wherein, The sizes of the first convolution kernel and the second convolution kernel are determined based on the scene in which the target object is located.

11. The method according to claim 2, wherein, The power corresponding to the plurality of clusters includes the cross-correlation power corresponding to each signal path included in the plurality of clusters, wherein the cross-correlation power is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window.

12. The method according to claim 1, wherein, Multiple clusters are determined based on the multiple signal paths, including: The plurality of clusters are determined based on the power information and time delay information corresponding to each of the plurality of signal paths, wherein the power information includes the cross-correlation power corresponding to each signal path, and the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window.

13. The method according to claim 12, wherein, The multiple clusters are determined based on the power information and time delay information of each signal path in the multiple signal paths, including: Based on the changing trend of the cross-correlation power of each signal path in the plurality of signal paths and the time delay information, the power peak value in the time domain is detected; The plurality of clusters are determined based on the power peak values.

14. The method according to claim 1, wherein, Determining multiple signal paths of the non-line-of-sight channel based on the channel impulse response includes: Based on the channel impulse response, the cross-correlation power of each signal path in the signal path set of the non-line-of-sight channel is determined, wherein the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window. Based on the cross-correlation power, at least one signal path is removed from the set of signal paths in the non-line-of-sight channel to determine multiple signal paths.

15. The method according to claim 14, wherein, Based on the cross-correlation power, at least one signal path is removed from the set of signal paths in the non-line-of-sight channel to determine multiple signal paths, including: Based on the cross-correlation power, determine the cross-correlation power threshold; Based on the cross-correlation power threshold, at least one signal path is removed from the signal path set of the non-line-of-sight channel, and the remaining signal paths after removing the at least one signal path are taken as the plurality of signal paths.

16. The method according to claim 15, wherein, The signal receiving device includes multiple antennas, and the cross-correlation power includes the cross-correlation power of each signal path corresponding to each of the multiple antennas. Determining the cross-correlation power threshold based on the cross-correlation power includes: The cross-correlation power of each signal path corresponding to the multiple antennas is averaged and arranged from high to low to obtain an average cross-correlation power queue. The average cross-correlation power corresponding to the starting position of the effective path in the average cross-correlation power queue is determined based on the predetermined time window and the cross-correlation power of each signal path corresponding to each antenna, and is used as the cross-correlation power threshold.

17. The method according to claim 1, wherein, Locating the target object based on the at least one cluster includes: Based on the power of the signal path, at least one first signal path is selected from each of the at least one cluster; Based on the relevant information of the signal paths selected from each cluster, the localization result is obtained using an artificial intelligence (AI) model; Based on the difference between the current positioning result and the positioning result obtained by the AI ​​model last time, determine whether to output the positioning result. If yes, obtain the location information of the target object based on the current positioning result. Otherwise, select at least one second signal path from each cluster based on the power of the signal path, and execute the step of obtaining the positioning result by using the AI ​​model based on the relevant information of the signal path selected from each cluster. The second signal path includes the first signal path.

18. The method according to claim 17, wherein, After selecting at least one second signal path from each cluster, the localization results are obtained using an AI model based on the relevant information of the signal paths selected from each cluster, including: Perform a first attention operation on the feature data corresponding to the relevant information of signal paths in different clusters to obtain the first attention result; A second attention operation is performed between the feature data corresponding to the relevant information of the signal paths in the same cluster to obtain the second attention result; Based on the first attention result and the second attention result, the localization result is obtained.

19. The method of claim 17, wherein, The relevant information includes angle information, power information, and time delay information corresponding to the signal path, obtained based on the channel impulse response corresponding to the signal path.

20. The method according to claim 19, wherein, The step of obtaining the location information of the target object based on the current location result includes: Perform a validity verification on the current location result; If the current location result is valid, then the location information is obtained based on the current location result; If the current location result is invalid, the location information is obtained based on the previous location result of the target object.

21. A method executed by a network node, comprising: Obtain the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device; Based on the channel impulse response, the cross-correlation power of each signal path in the signal path set of the non-line-of-sight channel is determined, wherein the cross-correlation power of each signal path is determined based on the cross-correlation operation between the channel impulse responses of each signal path at different times within a predetermined time window. Based on the cross-correlation power, at least one signal path is removed from the set of signal paths in the non-line-of-sight channel to determine multiple signal paths; The target object is located based on the multiple signal paths.

22. The method according to claim 21, wherein, Based on the cross-correlation power, at least one signal path is removed from the set of signal paths in the non-line-of-sight channel to determine multiple signal paths, including: Based on the cross-correlation power, determine the cross-correlation power threshold; Based on the cross-correlation power threshold, at least one signal path is removed from the signal path set of the non-line-of-sight channel, and the remaining signal paths after removing the at least one signal path are taken as the plurality of signal paths.

23. The method according to claim 22, wherein, The signal receiving device includes multiple antennas, and the cross-correlation power includes the cross-correlation power of each signal path corresponding to each of the multiple antennas. Based on the cross-correlation power, the cross-correlation power threshold is determined, including: The cross-correlation power of each signal path corresponding to the multiple antennas is averaged and arranged from high to low to obtain an average cross-correlation power queue. The average cross-correlation power corresponding to the starting position of the effective path in the average cross-correlation power queue is determined based on the predetermined time window and the cross-correlation power of each signal path corresponding to each antenna, and is used as the cross-correlation power threshold.

24. A method executed by a network node, comprising: Obtain the channel impulse response of the non-line-of-sight channel between the target object and the signal receiving device; Multiple signal paths of the non-line-of-sight channel are determined based on the channel impulse response; Multiple clusters are determined based on the multiple signal paths; Based on the power of the signal path, at least one first signal path is selected from each of the plurality of clusters; Based on the relevant information of the signal paths selected from each cluster, the localization result is obtained using an artificial intelligence (AI) model; Based on the difference between the current positioning result and the positioning result obtained by the AI ​​model last time, determine whether to output the positioning result. If yes, obtain the location information of the target object based on the current positioning result. Otherwise, select at least one second signal path from each cluster based on the power of the signal path, and execute the step of obtaining the positioning result by using the AI ​​model based on the relevant information of the signal path selected from each cluster. The second signal path includes the first signal path.

25. The method according to claim 24, wherein, After selecting at least one second signal path from each cluster, the localization results are obtained using an AI model based on the relevant information of the signal paths selected from each cluster, including: Perform a first attention operation on the feature data corresponding to the relevant information of signal paths in different clusters to obtain the first attention result; A second attention operation is performed between the feature data corresponding to the relevant information of the signal paths in the same cluster to obtain the second attention result; Based on the first attention result and the second attention result, the localization result is obtained.

26. The method of claim 25, wherein, The relevant information includes angle information, power information, and time delay information corresponding to the signal path, obtained based on the channel impulse response corresponding to the signal path.

27. A network node, comprising: transceiver; A processor, coupled to the transceiver and configured to perform the method as claimed in any one of claims 1 to 26.

28. An electronic device comprising: At least one processor; At least one transceiver and / or at least one memory coupled to the at least one processor The at least one processor is configured to perform the method as claimed in any one of claims 1 to 26.

29. A computer-readable storage medium storing a computer program or instructions that, when executed by at least one processor, cause the at least one processor to perform the method as claimed in any one of claims 1 to 26.

Images