Positioning method, apparatus, device, medium, and program product

Abstract

This application provides a positioning method, apparatus, device, medium, and program product, relating to the field of communication technology, for improving positioning accuracy. The specific technical solution is as follows: RSS measurements of a target user are acquired from at least one measurement point via an unmanned aerial vehicle (UAV) network; based on a first measurement point where the target user's RSS measurement value is measured, the similarity between the RSS measurement value of the first measurement point and the predicted RSS values ​​of each candidate location in the CKM database is determined; based on a second measurement point where the target user's RSS measurement value is not measured, the probability that the target user exists at the second measurement point is determined; based on the similarity between the first RSS measurement value and the predicted RSS values ​​of each candidate location in the CKM database, and the probability that the target user exists at the second measurement point, a positioning probability map is constructed; the first RSS measurement value and the positioning probability map are input into a local communication graph neural network to obtain the target user's location information. This application is applicable to scenarios requiring high-precision positioning in complex terrain.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a positioning method, apparatus, device, medium, and program product. Background Technology

[0002] In the complex scenarios of mountain disaster relief, locating the position of disaster victims can effectively ensure the smooth progress of disaster relief operations.

[0003] Currently, the location of disaster victims can be determined using fingerprint-based wireless network positioning methods. Specifically, in fingerprint-based wireless network positioning methods, a Channel Knowledge Map (CKM) database can be formed by pre-collecting or simulating Channel State Information (CSI) within the service area, and then matched with real-time observations to infer the user's location.

[0004] However, the aforementioned fingerprint-based wireless network positioning methods mainly focus on planar information. In mountainous environments with large terrain undulations and significant three-dimensional signal propagation characteristics, signal propagation is affected by blockage and multipath effects, resulting in highly similar Received Signal Strength (RSS) fingerprints from multiple different geographical locations. This makes it difficult for fingerprint-based wireless network positioning methods to accurately distinguish between them, leading to poor positioning accuracy. Summary of the Invention

[0005] This application provides a positioning method, apparatus, device, medium, and program product for improving positioning accuracy.

[0006] In a first aspect, embodiments of this application provide a positioning method, the method comprising: acquiring RSS measurements of a target user from at least one measurement point via an Unmanned Aerial Vehicle (UAV) network; when at least one measurement point includes a first measurement point, determining the similarity between the first RSS measurement and the predicted RSS values ​​of each candidate location in a CKM database, wherein the first measurement point is the measurement point from which the RSS measurement of the target user has been acquired, and the first RSS measurement is the RSS measurement obtained from the first measurement point, and each predicted RSS value in the CKM database is determined based on two-dimensional location information and measurement location information of at least one historical user; when at least one measurement point includes a second measurement point, determining the probability that the target user exists at the location corresponding to the second measurement point, wherein the second measurement point is the measurement point from which the RSS measurement of the target user has not been acquired; constructing a positioning probability map based on the similarity between the first RSS measurement and the predicted RSS values ​​of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point, the positioning probability map being used to indicate the location of the target user; and inputting the first RSS measurement and the positioning probability map into a local communication graph neural network to obtain the location information of the target user.

[0007] The technical solution provided in this application offers at least the following advantages: By constructing a location probability map, observation points receiving signals can be used as positive evidence, and observation points not receiving signals can be combined as negative evidence to eliminate conflicting areas, thereby accurately narrowing down the possible location range of the user; furthermore, the terrain elevation information used in the construction of this location probability map can adapt to the three-dimensional environment of mountainous areas. This improves the accuracy of positioning.

[0008] One possible implementation involves constructing a location probability map based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that a target user exists at the location corresponding to the second measurement point. This includes: determining a likelihood function based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that a target user exists at the location corresponding to the second measurement point; and constructing a location probability map based on the likelihood function.

[0009] Another possible implementation involves determining the similarity between the first RSS measurement and the predicted RSS value for each candidate location in the CKM database, including: determining the normalized cosine similarity between the first RSS measurement and the predicted RSS value for each candidate location based on the first RSS measurement and the predicted RSS value for each candidate location in the CKM database; and determining the similarity between the first RSS measurement and the predicted RSS value for each candidate location based on each normalized cosine similarity and an acceptable difference threshold.

[0010] Another possible implementation is that determining the probability that a target user exists at the location corresponding to the second measurement point includes: determining the probability that a target user exists at the location corresponding to the second measurement point based on the location corresponding to the second measurement point, the receiver detection threshold, and the probability of anomaly detection failure.

[0011] Another possible implementation is that the aforementioned local communication graph neural network includes a graph neural network and a convolutional network; the above-mentioned inputting the first RSS measurement value and the positioning probability map into the local communication graph neural network to obtain the target user's location information includes: inputting the first RSS measurement value into the graph neural network to obtain first feature information, which is a robust representation information with permutation invariance to receiver configuration extracted from the first RSS measurement value; inputting the positioning probability map into the convolutional network to obtain the target user's location feature information; and performing a fusion operation on the first feature information and the location feature information to obtain the target user's location information.

[0012] Secondly, embodiments of this application provide a positioning device, including: an acquisition module, a determination module, a construction module, and an execution module. The acquisition module is used to acquire the RSS measurement value of the target user from at least one measurement point via a drone network. The determination module is used to determine the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database when at least one measurement point includes a first measurement point, wherein the first measurement point is the measurement point where the RSS measurement value of the target user has been acquired, and the first RSS measurement value is the RSS measurement value acquired from the first measurement point, and each RSS prediction value in the CKM database is determined based on the two-dimensional location information and measurement location information of at least one historical user. The determination module is also used to determine the probability that the target user exists at the location corresponding to the second measurement point when at least one measurement point includes a second measurement point, wherein the second measurement point is the measurement point where the RSS measurement value of the target user has not been acquired. The construction module is used to construct a positioning probability map based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point, wherein the positioning probability map is used to indicate the location of the target user. The execution module is used to input the first RSS measurement value and the positioning probability map into a local communication graph neural network to obtain the location information of the target user.

[0013] One possible implementation is that the aforementioned construction module is specifically used to: determine the likelihood function based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point; and construct a location probability map based on the likelihood function.

[0014] Another possible implementation, the aforementioned determining module, is specifically used to: determine the normalized cosine similarity between the first RSS measurement value and the RSS prediction value of each candidate position based on the first RSS measurement value and the RSS prediction value of each candidate position in the CKM database; and determine the similarity between the first RSS measurement value and the RSS prediction value of each candidate position based on each normalized cosine similarity and an acceptable difference threshold.

[0015] Another possible implementation is that the aforementioned determining module is specifically used to determine the probability that a target user exists at the location corresponding to the second measurement point, based on the location corresponding to the second measurement point, the receiver detection threshold, and the probability of anomaly detection failure.

[0016] Another possible implementation is that the aforementioned local communication graph neural network includes a graph neural network and a convolutional network; the execution module is specifically used to: input the first RSS measurement value into the graph neural network to obtain first feature information, the first feature information being robust representation information with permutation invariance to receiver configuration extracted from the first RSS measurement value; input the positioning probability map into the convolutional network to obtain the target user's location feature information; and perform a fusion operation on the first feature information and the location feature information to obtain the target user's location information.

[0017] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.

[0018] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.

[0019] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.

[0020] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.

[0021] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description

[0022] Figure 1 A schematic diagram of a network architecture for a positioning method application provided in an embodiment of this application;

[0023] Figure 2 A flowchart illustrating a positioning method provided in an embodiment of this application;

[0024] Figure 3 A schematic diagram illustrating the dual-evidence input construction method provided in an embodiment of this application;

[0025] Figure 4 A flowchart illustrating another positioning method provided in an embodiment of this application;

[0026] Figure 5 A flowchart illustrating another positioning method provided in an embodiment of this application;

[0027] Figure 6 A flowchart illustrating another positioning method provided in an embodiment of this application;

[0028] Figure 7 A schematic diagram of the structure of the Local Communication Graph Neural Network (LocCGNN) provided in the embodiments of this application;

[0029] Figure 8 A flowchart illustrating another positioning method provided in an embodiment of this application;

[0030] Figure 9 A flowchart illustrating the implementation process of a positioning method provided in this application embodiment;

[0031] Figure 10 A schematic diagram illustrating the impact of negative evidence provided in the embodiments of this application on the results;

[0032] Figure 11 This is a schematic diagram of the structure of a positioning device provided in an embodiment of this application;

[0033] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0035] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0036] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."

[0037] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0038] The positioning method, apparatus, device, medium, and program product provided in this application embodiment can be applied to scenarios requiring high-precision positioning in complex terrain.

[0039] In existing technologies, locating the position of disaster victims is crucial for ensuring the smooth operation of disaster relief efforts in complex mountain disaster relief scenarios. However, mountains and vegetation severely obstruct radio signals and cause multipath effects. This obstruction not only drastically reduces the signal coverage of traditional emergency base stations but also severely interferes with the reception of Global Navigation Satellite System (GNSS) signals, significantly reducing positioning accuracy or even causing complete failure, making it difficult to achieve reliable communication and positioning coverage across the entire disaster area. Deploying emergency base stations using highly flexible and rapidly deployable drones to establish a real-time positioning network based on communication signals is an effective technological approach to meet this emergency positioning requirement.

[0040] Existing wireless network positioning methods are mainly divided into two categories: measurement-based and fingerprint-based. Measurement-based positioning techniques, such as Time Difference of Arrival (TDOA) positioning, calculate the distance difference by measuring the time difference of signal arrival at different base stations and determine the location based on the intersection of multiple hyperbolas; or Angle of Arrival (AOA) positioning, which measures the angle of arrival of signals through an antenna array and relies on the intersection of multiple directional lines for location. While these methods offer high positioning accuracy, a common drawback is the need for a precise ranging protocol and time synchronization mechanism between the user and the base station. Another measurement-based method utilizes RSS (Reverse Segmentation Path Loss), which estimates distance and performs trilateration by inverting the logarithmic distance path loss model of the signal. Furthermore, fingerprint-based positioning methods rely on a pre-collected signal strength database (fingerprint database) for pattern matching, requiring no additional communication protocols.

[0041] In recent years, location methods based on Channel State Information (CSI) have made significant progress. These methods form a CSI database by pre-collecting or simulating Channel State Information (CSI) within the service area and then matching it with real-time observations to infer the user's location. However, existing CSI-based location schemes, such as those using models like U-Net to process 2D channel maps, primarily focus on planar information and fail to effectively integrate crucial elevation data. This is a major limitation for their application in mountainous environments with significant terrain undulations and pronounced three-dimensional signal propagation characteristics. Meanwhile, Graph Neural Networks (GNNs) have shown advantages in the location field by fusing measurement data across receivers and reducing the impact of multipath effects.

[0042] Existing wireless network positioning technologies suffer from key technical shortcomings when applied to complex scenarios such as mountain disaster relief. First, signal strength fingerprint-based positioning methods face severe positioning ambiguity issues in the complex and variable terrain of mountainous areas. Due to signal propagation obstruction and multipath effects, multiple different geographical locations may generate highly similar RSS fingerprints, making it difficult for traditional fingerprint matching and CKM-based positioning methods to accurately distinguish them, severely impacting positioning reliability. Second, existing CKM positioning methods, such as those based on 2D convolutional networks, primarily focus on planar information and fail to fully utilize and integrate elevation data. This makes it unable to effectively model the complex three-dimensional signal propagation characteristics of mountainous environments, hindering high-precision positioning in areas with drastically changing terrain. Furthermore, when fusing real-time observation data from multiple UAV receivers, existing solutions lack an efficient and robust mechanism to extract permutation-invariant representations that are insensitive to receiver configuration changes, further limiting the system's accuracy and generalization capabilities.

[0043] To address the shortcomings of existing technologies in mountainous positioning, such as positioning ambiguity and poor adaptability to 3D environments, the primary objective of this invention is to solve the ambiguity problem in fingerprint positioning in mountainous areas. This invention aims to construct a user positioning probability map as input data using a dual-evidence method. Observation points receiving signals are used as positive evidence, while observation points without received signals are used as negative evidence to eliminate contradictory regions, thereby accurately narrowing down the possible user location range. Secondly, this invention aims to provide a deep learning positioning network, LocCGNN, which can effectively process CKM data containing terrain elevation information, thus overcoming the limitations of existing CKM methods in adapting to the 3D environment of mountainous areas. Finally, by introducing a GNN into LocCGNN, this invention aims to extract the permutation-invariant representation of real-time received RSS observations and effectively fuse it with spatial features in the probability map, thereby improving the robustness of the positioning system to changes in observation configuration and the final coordinate output accuracy.

[0044] To address the aforementioned technical problems, embodiments of this application provide a positioning method, apparatus, device, medium, and program product. By constructing a positioning probability map, observation points receiving signals can be used as positive evidence, and observation points not receiving signals can be combined as negative evidence to eliminate contradictory areas, thereby accurately narrowing down the possible location range of the user. Furthermore, the terrain elevation information used in the construction of this positioning probability map can adapt to the three-dimensional environment of mountainous areas. This improves positioning accuracy.

[0045] The positioning methods, apparatus, devices, media, and program products provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0046] Figure 1 The network architecture for a positioning method provided in an embodiment of this application is illustrated. For example... Figure 1 As shown, the network architecture includes a positioning device 101 and a terminal device 102. The positioning device 101 and the terminal device 102 are interconnected.

[0047] In some embodiments, the positioning device 101 may be a server, a computer, or a processor or processing unit within a server or computer. The server may be a single server or a server cluster comprising multiple servers. It should be noted that the specific device form of the positioning device 101 is not limited in the embodiments of this application. Figure 1 The example shown is a single server using the positioning device 101.

[0048] In some embodiments, the terminal device may be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, personal computer (PC), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and the embodiments of this application do not specifically limit it. Figure 1 The example shown is a mobile phone, with terminal device 102 as an example.

[0049] In some embodiments, the terminal device 102 continuously transmits wireless signals as a signal source; the positioning device 101 controls its drone network to fly to various measurement points, listen to and measure the signals from the terminal device 102, and converts successfully received signals into RSS measurement values, while the state of not receiving signals is also recorded; the positioning device 101 collects all measurement information and constructs a positioning probability map based on the channel knowledge map, and then processes and calculates it through a local communication graph neural network to finally obtain the location coordinates of the terminal device 102.

[0050] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0051] See Figure 2 This is a flowchart illustrating a positioning method provided in an embodiment of this application. Figure 2 As shown, the positioning method provided in this application embodiment can be implemented by the above-mentioned positioning device, specifically including the following steps 201 to 205.

[0052] Step 201: The positioning device obtains the target user's RSS measurement value from at least one measurement point through the UAV network.

[0053] In some embodiments, the aforementioned drone network refers to a rapidly deployable aerial mobile measurement platform consisting of multiple drones, responsible for detecting and measuring radio signals from the ground at different locations over the service area.

[0054] In some embodiments, the above-mentioned measurement point refers to a specific location in three-dimensional space where a single UAV is positioned to perform signal measurements.

[0055] In some embodiments, the target user refers to a disaster victim or their emergency beacon device that needs to be located in a mountainous disaster relief scenario. The precise two-dimensional horizontal position of the target user is unknown and needs to be estimated using this positioning system.

[0056] In some embodiments, the RSS measurement value mentioned above refers to the signal power from the target user equipment that is successfully captured and measured by the radio receiver on the UAV at a certain measurement point.

[0057] In some embodiments, the drone network can fly to a series of pre-defined measurement point locations over the service area according to pre-planned or real-time commands. While hovering at each measurement point, the receiver on the drone attempts to listen for and receive wireless signals transmitted by the target user equipment. For successfully received signals, the receiver measures their power intensity to obtain a specific RSS value; for measurement points that fail to receive a valid signal, a special "invalid" or "no signal" state is recorded. Finally, all the RSS observation data collected from all measurement points are aggregated, including valid values ​​and invalid states, and transmitted uniformly to the backend positioning device via a data link as input data for the entire positioning process.

[0058] This application embodiment considers a mountainous positioning scenario. The service area is... The ground user set is .user The unknown 2D position is Topographic elevation It can be determined from topographic data. Within a short time window. Inside, all users remain stationary.

[0059] In some embodiments, the drone network can be derived from... There are 1 measurement point, indexed as . The location is Acquire users RSS To ensure flight safety, each measurement point is maintained at a fixed altitude above the terrain. .

[0060] Step 202: In cases where at least one measurement point includes the first measurement point, the positioning device determines the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database.

[0061] In some embodiments, the first measurement point is the measurement point where the RSS measurement value of the target user is obtained.

[0062] In some embodiments, the first measurement point refers to the location of the drone that successfully detects and acquires the wireless signal emitted by the target user equipment in the drone network, and is a measurement point that provides effective observation data.

[0063] In some embodiments, the first RSS measurement value is the RSS measurement value obtained from the first measurement point.

[0064] In some embodiments, each RSS prediction value in the CKM database is determined based on two-dimensional location information and measurement location information of at least one historical user.

[0065] In some embodiments, the aforementioned CKM database refers to a pre-built radio environment fingerprint database covering a service area, storing the predicted long-term average received signal strength value for each candidate location and each possible combination of measurement point locations within the area. This database can be generated through historical measurement data or radio propagation model simulation.

[0066] In some embodiments, the above-mentioned candidate locations refer to the locations of all possible target users within the service area.

[0067] In some embodiments, the aforementioned RSS prediction value refers to the signal strength value that should theoretically be received at a certain measurement point for a specific candidate location, obtained from a query of the CKM database. It represents the predicted attenuation of the signal propagating from the candidate location to the measurement point under ideal long-term averaging conditions.

[0068] In some embodiments, the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database refers to the degree of consistency or matching between the real-time measured first RSS measurement value and the RSS prediction value obtained from the CKM database for each candidate location, and is used to assess how likely the target user is to be located at that candidate location.

[0069] In some embodiments, CKM serves as a fingerprint database representing any user location. and measurement location The long-term average RSS, denoted as the continuous field Real-time RSS measurements It can be represented as: in, It measures residuals, including small-scale fading and shadowing effects.

[0070] In some embodiments, to overcome the ambiguity of UAV positioning in complex mountainous terrain, we propose a dual-evidence input construction method to build a user positioning probability graph. .like Figure 3As shown, drones that successfully receive signals provide positive evidence, identifying potential areas through signal fingerprinting and assigning them high probability values. Drones that fail to receive valid signals provide negative evidence, suggesting that the target is outside their reception range, thus eliminating contradictory areas.

[0071] In some embodiments, combined with Figure 2 ,like Figure 4 As shown, step 202 above can be implemented through steps 202a and 202b.

[0072] Step 202a: When at least one measurement point includes the first measurement point, the positioning device determines the normalized cosine similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database.

[0073] In some embodiments, the normalized cosine similarity mentioned above refers to an index used to measure the directional consistency between two vectors, with a value range of [-1, 1]. In this application embodiment, the normalized cosine similarity can map the numerical difference between a real-time first RSS measurement value and an RSS prediction value queried from a CKM database into a probability weight representing the degree of similarity between the two. The core idea is that the closer the measurement value is to the prediction value, the closer its normalized cosine similarity is to 1, indicating that the candidate position is more likely to be the target user's true position.

[0074] Step 202b: The positioning device determines the similarity between the first RSS measurement and the RSS prediction of each candidate location based on each normalized cosine similarity and an acceptable difference threshold.

[0075] In some embodiments, the above-mentioned acceptable difference threshold may also be referred to as the maximum acceptable difference threshold, which is a preset threshold value, denoted as . The acceptable difference threshold is the maximum absolute deviation allowed between real-time measurements and database predictions during location determination.

[0076] In some embodiments, the aforementioned similarity specifically refers to the final similarity weight after threshold determination, denoted as . It combines the matching strength represented by normalized cosine similarity with the reasonableness range defined by an acceptable difference threshold, and outputs a deterministic evidence value for constructing a probabilistic graph.

[0077] In some embodiments, for receiving a valid RSS Measurement points We use a normalized similarity metric to calculate the real-time measured value and the CKM predicted value at each candidate location. The similarity at each location is calculated using the following formula:

[0078]

[0079] in, It is the normalized cosine similarity. It is the maximum acceptable difference threshold. It is an indicator function.

[0080] In some embodiments, for each candidate location, the positioning device can extract the predicted RSS value from that location to the current first measurement point from the CKM database. Then, the actual first RSS measurement value and the predicted value are treated as two one-dimensional vectors, and their cosine similarity is calculated. Since it is a one-dimensional scalar, its calculation simplifies to measuring the numerical closeness between the two, and is converted into a weight coefficient between 0 and 1 through normalization, denoted as . This coefficient represents the preliminary similarity score, before it has been assessed using a threshold.

[0081] In some embodiments, the positioning device obtains the normalized cosine similarity of each candidate location. Then, it can be compared with an acceptable difference threshold. Both are input into a decision function to calculate the absolute difference between the measured and predicted values. If this absolute difference is less than or equal to the threshold Then the positioning device can determine that the candidate position matches the observation of this first measurement point, and its final similarity is... The value is equal to the previously calculated value. If the absolute difference is greater than the threshold If the candidate location does not match the observation, its final similarity is set to 0 using the indicator function I(⋅). Ultimately, the localization device can generate a threshold-filtered similarity score for each candidate location based on the observation at the first measurement point.

[0082] In this way, by performing standardized similarity calculations and threshold filtering on successfully received signals, signal measurements are transformed into comparable probability weights. Unreliable data caused by channel fluctuations or measurement errors are filtered out through the threshold mechanism, thereby ensuring the accuracy of the positive evidence portion of the constructed positioning probability map. This enhances the positioning system's ability to identify and utilize effective signals in complex terrain and lays the groundwork for subsequent high-precision location estimation.

[0083] Step 203: If at least one measurement point includes the second measurement point, the positioning device determines the probability that the target user exists at the location corresponding to the second measurement point.

[0084] In some embodiments, the second measurement point is a measurement point where no RSS measurement value of the target user was obtained.

[0085] In some embodiments, for each such second measurement point, the positioning device may calculate a weighted probability, which represents the likelihood that the observation result of "no signal received" will occur if the target user is located at a candidate location q, and is used to subsequently construct a complete positioning probability map.

[0086] In some embodiments, combined with Figure 2 ,like Figure 5 As shown, step 203 above can be implemented through step 203a as follows.

[0087] Step 203a: When at least one measurement point includes the second measurement point, the positioning device determines the probability that the target user exists at the location corresponding to the second measurement point based on the location corresponding to the second measurement point, the receiver detection threshold, and the probability of anomaly detection failure.

[0088] In some embodiments, the receiver detection threshold is the minimum signal strength threshold required for the receiver to reliably detect and decode signals, typically denoted as [missing information]. If the signal strength is below this threshold, the receiver will report "No valid signal detected".

[0089] In some embodiments, the probability of the above-mentioned anomaly detection failure is a small probability value, such as 0.1, usually denoted as ϵ, used to model uncertainties in the actual system, such as occasional detection errors by the receiver, missed detections due to momentary signal obstruction or interference, and minor errors in the channel prediction model. Introducing ϵ can prevent the model from being too absolute and improve the robustness of the system.

[0090] In some embodiments, for measurement points that fail to receive a valid RSS... The probability calculation formula is:

[0091]

[0092] in It is the receiver detection threshold. This takes into account the low probability of anomaly detection failure.

[0093] In some embodiments, for each second measurement point For each candidate position q, perform the following steps:

[0094] 1. Query predicted value: From the CKM database, obtain the signal propagation value at candidate position q to the second measurement point. Predicted signal strength .

[0095] 2. Apply decision rules:

[0096] (1) If This means that, according to the prior knowledge provided by CKM, if the user is indeed at location q, then the signal propagating to the drone s_k should be very weak, i.e., below the threshold that the receiver can detect. Therefore, the observation that the drone "did not receive a signal" is very reasonable and consistent with the prior knowledge provided by CKM. Therefore, a high probability value (1-ϵ), such as 0.9, is assigned to this location.

[0097] (2) If This means that, based on the prior knowledge provided by CKM, if the user is at location q, the drone... A sufficiently strong signal should have been received. However, the actual observation showed that "no signal was received," which creates a contradiction. Therefore, the probability of this observation occurring is very low, and a low probability value ϵ, such as 0.1, is assigned to this location.

[0098] In this way, by assigning extremely low probabilities to candidate locations that "should have received a signal but did not," the positioning device effectively eliminates these unreasonable areas on the probability map, narrowing the positioning range. This complements positive evidence and together constitutes the "dual evidence" positioning logic, improving the accuracy of positioning.

[0099] Step 204: The positioning device constructs a positioning probability map based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point.

[0100] In some embodiments, the location probability map described above is used to indicate the location of the target user.

[0101] In some embodiments, the above-described localization probability map is a set of probability values ​​defined at all candidate locations q, which can be used... This indicates that it was generated by the dual-evidence input construction method.

[0102] A probability distribution visualization covering the entire area to be located is presented, where each point in the graph corresponds to a candidate location and is assigned a probability value. This probability value intuitively indicates the likelihood that the target user is located at that location.

[0103] In some embodiments, combined with Figure 2 ,like Figure 6 As shown, step 204 above can be implemented through steps 204a and 204b.

[0104] Step 204a: The positioning device determines the likelihood function based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point.

[0105] In some embodiments, the likelihood function mentioned above refers to the conditional probability or likelihood value of the hypothesis that "the target user is located at position q" given all current observation data, i.e., the signal conditions measured by all UAVs. It is not a probability distribution, but its magnitude directly reflects the relative reasonableness of different positions q.

[0106] Step 204b: The positioning device constructs a positioning probability map based on the likelihood function.

[0107] In some embodiments, the complete likelihood function combining positive and negative evidence is:

[0108]

[0109] In this way, by using the likelihood function, positive and negative evidence are fused together using the probability multiplication rule, which not only allows us to find possible regions from successfully received signals, but also eliminates contradictory regions from failed signals, thus improving positioning accuracy.

[0110] Step 205: The positioning device inputs the first RSS measurement value and the positioning probability map into the local communication graph neural network to obtain the location information of the target user.

[0111] In some embodiments, the Local Communication Graph Neural Network (LocCGNN) described above is a deep learning model designed specifically for this localization task. Its core feature is that it includes two branches: GNN and Convolutional Neural Network (CNN), which process input data of different forms respectively.

[0112] In some embodiments, the positioning device can combine the first RSS measurement value and the positioning probability. Figure 1 The two types of information are input into a pre-trained LocCGNN model. The model performs in-depth feature extraction and fusion on these two types of information and directly outputs a coordinate as the location information of the target user.

[0113] In some embodiments, the user localization problem is modeled as a supervised learning task, and a training function is used. Predicting user location by minimizing the expected squared location error:

[0114]

[0115] The model inputs include the measured RSSI values. , corresponding measurement point locations and CKM prior .Model It is implemented as a permutation-invariant set function, enabling it to handle any number and order of measurement points.

[0116] In some embodiments, LocCGNN employs a dual-branch architecture, including a measurement set encoder and a CKM prior encoder, such as... Figure 7 As shown, where,

[0117] Measurement Branch (GNN): Used to receive features from multiple UAV measurement points, construct a star diagram, and generate an order-invariant central embedding through two layers of GAT processing to summarize the measurement set.

[0118] Environment branch (CNN): Used to encode the target probability map using a convolutional network to obtain a priori representation of the environment.

[0119] In some embodiments, the two embeddings are fused through a multiplication gating mechanism and then fed into a multilayer perceptron (MLP) to regress to normalized coordinates, and finally mapped to a physical or raster coordinate system through a known affine transformation.

[0120] In some embodiments, the aforementioned local communication graph neural network includes a graph neural network and a convolutional network. (Combined) Figure 2 ,like Figure 8 As shown, step 205 above can be implemented through steps 205a to 205c.

[0121] Step 205a: The positioning device inputs the first RSS measurement value into the graph neural network to obtain the first feature information.

[0122] In some embodiments, the first feature information is robust representation information with permutation invariance to receiver configuration extracted from the first RSS measurement value.

[0123] In some embodiments, the aforementioned permutation invariance means that regardless of the order in which the data reported by multiple UAV measurement points are arranged, the first feature information extracted by the GNN is the same. This is a key attribute for processing ensemble data and is guaranteed by the star graph structure and the aggregation mechanism of the GNN.

[0124] In some embodiments, the GNN described above is a neural network specifically designed for processing graph-structured data.

[0125] In some embodiments, the first feature information mentioned above refers to the embedding of the center node of the GNN branch output.

[0126] In some embodiments, input diagram Using a star topology, including One measurement node that received a valid RSS and one central node. Each edge connects the measurement node and the center node bidirectionally. Input node feature matrix. Contains the feature vectors of all nodes, with the center node as the center node. .

[0127] In some embodiments, the positioning device uses a two-layer GAT encoder to generate node embeddings:

[0128]

[0129] Central node embedding It is a measurement set The permutation-invariant summary. The star topology ensures strict permutation invariance to the order of measurement points through its symmetric connectivity pattern, while achieving adaptive, feature-based weighting across measurement links through an attention mechanism.

[0130] Step 205b: The positioning device inputs the positioning probability map into the convolutional network to obtain the location feature information of the target user.

[0131] In some embodiments, the CNN described above is a neural network that is good at processing gridded data and can effectively extract spatial local features and hierarchical patterns.

[0132] In some embodiments, the aforementioned location feature information refers to the feature vector output by the CNN.

[0133] In some embodiments, a user location probability graph Spatial features are extracted through encoding via convolutional branches. Input After processing through three convolutional blocks, the final output is... Transformed into feature vectors through a linear layer , as a priori for the GNN branch.

[0134] Step 205c: The positioning device performs a fusion operation on the first feature information and the location feature information to obtain the location information of the target user.

[0135] In some embodiments, the above-described fusion operation refers to fusion via a multiplication gating mechanism.

[0136] In some embodiments, the above multiplication gating mechanism is a feature fusion method, typically in the form of: ,in It is a location feature. It is the first feature. This element-wise multiplication allows the network to learn from environmental priors. To dynamically modulate or weighted measure features To achieve adaptive fusion.

[0137] In some embodiments, the target user's location information is the final output, i.e., the estimated coordinates of the user's location. .

[0138] In some embodiments, the central node feature and Multiplication gating is used for fusion. A multilayer perceptron is used to regress the target user's location.

[0139]

[0140] In some embodiments, during the model training phase, the parameters of LocCGNN can be optimized by minimizing the mean absolute error (MAE) between the predicted and true location coordinates. Model parameters Optimize by minimizing the mean squared error (MAE) between the estimated and true values:

[0141]

[0142] in It is a mini-batch. It is a rounding operation that maps continuous coordinates to the nearest discrete grid point.

[0143] In this way, the positioning device can integrate the feature information output by GNN and CNN through the multiplication-gated fusion mechanism, thereby fusing the measurement branch with the environmental branch and achieving high-precision and high-reliability user location estimation in complex terrain.

[0144] The positioning method provided in this application, by constructing a positioning probability map, can utilize observation points that receive signals as positive evidence and combine them with observation points that do not receive signals as negative evidence to eliminate contradictory areas, thereby accurately narrowing down the possible location range of the user; furthermore, the terrain elevation information used in the construction of this positioning probability map can adapt to the three-dimensional environment of mountainous areas. This improves positioning accuracy.

[0145] The positioning method of this application will be described below through specific embodiments.

[0146] like Figure 9 As shown, the implementation process of the positioning method provided in this application embodiment includes the following S1 to S7:

[0147] S1. Obtain the RSS measurement value of the target user from K measurement points through the drone network.

[0148] S2. For each first measurement point that receives a valid signal, the positioning device calculates the similarity weight of that measurement point to each candidate location.

[0149] S3. For each second measurement point that has not received a valid signal, the positioning device calculates the probability weight of the target user at each candidate location for that measurement point.

[0150] S4. The positioning device merges the weights calculated for each measurement point with respect to all candidate locations, and constructs the complete likelihood function value for each candidate location through a multiplication method to obtain the positioning probability map.

[0151] S5. The positioning device inputs the RSS measurement values ​​from all the first measurement points and their corresponding measurement point location information into the GNN, and outputs the first feature information.

[0152] S6. The positioning device inputs the positioning probability map into a CNN branch, extracts the spatial environment features, and obtains the location feature information.

[0153] S7. The positioning device fuses the first feature information and the position feature information through a multiplication gating mechanism, and inputs the fused features into a multilayer perceptron to obtain the estimated position coordinates of the target user.

[0154] In this way, a more reliable positioning probability map can be constructed by fusing the two types of observations, "signal received" and "no signal received," thereby effectively eliminating the positional ambiguity caused by terrain obstruction. Furthermore, the terrain elevation information used in the construction of this positioning probability map can be adapted to the three-dimensional environment of mountainous areas, thus achieving higher-precision positioning in complex mountainous environments.

[0155] It should be noted that the descriptions of each step S1 to S7 in this embodiment can be found in the descriptions in the above embodiments, and will not be repeated here.

[0156] In some embodiments, to test the accuracy of the positioning method provided in this application, a channel map dataset named Mountain-CKM was created to evaluate the performance of the proposed network in mountainous terrain. This dataset was created for a typical 10km × 10km mountainous forest area using ray tracing simulation based on digital elevation model (DEM) data.

[0157] In some embodiments, the terrain is divided into a 100m × 100m grid as potential user locations, and the UAV receiver is deployed on a 200m × 200m grid in the air. Each sample in the dataset contains five valid UAV received signal maps, corresponding RSS values, five invalid measurement points, and a label indicating the user's actual ground grid location.

[0158] Experiment 1 focuses on the impact of negative evidence. This experiment aims to verify the effect of negative evidence on model performance. Figure 10 The impact of negative evidence on the outcome, such as Figure 10 As shown, the minimum MAE on the test set reaches 0.07 when negative evidence is included. In contrast, the minimum MAE is 0.14 when negative evidence is not included. The results indicate that integrating negative evidence can effectively reduce localization error.

[0159] Experiment 2 focuses on algorithm performance comparison. This experiment compares the performance of LocCGNN with other localization methods on the Mountain-CKM dataset and the Deterministic Propagation Model (DPM) (city) dataset, as well as the inference time for 128 users.

[0160] The baseline method is as follows:

[0161] LocUnet: A deep learning localization network designed specifically for dense urban scenes. Due to elevation variations in mountainous areas, we modified the environment map to a grayscale image that includes elevation information.

[0162] GNN: A network consisting of two layers of GAT and one MLP. The graph structure is consistent with LocCGNN, and the input is RSS and the drone's location.

[0163] Maximum Likelihood Estimation (MLE): Assuming a free-space path loss model and Gaussian noise distribution, the target location is estimated by maximizing the likelihood function.

[0164] The experimental results are compared in Table (1) below:

[0165]

[0166] Table (1)

[0167] The results are shown in Table (1). The LocCGNN method outperforms all other methods on both mountainous and urban datasets, achieving the lowest localization error. LocUnet performs comparably to LocCGNN on 2D urban datasets, but its performance is significantly worse on mountainous terrain, indicating a lack of ability to utilize key elevation information. Furthermore, the LocCGNN model has an average localization time of 17.36 ms for 128 users, demonstrating high efficiency sufficient to support fast, large-scale user localization.

[0168] This application addresses the inherent ambiguity in fingerprint positioning in mountainous areas and the insufficient adaptability of existing technologies to complex 3D environments by proposing two mutually supporting core innovations. First, this invention proposes a dual-evidence method to preprocess input data and construct a user location probability map. This method innovatively combines CKM data, using observation points that have received user RSS feeds as positive evidence to identify possible user regions, while simultaneously using observation points that have not received user RSS feeds as negative evidence to eliminate contradictory regions, thus effectively solving the ambiguity of positioning in mountainous areas. Second, based on this probability map, this invention proposes and constructs the LocCGNN fusion positioning network. LocCGNN combines the advantages of Graph Neural Networks (GNNs) and CNNs: GNNs are specifically used to extract robust representations with permutation invariance to receiver configuration from real-time observation data, while CNNs are responsible for extracting potential user location features from the dual-evidence probability map. Finally, LocCGNN efficiently integrates these two key features through GNNs, outputting accurate user coordinates, thereby significantly improving positioning accuracy and robustness in complex mountainous environments.

[0169] It should be noted that the above-described method embodiments, or the various possible implementations of the method embodiments, can be executed individually, or, provided there is no conflict, they can be combined with each other. The specific implementation can be determined according to actual usage requirements, and this application embodiment does not impose any restrictions on this.

[0170] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0171] This application embodiment can divide the positioning device into functional modules according to the above method example. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0172] In some embodiments, this application also provides a positioning device. The positioning device may include one or more functional modules for implementing the positioning method of the above method embodiments.

[0173] For example, Figure 11 This is a schematic diagram of a positioning device provided in an embodiment of this application. Figure 11 As shown, the positioning device 900 includes: an acquisition module 901, a determination module 902, a construction module 903, and an execution module 904.

[0174] The acquisition module 901 is used to acquire the RSS measurement value of the target user from at least one measurement point through the drone network.

[0175] The aforementioned determining module 902 is used to determine the similarity between a first RSS measurement value and the RSS prediction value of each candidate location in the CKM database when at least one measurement point includes a first measurement point. The first measurement point is the measurement point where the RSS measurement value of the target user is obtained. The first RSS measurement value is the RSS measurement value obtained from the first measurement point. Each RSS prediction value in the CKM database is determined based on the two-dimensional location information and measurement location information of at least one historical user.

[0176] The aforementioned determining module 902 is further configured to determine the probability that a target user exists at the location corresponding to the second measuring point when at least one measuring point includes the second measuring point; the second measuring point is a measuring point where no RSS measurement value of the target user has been obtained.

[0177] The aforementioned construction module 903 is used to construct a location probability map based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point. This location probability map is used to indicate the location of the target user.

[0178] The aforementioned execution module 904 is used to input the first RSS measurement value and the positioning probability map into the local communication graph neural network to obtain the location information of the target user.

[0179] In some embodiments, the above-mentioned construction module 903 is specifically used to: determine a likelihood function based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that a target user exists at the location corresponding to the second measurement point; and construct a location probability map based on the likelihood function.

[0180] In other embodiments, the determining module 902 is specifically used to: determine the normalized cosine similarity between the first RSS measurement value and the RSS prediction value of each candidate position based on the first RSS measurement value and the RSS prediction value of each candidate position in the CKM database; and determine the similarity between the first RSS measurement value and the RSS prediction value of each candidate position based on each normalized cosine similarity and an acceptable difference threshold.

[0181] In some other embodiments, the determination module 902 is specifically used to determine the probability that a target user exists at the location corresponding to the second measurement point based on the location corresponding to the second measurement point, the receiver detection threshold, and the probability of abnormal detection failure.

[0182] In some other embodiments, the aforementioned local communication graph neural network includes a graph neural network and a convolutional network; the execution module 904 is specifically used to: input a first RSS measurement value into the graph neural network to obtain first feature information, the first feature information being robust representation information with permutation invariance to receiver configuration extracted from the first RSS measurement value; input a positioning probability map into the convolutional network to obtain the location feature information of the target user; and perform a fusion operation on the first feature information and the location feature information to obtain the location information of the target user.

[0183] It should be noted that the positioning device can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0184] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 12 As shown, the electronic device 90 includes: a processor 92, a communication interface 93, and a bus 94. Optionally, the electronic device 90 may also include a memory 91.

[0185] Processor 92 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 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.

[0186] Communication interface 93 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.

[0187] The memory 91 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0188] As one possible implementation, the memory 91 can exist independently of the processor 92. The memory 91 can be connected to the processor 92 via a bus 94 and is used to store instructions or program code. When the processor 92 calls and executes the instructions or program code stored in the memory 91, it can implement the positioning method provided in the embodiments of this application.

[0189] In another possible implementation, memory 91 can also be integrated with processor 92.

[0190] Bus 94 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 94 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 12 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.

[0191] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.

[0192] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described positioning method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0193] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0194] This application also provides a readable storage medium storing a program or instructions that, when executed by a computer, implement the positioning method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware; the readable storage medium can be any of the foregoing embodiments or memory; the readable storage medium can also be an external storage device of the service invocation device, such as a pluggable hard drive, Smart Media Card (SMC), Secure Digital (SD) card, flash card, etc., equipped on the service invocation device. Further, the readable storage medium can include both internal storage units of the service invocation device and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the service invocation device. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0195] This application also provides a computer program product, which is stored in a storage medium and implements the positioning method provided in the above embodiments when the computer program product is executed by a computer.

[0196] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0197] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0198] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A positioning method, characterized in that, include: The RSS (Received Signal Strength) measurement value of the target user is obtained from at least one measurement point using a drone network. In the case where the at least one measurement point includes a first measurement point, the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the Channel Knowledge Map (CKM) database is determined. The first measurement point is the measurement point from which the RSS measurement value of the target user is obtained. The first RSS measurement value is the RSS measurement value obtained from the first measurement point. Each RSS prediction value in the CKM database is determined based on the two-dimensional location information and measurement location information of at least one historical user. In the case that at least one measurement point includes a second measurement point, the probability that the target user exists at the location corresponding to the second measurement point is determined; the second measurement point is a measurement point where the RSS measurement value of the target user has not been obtained; Based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point, a location probability map is constructed, which is used to indicate the location of the target user. The first RSS measurement value and the location probability map are input into a local communication graph neural network to obtain the location information of the target user.

2. The positioning method according to claim 1, characterized in that, The step of constructing a location probability map based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point, includes: Based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point, the likelihood function is determined. Based on the likelihood function, a localization probability map is constructed.

3. The positioning method according to claim 1, characterized in that, Determining the similarity between the first RSS measurement and the RSS prediction value for each candidate position in the CKM database includes: Based on the first RSS measurement value and the RSS prediction value of each candidate position in the CKM database, the normalized cosine similarity between the first RSS measurement value and the RSS prediction value of each candidate position is determined. Based on each of the normalized cosine similarities and acceptable difference thresholds, the similarity between the first RSS measurement and the RSS prediction value at each candidate location is determined.

4. The positioning method according to claim 1, characterized in that, Determining the probability that the target user exists at the location corresponding to the second measurement point includes: Based on the location corresponding to the second measurement point, the receiver detection threshold, and the probability of anomaly detection failure, the probability that the target user exists at the location corresponding to the second measurement point is determined.

5. The positioning method according to claim 1, characterized in that, The local communication graph neural network includes graph neural networks and convolutional networks; The step of inputting the first RSS measurement value and the positioning probability map into a local communication graph neural network to obtain the location information of the target user includes: The first RSS measurement value is input into a graph neural network to obtain the first feature information, which is a robust representation information with permutation invariance to receiver configuration extracted from the first RSS measurement value. The location probability map is input into the convolutional network to obtain the location feature information of the target user; A fusion operation is performed on the first feature information and the location feature information to obtain the location information of the target user.

6. A positioning device, characterized in that, include: Acquire modules, determine modules, construct modules, and execute modules; The acquisition module is used to acquire the RSS measurement value of the received signal strength of the target user from at least one measurement point through the UAV network; The determining module is configured to determine the similarity between a first RSS measurement value and an RSS prediction value for each candidate location in the Channel Knowledge Map (CKM) database, provided that at least one measurement point includes a first measurement point. The first measurement point is the measurement point from which the RSS measurement value of the target user is obtained. The first RSS measurement value is the RSS measurement value obtained from the first measurement point. Each RSS prediction value in the CKM database is determined based on two-dimensional location information and measurement location information of at least one historical user. The determining module is further configured to determine, when the at least one measurement point includes the second measurement point, the probability that the target user exists at the location corresponding to the second measurement point; the second measurement point is a measurement point where the RSS measurement value of the target user has not been obtained; The construction module is used to construct a location probability map based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point. The location probability map is used to indicate the location of the target user. The execution module is used to input the first RSS measurement value and the positioning probability map into a local communication graph neural network to obtain the location information of the target user.

7. The positioning device according to claim 6, characterized in that, The building module is specifically used for: Based on the similarity between the first RSS measurement value and the RSS prediction value of each candidate location in the CKM database, and the probability that the target user exists at the location corresponding to the second measurement point, the likelihood function is determined. Based on the likelihood function, a localization probability map is constructed.

8. The positioning device according to claim 6, characterized in that, The determining module is specifically used for: Based on the first RSS measurement value and the RSS prediction value of each candidate position in the CKM database, the normalized cosine similarity between the first RSS measurement value and the RSS prediction value of each candidate position is determined. Based on each of the normalized cosine similarities and acceptable difference thresholds, the similarity between the first RSS measurement and the RSS prediction value at each candidate location is determined.

9. The positioning device according to claim 6, characterized in that, The determining module is specifically used to determine the probability that the target user exists at the location corresponding to the second measurement point, based on the location corresponding to the second measurement point, the receiver detection threshold, and the probability of abnormal detection failure.

10. The positioning device according to claim 6, characterized in that, The local communication graph neural network includes graph neural networks and convolutional networks; The execution module is specifically used for: The first RSS measurement value is input into a graph neural network to obtain the first feature information, which is a robust representation information with permutation invariance to receiver configuration extracted from the first RSS measurement value. The location probability map is input into the convolutional network to obtain the location feature information of the target user; A fusion operation is performed on the first feature information and the location feature information to obtain the location information of the target user.

11. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the positioning method as described in any one of claims 1-5.

12. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a computer, implement the positioning method as described in any one of claims 1-5.

13. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when executed by a computer, the computer program product implements the positioning method as described in any one of claims 1-5.

Images