Ultra-wideband sensor air-ground cooperative positioning method and device and storage medium

Abstract

The application discloses an ultra-wideband sensor air-ground cooperative positioning method and device and a storage medium, and relates to the technical field of positioning. The method comprises the following steps: after determining an A node and a T node, collecting D node information between the A node and the T node; establishing a heterogeneous graph according to the A node, the T node and the D node information, and defining a meta-path according to the heterogeneous graph; fusing the characteristics of the A node, the T node and the D node according to the meta-path to obtain fused node characteristics; and based on the fused node characteristics, obtaining a predicted three-dimensional coordinate of the T node and a predicted scalar distance of the D node. The application can improve positioning accuracy and robustness, reduce deployment cost, adapt to multiple dynamic scenes, and balance real-time performance and generalization ability, thereby providing reliable support for technology engineering landing and cross-device promotion.

Description

Technical Field

[0001] This invention relates to the field of positioning technology, specifically to an ultra-wideband sensor-based air-to-ground cooperative positioning method, device, and storage medium. Background Technology

[0002] Ultra-wideband (UWB) has been widely used in indoor positioning and robot navigation due to its high accuracy and real-time performance. Current UWB positioning methods are mostly based on time of arrival (TOA / TDOA) or two-way ranging (TWR), combined with extended Kalman filtering (EKF), unscented Kalman filtering (UKF), or particle filtering to achieve single-label position estimation. In recent years, deep learning models such as convolutional networks, recurrent neural networks, and Transformers have also been introduced to learn temporal features from raw ranging sequences, achieving some success in certain scenarios, but still have the following limitations:

[0003] First, it relies on fixed base station deployment, which requires a lot of resources to maintain the global map and base station equipment, resulting in high deployment and maintenance costs and poor adaptability.

[0004] Second, the spatiotemporal features are fragmented, ranging data is processed in isolation, there is a lack of effective fusion of information across time steps, the anti-interference ability is weak, and the positioning accuracy and robustness are poor.

[0005] Third, insufficient pose observation dimensions make it easy for positioning failure to occur due to the lack of data from a single node, making it difficult to achieve joint pose observability.

[0006] Fourth, it has poor adaptability to dynamic scenarios, a fixed topology, and cannot flexibly match the platform's motion state, thus limiting its application boundaries.

[0007] Fifth, the feature processing process is complex and computationally intensive, making it difficult to balance real-time performance and positioning capabilities, thus hindering the implementation of the project. Summary of the Invention

[0008] In view of the deficiencies in the existing technology, the technical problem to be solved by this application is: how to enable ultra-wideband positioning technology to simultaneously meet the requirements of high-precision positioning, low-cost deployment and dynamic environment adaptation.

[0009] To achieve the above objectives, in a first aspect, embodiments of this application provide an ultra-wideband sensor air-to-ground cooperative positioning method, the method comprising the following steps:

[0010] After determining nodes A and T, collect the information of node D between nodes A and T;

[0011] A heterogeneous graph is constructed based on the information of nodes A, T, and D, and meta-paths are defined based on the heterogeneous graph.

[0012] Based on the characteristics of nodes A, T, and D in the meta-path fusion, the characteristics of the fused nodes are obtained.

[0013] The predicted 3D coordinates of node T and the predicted scalar distance of node D are obtained based on the feature regression of the fused nodes.

[0014] In conjunction with the first aspect, in one implementation, the process of collecting data on node D between node A and node T includes:

[0015] Deploy N ultra-wideband sensors at node A and M ultra-wideband sensors at node T;

[0016] Collect continuous ranging values ​​between node A and node T, and set each continuous ranging value as a measurement sample within a time step;

[0017] Collect the absolute timestamp corresponding to each consecutive ranging value in the measurement sample, and obtain the adjacent time difference between adjacent consecutive ranging values ​​based on the absolute timestamp;

[0018] Get the D node containing continuous distance values, absolute timestamps, and adjacent time differences.

[0019] In conjunction with the first aspect, in one implementation, the process of establishing the heterogeneous graph includes:

[0020] The nodes in the heterogeneous graph are determined to include A nodes, T nodes, and D nodes, and each pair of A nodes and T nodes uniquely corresponds to a D node; within the same time step, the heterogeneous graph contains N A nodes, M T nodes, and N×M D nodes.

[0021] Within the same time step, node A is connected to its corresponding node D, and node T is connected to its corresponding node D through spatial edges one by one;

[0022] At different time steps, the same node A and / or the same node T are connected by time-series edges.

[0023] In conjunction with the first aspect, in one implementation, the connection of the time-series edge is determined based on the movement states of node A and node T, and the determination process includes:

[0024] When node A is stationary and node T moves, only time-series edges are connected between nodes T.

[0025] When both node A and node T move, time-series edges are connected between nodes A and between nodes T.

[0026] In conjunction with the first aspect, in one implementation, a node feature extraction model and a meta-path aggregation model are constructed before fusing the features of nodes A, T, and D based on the meta-path.

[0027] The feature extraction model includes a temporal encoder, a preprocessing mapping submodule, and a unified dimension projection submodule;

[0028] The meta-path aggregation model includes intra-meta-path aggregation model and inter-meta-path aggregation model.

[0029] In conjunction with the first aspect, in one implementation, the process of obtaining the fusion node features includes:

[0030] S1. Map the absolute timestamp and adjacent time difference into a time embedding vector using a time encoder;

[0031] S2. The preprocessing mapping submodule concatenates, linearly maps, and transforms the original feature vectors of nodes A, T, and D with their corresponding time embedding vectors to obtain intermediate representation vectors.

[0032] S3. The intermediate representation vector is mapped to the unified hidden space and normalized using the unified dimension projection submodule to obtain the hidden features of nodes A, T, and D in the same channel dimension. The hidden features of node D are derived from... Feature transformation;

[0033] S4. The hidden features of nodes A, T, and D on any metapath are arranged into a sequence according to the metapath order using the metapath aggregation model. After sequence encoding and pooling, path-level aggregated features are obtained.

[0034] S5. The path-level aggregated features of multiple meta-paths corresponding to the same end node are stacked using the meta-path aggregation model. After fusion encoding, pooling and linear mapping, the fused node features corresponding to the end node are obtained.

[0035] In conjunction with the first aspect, in one implementation, the process of obtaining the predicted 3D coordinates of node T and the predicted scalar distance of node D based on fused node feature regression includes:

[0036] The last frame time step is recorded as Extract the fused node features of all T nodes and D nodes at this time step;

[0037] Input the fused node features of node T into the location regression head to obtain the predicted 3D coordinates of node T;

[0038] Input the fused node features of node D into the distance regression head to obtain the predicted scalar distance of node D.

[0039] In conjunction with the first aspect, in one implementation, the node feature extraction model and the meta-path aggregation model use a total loss function as the optimization objective, which is:

[0040] ;

[0041] In the formula, For the total loss function, For the location loss function, Let distance be the loss function. Let the geometric consistency loss function be... All are weighting coefficients.

[0042] Secondly, embodiments of this application provide an ultra-wideband sensor air-to-ground cooperative positioning device, the ultra-wideband sensor air-to-ground cooperative positioning device including a processor, a memory, and an ultra-wideband sensor air-to-ground cooperative positioning program stored in the memory and executable by the processor, wherein when the ultra-wideband sensor air-to-ground cooperative positioning program is executed by the processor, it implements the steps of the method provided in the first aspect.

[0043] Thirdly, embodiments of this application provide a computer-readable storage medium storing an ultra-wideband sensor air-to-ground cooperative positioning program, wherein when the ultra-wideband sensor air-to-ground cooperative positioning program is executed, it implements the steps of the method provided in the first aspect.

[0044] Compared with the prior art, the advantages of this application are:

[0045] (1) By switching between the A node and T node of the drone and the unmanned vehicle, the anchor point and the tag can be dynamically switched. Combined with the real-time topology update of the heterogeneous graph, the fixed base station can be replaced, saving the tedious work of site deployment, calibration and equipment maintenance. Furthermore, it can adapt to different motion scenarios through differentiated time-series edge rules, without relying on fixed infrastructure, reducing deployment and maintenance costs, and adapting to lightweight requirements such as temporary operations and multi-area reconstruction.

[0046] (2) By constructing the topology of three types of nodes, A, T and D, through heterogeneous graphs, the spatial edges are used to model physical associations and the temporal edges are used to connect cross-time step information. The meta-path is combined to realize the structured aggregation of spatiotemporal features. At the same time, the absolute timestamp and adjacent time difference are transformed into time embedding vectors, so that the feature representation has both ranging data and temporal semantics, effectively resisting the impact of abnormal scenarios such as frame loss and multipath interference, improving positioning accuracy and enhancing robustness.

[0047] (3) Based on the configuration of N A nodes, M T nodes and N×M D nodes, the multi-sensor collaborative information and rigid body geometric constraints are deeply integrated through the meta-path aggregation model. Without increasing the hardware cost, the limitations of insufficient traditional positioning pose observation dimensions are overcome, and the position and attitude are jointly observable. This avoids positioning failure caused by missing data from a single node and improves the completeness and reliability of the results.

[0048] (4) By strongly binding the meta-path, temporal edge design with the node motion state, the actual scene topology is dynamically matched, breaking the dependence of traditional technology on fixed environment. It can run stably in scenarios without prior site deployment, such as emergency rescue, harsh environment operation, and temporary collaborative tasks, expanding the application scenario coverage and improving versatility and adaptability.

[0049] (5) A three-level feature extraction strategy of time coding, preprocessing mapping and dimensional projection and a two-level aggregation strategy of intra-path and inter-path are adopted to capture key dependencies while accurately controlling computational complexity and improving feature processing efficiency to meet real-time response requirements, take into account high precision and engineering practicality, and facilitate deployment on various air-ground collaborative devices. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the ultra-wideband sensor air-to-ground cooperative positioning method of this application;

[0052] Figure 2 This is the heterogeneous graph network formed by time steps Step1 and Step2 when node A is stationary in this application;

[0053] Figure 3 This is the heterogeneous graph network formed by time steps Step1 and Step2 when both node A and node T move in this application;

[0054] Figure 4 This is a schematic diagram of the process for obtaining the fusion node characteristics in this application;

[0055] Figure 5 A flowchart illustrating the method provided for the first aspect of this application;

[0056] Figure 6 This is a schematic diagram of the hardware structure of the ultra-wideband sensor air-to-ground cooperative positioning device involved in the embodiments of this application. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0059] First, some of the technical terms used in this application will be explained to help those skilled in the art understand this application.

[0060] A (Anchor) node: A drone or unmanned vehicle that serves as a base station;

[0061] T (Tag) node: A drone or unmanned vehicle used as a tag;

[0062] D (Distance) node: The carrier of distance measurement information between A node and T node;

[0063] Metapath: The route by which information is transmitted between nodes A, D, and T.

[0064] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0065] Firstly, referring to Figure 1 , Figure 1 This is a flowchart illustrating the ultra-wideband sensor air-to-ground cooperative positioning method of this application. Figure 1 As shown in the figure, this application provides an ultra-wideband sensor air-to-ground cooperative positioning method, the method comprising the following steps:

[0066] After determining nodes A and T, collect the information of node D between nodes A and T;

[0067] A heterogeneous graph is constructed based on the information of nodes A, T, and D, and meta-paths are defined based on the heterogeneous graph.

[0068] Based on the characteristics of nodes A, T, and D in the meta-path fusion, the characteristics of the fused nodes are obtained.

[0069] The predicted 3D coordinates of node T and the predicted scalar distance of node D are obtained based on the feature regression of the fused nodes.

[0070] This allows for the free switching between A-nodes and T-nodes via drones and unmanned vehicles, enabling dynamic switching of anchor points and tags. Combined with real-time topology updates of heterogeneous graphs, it replaces fixed base stations, eliminating tedious site deployment, calibration, and equipment maintenance. Furthermore, it adapts to different motion scenarios through differentiated temporal edge rules, eliminating the need for fixed infrastructure, reducing deployment and maintenance costs, and adapting to lightweight requirements such as temporary operations and multi-region reconstruction.

[0071] In one embodiment, the D-node information in the above method refers to the D-node information between node A and node T at one time step. In this case, the process of collecting the D-node information between node A and node T includes:

[0072] Deploy N ultra-wideband sensors at node A (e.g., deploy N ultra-wideband sensors on a drone), and deploy M ultra-wideband sensors at node T (e.g., deploy M ultra-wideband sensors on an unmanned vehicle).

[0073] The coordinates of all ultra-wideband sensors are calibrated to ensure a consistent ranging benchmark.

[0074] After calibration, collect continuous ranging values ​​between node A and node T, with each set number (e.g., 10) of continuous ranging values ​​constituting one time step. Measurement samples within;

[0075] Collect the absolute timestamp corresponding to each consecutive ranging value in the measurement sample. Based on absolute timestamp Obtain the time difference between adjacent consecutive ranging values ;

[0076] Get the D node containing continuous distance values, absolute timestamps, and adjacent time differences.

[0077] In this embodiment, the measured sample satisfies:

[0078] ;

[0079] At this point, the set of continuous distance measurements in the measurement sample The dimensions are: N×M×10, where, Indicates the first The T-node and the first Continuous distance measurements between nodes A.

[0080] By deploying and calibrating node sensors, a unified ranging benchmark is ensured. The continuous set of ranging values, absolute timestamps, and adjacent time differences are integrated into D-node information to synchronously collect time series and ranging data. This provides high-quality data support for subsequent heterogeneous graph modeling, overcomes the pain points of traditional data isolation and insufficient accuracy, and improves the reliability of feature representation.

[0081] In one embodiment, the process of establishing the heterogeneous graph in the above method includes:

[0082] The nodes in the heterogeneous graph are determined to include nodes A, T, and D, and each pair of nodes A and T uniquely corresponds to a node D.

[0083] Within the same time step, the heterogeneous graph contains N A nodes: M T nodes: And N×M D nodes: ,in This is the current time step;

[0084] Within the same time step, node A is connected to the corresponding node D, and node T is connected to the corresponding node D through spatial edges one by one, forming a spatial association link of node A-spatial edge-node D-spatial edge-node T;

[0085] At different time steps, the same node A and / or the same node T are connected by a time-series edge; such as and , and Connecting the time-series edges between them, where for The connection of time-series edges is determined based on the movement states of nodes A and T at adjacent time steps. The determination process includes:

[0086] When node A is stationary and node T moves, only temporal edges are connected between nodes T, as shown in the reference. Figure 2 , Figure 2 In this application, the heterogeneous graph network formed by time steps Step1 and Step2 when node A is stationary is defined as follows:

[0087] - - - - At this point, the metapath is connected sequentially only through spatial edges, and the information of node T in the next time step is updated based on continuous ranging values. Node D carries the set of continuous ranging values ​​for the corresponding time step. ;

[0088] - - - At this point, the metapath is connected alternately by spatial and temporal edges, used to monitor the information of the D node in reverse through the updated T node. The D nodes in adjacent time steps each carry the set of continuous ranging values ​​for the corresponding time step. ;

[0089] When both node A and node T move, temporal edges are established between nodes A and between nodes T, as shown in the reference. Figure 3 , Figure 3 In this application, the heterogeneous graph network formed by time steps Step1 and Step2 when both node A and node T move is defined. The meta-path flow then includes:

[0090] - - - - - At this point, the metapath is connected alternately through spatial and temporal edges, relying on continuous ranging values ​​to update the information of node T in the next time step, and node D carries the set of continuous ranging values ​​for the corresponding time step. ;

[0091] - - - It uses alternating spatial and temporal edges for reverse monitoring of node D information via updated node T. Node D in adjacent time steps carries a set of continuous ranging values ​​for the corresponding time step. .

[0092] In this way, three types of node topologies, A, T, and D, are constructed through heterogeneous graphs. Spatial edges are used to model physical relationships, and temporal edges are used to connect cross-time step information. Combined with meta-paths, spatiotemporal features are aggregated in a structured manner.

[0093] In one embodiment, before fusing the features of nodes A, T, and D based on the metapath, a node feature extraction model and a metapath aggregation model are constructed; wherein the feature extraction model includes a time encoder (TimeMLP), a preprocessing mapping submodule (pre), and a unified dimension projection submodule (proj); the metapath aggregation model includes an intra-metapath aggregation model (Intra) and an inter-metapath aggregation model (Inter).

[0094] Based on this, refer to Figure 4 , Figure 4 This is a schematic diagram illustrating the process of obtaining the fusion node characteristics in this application. For example... Figure 4 As shown, the process for obtaining fused node features includes:

[0095] S1. Map the absolute timestamp and adjacent time difference into a time embedding vector using a time encoder, for any node type. Temporal embedding vector satisfy:

[0096] ;

[0097] In the formula, For time embedding vectors, It is a non-linear activation function. For nodes absolute timestamp, For nodes Adjacent time difference, , , , ;

[0098] S2. The preprocessing mapping submodule concatenates, linearly maps, and nonlinearly transforms the original feature vectors of nodes A, T, and D with their corresponding time embedding vectors to obtain intermediate representation vectors. satisfy:

[0099] ;

[0100] In the formula, This is an intermediate representation vector. For node A, The input vector is a concatenation of a nonlinear activation function, T nodes, and D nodes. The weight matrix has the number of rows equal to the output dimension. Number of columns = input dimension , This is the bias value;

[0101] S3. The intermediate representation vector is mapped to the unified hidden space and normalized using the unified dimension projection submodule to obtain the hidden features of nodes A, T, and D in the same channel dimension. The hidden features of node D are derived from... Feature transformation; hidden features satisfy:

[0102] ;

[0103] In the formula, This is an intermediate representation vector. , These are the learnable weights and biases, respectively. It is a non-linear activation function. For layer normalization operators, To unify hidden dimensions;

[0104] It is important to note that for any node type The set of parameters appearing in S1-S3 They do not share resources;

[0105] S4. Using the metapath aggregation model, the hidden features of nodes A, T, and D on any metapath are arranged into a sequence according to the metapath order. ( The length of the metapath. To unify the hidden dimensions, sequence encoding and pooling are performed to obtain path-level aggregated features:

[0106] ;

[0107] In the formula, This is a path-level aggregation feature. For sequence encoders, For sequence pooling operators;

[0108] S5. Using the meta-path aggregation model to aggregate the same end node The corresponding path-level aggregated features of multiple meta-paths are stacked to obtain the path dimension: After fusion encoding, pooling, and linear mapping, the end node is obtained. Corresponding fusion node features:

[0109] ;

[0110] In the formula, End node Corresponding fusion node features, For end nodes The number of meta-paths to the target. To unify hidden dimensions, For fusion encoders, It is a linear mapping tool.

[0111] This embodiment employs a three-level feature extraction strategy of time coding, preprocessing mapping, and dimensional projection, along with a two-level aggregation strategy of intra-path and inter-path aggregation. This strategy captures key dependencies while precisely controlling computational complexity and improving feature processing efficiency to meet real-time response requirements. It balances high precision with engineering practicality and facilitates deployment on various air-ground collaborative devices.

[0112] In one embodiment, the process of obtaining the predicted 3D coordinates of node T and the predicted scalar distance of node D based on fused node feature regression includes:

[0113] The last frame time step is recorded as Extract the fused node features of all T nodes and D nodes at this time step;

[0114] The fused node features of node T are input into the location regression head to obtain the predicted 3D coordinates of node T, which satisfy the following:

[0115] ;

[0116] In the formula, This refers to the fusion node characteristics of node T. It is a nonlinear activation function (such as SiLU or ReLU). , , , These are the weights and biases for each layer;

[0117] Input the fused node features of node D into the distance regression head to obtain the predicted scalar distance of node D, which satisfies:

[0118] ;

[0119] In the formula, This refers to the fusion node characteristics of node D. It is a nonlinear activation function (such as SiLU or ReLU). , , , These represent the weights and biases of each layer. , , , .

[0120] Furthermore, a dataset is constructed based on the pre-collected true 3D coordinates of node T, the true scalar distance of node D, the predicted 3D coordinates of node T, and the predicted scalar distance of node D. This dataset is used to train the node feature extraction model and the meta-path aggregation model. A total loss function is established as the optimization objective to improve localization accuracy. The formula for the total loss function is:

[0121] ;

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] In the formula, For the total loss function, A smooth L1 loss function is used to calculate the error between the predicted and the true values. (In this embodiment, to enhance the model's robustness to outlier values, a smooth L1 loss function is preferred. However, in other embodiments, other regression loss functions such as mean squared error (MSE), mean absolute error (MAE / L1), or Huber loss can also be used, and this invention does not limit this.) For the location loss function, To predict the number of nodes T, Let T be the predicted 3D coordinates of node T. The true 3D coordinates of node T Let distance be the loss function. The number of consecutive ranging values ​​within a time step. Let D be the predicted scalar distance. The true scalar distance to node D. Let the geometric consistency loss function be... Let T be the distance from the predicted 3D coordinates of node T to node A, where For the predicted 3D coordinates of the label node, Given the known three-dimensional coordinates of the base station node. It is the Euclidean norm (i.e., the straight-line distance between two points). All are weighting coefficients.

[0127] This constructs a triple constraint system of location, distance, and geometric consistency, ensuring that the prediction results conform to physical geometric laws and effectively avoiding the influence of data noise and distribution offset. It enhances the model's generalization ability, balances positioning accuracy and result reasonableness, and adapts to the needs of different complex scenarios.

[0128] In one embodiment, reference is made to Figure 5 , Figure 5 A flowchart illustrating the method provided for the first aspect of this application. (See attached diagram.) Figure 5 As shown, the process of the method provided by the first aspect includes:

[0129] S1. The drone acts as a base station, denoted as node A, and the unmanned vehicle acts as a tag, denoted as node T. Collect the D node information between node A and node T (see the above process for collecting the D node information between node A and node T).

[0130] S2. Establish a heterogeneous graph based on the information of nodes A, T, and D, and define meta-paths based on the heterogeneous graph (see the above process for establishing a heterogeneous graph).

[0131] S3. Construct a node feature extraction model and a meta-path aggregation model;

[0132] S4. The node feature extraction model and meta-path aggregation model obtain the fused node features by fusing the features of nodes A, T, and D based on the meta-path (see the above process for obtaining fused node features).

[0133] S5. Obtain the predicted 3D coordinates of node T and the predicted scalar distance of node D based on the feature regression of fused nodes (see the above process for obtaining the predicted 3D coordinates of node T and the predicted scalar distance of node D based on the feature regression of fused nodes).

[0134] S6. Optimize the node feature extraction model and the meta-path aggregation model with the total loss function (see the total loss function calculation formula) as the objective.

[0135] Secondly, embodiments of this application provide an ultra-wideband sensor air-to-ground cooperative positioning device, which can be a personal computer (PC), laptop computer, server, or other device with data processing capabilities.

[0136] Reference Figure 6 , Figure 6 This is a schematic diagram of the hardware structure of the ultra-wideband sensor air-to-ground cooperative positioning device involved in the embodiments of this application. In the embodiments of this application, the ultra-wideband sensor air-to-ground cooperative positioning device may include a processor, a memory, a communication interface, and a communication bus.

[0137] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0138] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of the ultra-wideband sensor air-to-ground cooperative positioning device, as well as interfaces used for interconnecting the ultra-wideband sensor air-to-ground cooperative positioning device with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0139] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0140] The processor can be a general-purpose processor, which can call the ultra-wideband sensor air-to-ground cooperative positioning program stored in the memory and execute the ultra-wideband sensor air-to-ground cooperative positioning method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the ultra-wideband sensor air-to-ground cooperative positioning program is called can be referred to in the various embodiments of the ultra-wideband sensor air-to-ground cooperative positioning method of this application, and will not be repeated here.

[0141] Those skilled in the art will understand that Figure 6 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0142] Thirdly, embodiments of this application also provide a computer-readable storage medium.

[0143] The computer-readable storage medium of this application stores an ultra-wideband sensor air-to-ground cooperative positioning program, wherein when the ultra-wideband sensor air-to-ground cooperative positioning program is executed by a processor, it implements the steps of the ultra-wideband sensor air-to-ground cooperative positioning method as described above.

[0144] The method implemented when the ultra-wideband sensor air-to-ground cooperative positioning program is executed can be referred to in the various embodiments of the ultra-wideband sensor air-to-ground cooperative positioning method of this application, and will not be repeated here.

[0145] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0146] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0147] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0148] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0149] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0150] 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 software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0151] The above are merely specific embodiments of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the scope of the claims.

Claims

1. A method for ultra-wideband sensor-based air-to-ground cooperative positioning, characterized in that, The method includes the following steps: After determining nodes A and T, collect the information of node D between nodes A and T; A heterogeneous graph is constructed based on the information of nodes A, T, and D, and meta-paths are defined based on the heterogeneous graph. Based on the characteristics of nodes A, T, and D in the meta-path fusion, the characteristics of the fused nodes are obtained. The predicted 3D coordinates of node T and the predicted scalar distance of node D are obtained based on the feature regression of the fused nodes.

2. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 1, characterized in that, The process of collecting data on node D between node A and node T includes: Deploy N ultra-wideband sensors at node A and M ultra-wideband sensors at node T; Collect continuous ranging values ​​between node A and node T, and set each continuous ranging value as a measurement sample within a time step; Collect the absolute timestamp corresponding to each consecutive ranging value in the measurement sample, and obtain the adjacent time difference between adjacent consecutive ranging values ​​based on the absolute timestamp; Get the D node containing continuous distance values, absolute timestamps, and adjacent time differences.

3. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 2, characterized in that, The process of establishing the heterogeneous graph includes: The nodes in the heterogeneous graph are determined to include A nodes, T nodes, and D nodes, and each pair of A nodes and T nodes uniquely corresponds to a D node; within the same time step, the heterogeneous graph contains N A nodes, M T nodes, and N×M D nodes. Within the same time step, node A is connected to its corresponding node D, and node T is connected to its corresponding node D through spatial edges one by one; At different time steps, the same node A and / or the same node T are connected by time-series edges.

4. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 3, characterized in that, The connection of the time-series edge is determined based on the movement states of node A and node T. The determination process includes: When node A is stationary and node T moves, only time-series edges are connected between nodes T. When both node A and node T move, time-series edges are connected between nodes A and between nodes T.

5. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 2, characterized in that, Before fusing the features of nodes A, T, and D based on metapaths, a node feature extraction model and a metapath aggregation model are constructed. The feature extraction model includes a temporal encoder, a preprocessing mapping submodule, and a unified dimension projection submodule; The meta-path aggregation model includes intra-meta-path aggregation model and inter-meta-path aggregation model.

6. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 5, characterized in that, The process for obtaining the features of the fusion node includes: S1. Map the absolute timestamp and adjacent time difference into a time embedding vector using a time encoder; S2. The preprocessing mapping submodule concatenates, linearly maps, and transforms the original feature vectors of nodes A, T, and D with their corresponding time embedding vectors to obtain intermediate representation vectors. S3. The intermediate representation vector is mapped to the unified hidden space and normalized through the unified dimension projection submodule to obtain the hidden features of nodes A, T and D in the same channel dimension. S4. The hidden features of nodes A, T, and D on any metapath are arranged into a sequence according to the metapath order using the metapath aggregation model. After sequence encoding and pooling, path-level aggregated features are obtained. S5. The path-level aggregated features of multiple meta-paths corresponding to the same end node are stacked using the meta-path aggregation model. After fusion encoding, pooling and linear mapping, the fused node features corresponding to the end node are obtained.

7. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 6, characterized in that, The process of obtaining the predicted 3D coordinates of node T and the predicted scalar distance of node D based on fusion node feature regression includes: The last frame time step is recorded as Extract the fused node features of all T nodes and D nodes at this time step; Input the fused node features of node T into the location regression head to obtain the predicted 3D coordinates of node T; Input the fused node features of node D into the distance regression head to obtain the predicted scalar distance of node D.

8. The ultra-wideband sensor air-to-ground cooperative positioning method according to claim 7, characterized in that, The node feature extraction model and the meta-path aggregation model use a total loss function as the optimization objective. This total loss function is: ； In the formula, For the total loss function, For the location loss function, Let distance be the loss function. Let the geometric consistency loss function be... All are weighting coefficients.

9. An ultra-wideband sensor-based air-to-ground cooperative positioning device, characterized in that, The ultra-wideband sensor air-to-ground cooperative positioning device includes a processor, a memory, and an ultra-wideband sensor air-to-ground cooperative positioning program stored in the memory and executable by the processor, wherein when the ultra-wideband sensor air-to-ground cooperative positioning program is executed by the processor, it implements the steps of the ultra-wideband sensor air-to-ground cooperative positioning method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an ultra-wideband sensor air-to-ground cooperative positioning program, wherein when the ultra-wideband sensor air-to-ground cooperative positioning program is executed, it implements the steps of the ultra-wideband sensor air-to-ground cooperative positioning method as described in any one of claims 1 to 8.

Images