Radio map extrapolation method and system based on simulation correction and residual propagation

Abstract

This invention discloses a method and system for radio map extrapolation based on simulation correction and residual propagation, relating to wireless communication map construction technology. It combines building and transmitter information to conduct ray tracing simulations to obtain global simulation data. Based on the distribution of measured points, it divides the measured areas into areas to be extrapolated, including those with continuous missing regions. A linear correction model is constructed using simulation and measured data at the same points to calibrate the global simulation values, and the simulation-measured residuals of the measured points are calculated. Multi-dimensional propagation features such as structural visibility, wall penetration, and spatial partitioning are extracted. A weighted graph model is constructed by fusing spatial proximity, propagation similarity, and power differences. The residuals of missing regions are extrapolated using graph Laplace smoothing. After superimposing residual compensation simulation data, a complete radio map is generated. This invention decouples global and local propagation errors in a layered manner. Based on the residual propagation constrained by building structures, it requires only a small amount of measured data to accurately extrapolate large areas of missing regions, making it suitable for various communication scenarios, including indoor and park environments.

Description

Technical Field

[0001] This invention relates to the field of wireless communication map construction technology, specifically to a method and system for radio map extrapolation based on simulation correction and residual propagation. Background Technology

[0002] Radio maps can intuitively represent the spatial distribution patterns of wireless propagation parameters such as received signal strength and reference signal received power within a target area. They are core foundational data for wireless network planning, coverage blind spot investigation, interference suppression, high-precision indoor positioning, and iterative optimization of communication networks. Traditional radio map construction relies on point-by-point on-site sampling across the entire area, requiring significant investment of manpower, testing equipment, and time. It is completely unsuitable for scenarios that are inaccessible, such as enclosed indoor spaces, high-risk industrial areas, and restricted, closed spaces. To reduce the workload of on-site sampling, the industry has successively developed three mainstream radio map construction solutions: spatial interpolation, deep learning data-driven approach, and ray tracing simulation. However, all three solutions have inherent defects: pure spatial interpolation relies solely on the spatial distance between points to infer signals in unknown areas, which cannot adapt to signal abrupt changes caused by wall partitions and room zoning; data-driven models such as CNN and graph neural networks are highly dependent on sufficient, uniform training samples that match the target scene. When sampling is sparse and there are large areas with no continuous missing samples, overfitting and a sharp drop in generalization ability are very likely to occur; ray tracing simulation can output a complete prior signal across the entire domain, but the simulation parameters have a global scale offset from the real environment. At the same time, wall materials and local occlusions will cause spatial non-uniform local errors between simulation and actual measurement. Simply using uniform scaling and offset correction cannot eliminate local deviations in different zones. Summary of the Invention

[0003] To address the aforementioned technical problems, this application discloses a method and system for radio map extrapolation based on simulation correction and residual propagation; the method for radio map extrapolation based on simulation correction and residual propagation includes the following steps:

[0004] The system acquires building structure information, transmitter location information, and local measured data for the target area, and performs wireless propagation simulation based on the building structure information and transmitter location information to obtain simulation data covering the entire area.

[0005] Based on the spatial distribution of the local measured data, the target area is divided into a measured area and an area to be extrapolated, wherein the area to be extrapolated consists of one or more consecutive missing areas.

[0006] A linear correction model is constructed using the measured data within the measured area and the simulation data at the corresponding location. The simulation data for the entire area is then corrected to obtain the corrected simulation data.

[0007] Calculate the difference between the measured value and the corresponding corrected simulation value at each measured point in the measured area to obtain the simulation-measured residual;

[0008] Based on the building structure information and the transmitter location information, propagation state features characterizing the propagation characteristics of wireless signals are extracted. The propagation state features include at least one of the following: structural visibility status, number of direct-connection wires passing through walls, and spatial partitioning relationship.

[0009] Using the simulation-measured residuals within the measured area as constraints, a graph model is constructed by integrating spatial proximity relationships, the similarity of the propagation state features, and the corrected power differences. The simulation-measured residuals are then propagated from the measured area to the extrapolation area using a residual propagation algorithm to obtain the estimated residuals of the extrapolation area.

[0010] The estimated residuals of the region to be extrapolated are superimposed onto the corresponding corrected simulation data and combined with the measured data of the measured region to generate a radio map of the target region.

[0011] Preferably, several reference points are pre-distributed in a uniform grid pattern within the target area, and all reference points together constitute a complete sampling grid for the target area; only some reference points collect local measured data, corresponding to the measured area. Reference points for which no local measured data was collected are clustered according to spatial adjacency to form continuous missing regions. All continuous missing regions together constitute the region to be extrapolated. ,satisfy , A single contiguous missing region consists of at least two spatially adjacent reference points that are to be extrapolated and for which there is no valid local measured data.

[0012] The execution process of the wireless propagation simulation includes: building a simulation scene based on building structure information, recording the transmitter location, frequency band, transmission power, antenna parameters, and propagation model parameters in the simulation scene, and using a ray tracing algorithm to traverse all reference points to calculate the simulation data corresponding to each reference point.

[0013] Preferably, the linear correction model is a globally unified linear transformation model, and its model expression is:

[0014]

[0015] In the formula, Reference point numbering, For the first Simulation data at one reference point, For the first Simulation data corrected for each reference point The scaling factor. This is a global bias term;

[0016] The scaling coefficient is solved with the objective of minimizing the squared error between the corrected simulation data and the measured data for all reference points in the measured area. Bias terms The optimization objective formula is:

[0017]

[0018] In the formula, For the first Measured data collected from several reference points Representing the Each reference point belongs to the measured area;

[0019] The scale coefficient obtained by the solution Bias terms Substitute the linear correction model and traverse all reference points in the target area to complete the global simulation data correction.

[0020] Preferably, the simulation-measured residual is calculated only for reference points within the measured area that have measured data, and the residual calculation formula is:

[0021]

[0022] In the formula, For the first Simulation-measured residuals corresponding to each measured reference point;

[0023] The simulation-measured residual is used to characterize the local numerical deviation between the simulation data after linear correction and the real environment caused by wall materials, local occlusion, equipment response differences, and multipath disturbances. The residuals corresponding to all measured reference points together form the residual dataset.

[0024] Preferably, the similarity of propagation states between pairwise reference points Calculated using the following formula:

[0025]

[0026] Differences in propagation status ;

[0027] , From the transmitter to the reference point , Spatial distance, , The structure is visible. , This refers to the number of straight-through cables passing through walls. , Spatial partitioning relationship, For indicator functions, , For normalization parameters, , , , These are the weighting coefficients.

[0028] Preferably, the graph model uses each reference point as an independent graph node, and establishes connection edges only between graph nodes within the spatial neighborhood, with connection weights between two nodes. The calculation formula is:

[0029] ;

[0030] In the formula, , For reference point , spatial coordinates, For spatial distance scale parameters, This is the power difference scale parameter.

[0031] Preferably, the graph Laplacian smoothing algorithm is used to solve for the estimated residuals of all graph nodes in the region to be extrapolated, and a residual smoothing optimization objective is constructed:

[0032] ;

[0033] In the formula, For observation constraint weights, Given the known residuals at the measured reference point, To smooth the constraint strength, The target is the entire domain;

[0034] After converting the optimization objective into matrix form, solve for the closed-form solution: , To observe the constraint weight diagonal matrix, For the graph Laplace matrix, For the graph adjacency matrix The corresponding degree matrix.

[0035] A radio map extrapolation system based on simulation correction and residual propagation, used to perform any of the radio map extrapolation methods described above, including:

[0036] The data acquisition module acquires building structure information, transmitter location information, and local measured data of the target area, and performs ray tracing wireless propagation simulation based on the building structure information and transmitter location information, outputting simulation data covering all reference points in the entire area;

[0037] The region division module divides the measured region and the extrapolation region, which consists of several consecutive missing regions, based on the distribution of local measured data at the reference point.

[0038] The simulation correction module extracts measured data and simulation data of the same reference point within the measured area, constructs and solves a global linear correction model, and uniformly corrects the simulation data across the entire domain to obtain the corrected simulation data.

[0039] The residual calculation module iterates through all reference points in the measured area, calculates the simulation-measured residuals for each point based on the measured data and the corrected simulation data, and generates a residual dataset.

[0040] The feature extraction module combines the coordinates of the building walls and the geometric positions of the transmitter and reference points to extract the structural visibility status, number of straight lines passing through the walls, and spatial zoning relationship propagation status features of each reference point.

[0041] The residual propagation module uses the residual dataset as a constraint and combines spatial proximity, similarity of propagation state, and difference of power after correction to construct a weighted graph model. The estimated residuals of each reference point in the region to be extrapolated are solved by the graph Laplace smoothing algorithm.

[0042] The map generation module overlays the estimated residuals of each reference point in the area to be extrapolated onto the corresponding corrected simulation data, combines the predicted data of the extrapolated area with the measured data of the measured area, and outputs a complete radio map of the target area.

[0043] An electronic device includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the radio map extrapolation method based on simulation correction and residual propagation.

[0044] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the radio map extrapolation methods based on simulation correction and residual propagation.

[0045] Compared with the prior art, the technical solution of this application has the following technical effects:

[0046] This invention constructs a complete spatial prior based on global wireless propagation simulation data, and builds a global linear correction model in conjunction with local measured data. It uniformly calibrates the overall scale and numerical offset of the simulation data, so that the signal distribution trend of the simulation output conforms to the real communication environment. This provides benchmark data for subsequent residual calculation and regional extrapolation, and completely preserves the signal change law of the target area in the whole space without the need for additional global sampling samples.

[0047] This invention separates global scale bias from local environmental residuals, calculates the simulation and measurement residuals of each measured point separately, and accurately extracts local numerical differences caused by factors such as wall material, local occlusion, and equipment response to form an independent residual dataset. It quantifies the spatial distribution bias that cannot be eliminated by global correction separately, and provides accurate local error basis for signal compensation in the area to be extrapolated.

[0048] This invention extracts multi-dimensional propagation features such as the visible state of the structure, the number of direct-connection lines penetrating walls, and spatial partitioning relationships. It quantifies the differences in wireless propagation environments at different locations, calculates the similarity of propagation states based on the features, and constructs a weighted graph model by combining spatial distance and simulated power differences. This achieves physical constraints on the residual propagation process, allowing residual transmission to match the signal attenuation pattern caused by building partitions.

[0049] This invention uses the Laplace smoothing algorithm to solve the residuals of the region to be extrapolated, and then superimposes the constrained residuals onto the corrected simulation data to complete the global map stitching. It fully integrates the global trend of the simulation with the actual local deviation, automatically adapts to the signal extrapolation of large areas of continuous missing regions, and outputs a complete radio map that covers the entire target space and conforms to the building propagation law.

[0050] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0051] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0053] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0054] Figure 1 This is a schematic diagram of the radio map extrapolation method based on simulation correction and residual propagation of the present invention;

[0055] Figure 2This is a verification diagram of the radio map extrapolation method based on simulation correction and residual propagation according to an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram illustrating the division of the target region, the measured region, and the continuously missing region provided by the present invention.

[0057] Figure 4 This is a schematic diagram of the graph-based residual propagation model provided by the present invention;

[0058] Figure 5 This is a schematic diagram of the radio map extrapolation system based on simulation correction and residual propagation of the present invention. Detailed Implementation

[0059] 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. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0060] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0061] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0062] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0063] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0064] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0065] Example 1

[0066] This embodiment mainly describes a radio map extrapolation method based on simulation correction and residual propagation, such as... Figure 1 As shown, it includes the following steps:

[0067] The system acquires building structure information, transmitter location information, and local measured data for the target area, and performs wireless propagation simulation based on the building structure information and transmitter location information to obtain simulation data covering the entire area.

[0068] Based on the spatial distribution of the local measured data, the target area is divided into a measured area and an area to be extrapolated, wherein the area to be extrapolated consists of one or more consecutive missing areas.

[0069] A linear correction model is constructed using the measured data within the measured area and the simulation data at the corresponding location. The simulation data for the entire area is then corrected to obtain the corrected simulation data.

[0070] Calculate the difference between the measured value and the corresponding corrected simulation value at each measured point in the measured area to obtain the simulation-measured residual;

[0071] Based on the building structure information and the transmitter location information, propagation state features characterizing the propagation characteristics of wireless signals are extracted. The propagation state features include at least one of the following: structural visibility status, number of direct-connection wires passing through walls, and spatial partitioning relationship.

[0072] Using the simulation-measured residuals within the measured area as constraints, a graph model is constructed by integrating spatial proximity relationships, the similarity of the propagation state features, and the corrected power differences. The simulation-measured residuals are then propagated from the measured area to the extrapolation area using a residual propagation algorithm to obtain the estimated residuals of the extrapolation area.

[0073] The estimated residuals of the region to be extrapolated are superimposed onto the corresponding corrected simulation data and combined with the measured data of the measured region to generate a radio map of the target region.

[0074] like Figure 2 As shown, the indoor floor plan and scene parameters are obtained. The scene parameters include wall coordinates, room labels, wireless transmitter parameters, and reference point positions.

[0075] Wireless power data for each reference point is generated based on the Wireless InSite ray tracing platform;

[0076] Measured wireless power data were collected at some reference points in the target area;

[0077] Based on the distribution of measured wireless power data in the reference points, the measured area and the area to be extrapolated are determined. The area to be extrapolated consists of at least one consecutive missing area, and each consecutive missing area consists of multiple spatially adjacent reference points to be extrapolated that all lack valid measured radio data.

[0078] Based on the simulated power value and the measured power value at the reference point, the scaling factor and bias term are solved, and the simulated power is linearly corrected.

[0079] Calculate the simulation-measured residual based on the difference between the measured power value at the reference point and the corrected simulated power value;

[0080] Based on the geometric relationship between the wireless transmitter, the reference point and the wall boundary, the spatial distance, the visible state of the structure, the number of straight-through wires and the spatial zoning relationship are extracted.

[0081] Using reference points as graph nodes, a graph Laplace residual propagation model is constructed by combining spatial distance, power difference, and propagation state similarity.

[0082] The residual estimates of each reference point to be extrapolated within the extrapolation area are compensated and corrected, and the simulated power values ​​are combined with the radio data of the measured area to obtain a complete radio map covering the target area.

[0083] Taking an indoor wireless network coverage scenario as an example, the target area is an indoor planar area containing several rooms, corridors, walls, and doors and windows. The wireless transmitter can be a Wi-Fi access point, a cellular small base station, or other wireless transmitting equipment, and the receiver corresponds to discretely set reference points within the target area. In this embodiment, the target area plan, wall coordinates, room labels, wireless transmitter location, transmission power parameters, and reference point location are used as scenario inputs. The measured power can be the received signal strength (RSS) or the reference signal received power (RSRP). For ease of explanation, the power observation value P will be used below to uniformly represent RSS or RSRP data.

[0084] Using the Wireless InSite ray tracing platform, a simulated radio map of the target area is generated based on the actual floor plan scale, wall coordinates, room labels, wireless transmitter location, reference point location, transmission power parameters, and wireless propagation parameters. The simulated radio map includes the simulated power corresponding to each reference point within the complete spatial range of the target area.

[0085] For reference points in the measured area, measured wireless power data is obtained through on-site collection. The measured power can be either the RSS value or the RSRP value.

[0086] Let the set of all reference points within the target area be: ,in, Indicates the first The spatial location of a reference point This represents the total number of reference points within the target area. Based on the distribution of measured power among the reference points, the target area is divided into a measured area and an extrapolation area. The extrapolation area consists of at least one continuously missing region, wherein: , . This indicates that there is a measured area with actual power. This represents the entire extrapolation region consisting of reference points to be extrapolated that lack measured power support, including at least one continuous missing region, such as... Figure 3 As shown.

[0087] To reduce the scale difference and overall bias between simulated radio maps and actual measured radio maps, in the measured area Establish a linear correction model:

[0088]

[0089] in, and They represent the first Simulated power at each reference point and the corrected simulated power Represents the scaling factor. This indicates the bias term.

[0090] Scaling coefficient and bias terms This can be obtained by minimizing the squared error between the corrected simulated power and the measured power in the measured region:

[0091]

[0092] For ease of representation, let

[0093]

[0094] in, Represented as the first Measured power at each reference point Let represent the number of reference points within the measured area. Then, the linear correction parameter can be obtained from the least squares closed-form solution:

[0095]

[0096] Through the above linear correction, the simulated radio map can be adjusted to a numerical scale that is more consistent with the measured radio power, reducing the systematic bias caused by directly using simulation data.

[0097] After completing the simulation prior correction, calculate the simulation-measured residuals for each reference point in the measured area:

[0098]

[0099] in, Indicates the first The simulation-measured residuals at the measured reference points are used to characterize local environmental errors that the corrected simulation radio map still fails to describe, such as differences in wall materials, temporary obstructions, furniture arrangement, differences in equipment response, or local multipath propagation disturbances.

[0100] Based on the building structure information and the wireless signal transmission and reception location relationship in the target area, propagation state characteristics are constructed for each reference point. For a reference point, its propagation state characteristics can be expressed as: .in, Indicates the wireless transmitter and reference point Spatial distance between them This indicates the visible state of the structure between the transmitter and the reference point. (If the wireless transmitter and reference point are connected...) If the straight line segment does not intersect the wall boundary, then take... Otherwise take ; Indicates the connection between the wireless transmitter and the reference point. The number of different walls that the straight segment passes through; This represents the spatial partitioning relationship between the spatial partition where the wireless transmitter is located and the spatial partition where the i-th reference point is located. The spatial partitioning relationship includes relationships within the same room, relationships within the same corridor, relationships between different rooms, and room-corridor relationships.

[0101] like Figure 4 As shown, this application first obtains the planar layout information of the indoor target area, divides the measured area containing the measured points into the extrapolation area containing continuous blank points, and maps all reference points in the area to nodes in the graph structure. Black nodes represent measured points, and white nodes represent extrapolation points. Combining the three propagation state characteristics of visible structure, number of direct lines passing through walls, and spatial partitioning relationship, the connection weights between nodes are determined by combining spatial distance and the difference in corrected simulation power. The residual propagation calculation constrained by the building propagation law is carried out, and finally the estimated residuals of each node in the extrapolation area are obtained. The numerical extrapolation of residuals in the continuously missing area is completed, supporting the generation of a full-area radio map.

[0102] Specifically, based on the location of the wireless transmitter, the location of the reference point, and the wall coordinates on the target area plan, it is determined whether the straight line segment connecting the wireless transmitter and the reference point intersects with the wall boundary. If no intersection occurs, it is determined to be structurally visible; if an intersection occurs, it is determined to be structurally invisible. The number of different walls through which the straight line passes is counted to obtain the number of walls through which the straight line passes. The terms "structurally visible" and "structurally invisible" indicate whether the transmission and reception locations are obstructed by walls or building partitions, which differs from the line-of-sight or non-line-of-sight propagation states caused by furniture, people, or temporary obstacles in the context of wireless propagation. Room labels and spatial area attributes can be obtained from the target area plan or preset spatial partition information.

[0103] In one alternative implementation, the similarity of propagation states can be determined based on the differences between propagation state features. For example, for two reference points... and The difference in propagation state can be expressed as:

[0104]

[0105] in, and This indicates the spatial distance from the wireless transmitter to the corresponding reference point. and Indicates the visible state of the structure. and This indicates the number of straight-through cables passing through walls. and These represent the spatial partitioning relationships between the spatial partition where the wireless transmitter is located and the spatial partitions where the i-th and j-th reference points are located, respectively. For indicator functions, and For normalization parameters, , , , Here are the weighting coefficients. The similarity of propagation states can be expressed as:

[0106]

[0107] A residual propagation weighted graph is constructed by combining spatial proximity, similarity of propagation states, and power differences between the corrected simulated powers. Reference points within the target region are used as graph nodes; if two reference points satisfy the neighborhood connectivity condition, a graph edge is established between them. (Reference points) and The edge weights between them can be represented as:

[0108]

[0109] in, and Indicates the spatial location of node i and node j. This represents a spatial distance scale parameter. The power difference scale parameter represents the absolute value of the difference between the corrected simulated power at node i and node j.

[0110] set up For the graph adjacency matrix, Let be a degree matrix, where The Graph Laplace matrix is To estimate contiguous missing regions The residuals in the data can be propagated using a semi-supervised graphical propagation model. Let the residual vector across the entire region be... The observed residuals for the measured area are The following optimization objective is constructed:

[0111]

[0112] in, Represents the observation constraint weights. This indicates the strength of the graph smoothing constraint.

[0113] The above optimization objective can be written in matrix form: .in, Let be the diagonal matrix formed by the observation constraint weights. Its closed-form solution can be expressed as: Using the above graph Laplace propagation model, we can obtain the residual estimates for each reference point in the continuously missing region.

[0114] Reference points in a contiguous missing region Its predicted radio power can be expressed as:

[0115]

[0116] in, This represents the final predicted power of the i-th reference point in the region to be extrapolated. This represents the residual estimation result obtained through the propagation state-constrained residual propagation model.

[0117] In another alternative implementation, the residual propagation model can also be implemented using a learning model. Specifically, the spatial location of each reference point, the corrected simulation power, and the propagation state characteristics can be used as inputs, and the simulation-measured residuals in the measured area can be used as supervision labels to train the residual estimation model.

[0118]

[0119] in, This represents the residual estimation model to be trained. The parameters represent the model parameters. The residual estimation model can be implemented using models such as multilayer perceptron, graph neural network, random forest, or gradient boosting tree.

[0120] In a specific model example, a graph neural network can be used as the residual estimation model. The graph neural network takes a reference point graph structure as input and node features as input. As the initial node features, the latent representation of node i at layer k is updated through neighborhood aggregation: ,in To reference point The set of neighboring nodes connected to the corresponding graph node. The model output is the estimated node residual. For the measured region, the model training loss can be expressed as: .

[0121] The above-described graph Laplacian propagation method is the primary embodiment, while the graph neural network residual estimation method is an optional embodiment. Both can estimate the residuals of each reference point to be extrapolated in a continuously missing region based on the spatial proximity between reference points, the similarity of propagation states, and the power differences between the corrected simulated powers. Generate radio map extrapolation results for continuously missing regions.

[0122] This embodiment obtains the global simulation prior through ray tracing, eliminates the overall simulation bias through linear correction, extracts multi-dimensional building propagation features to construct a weighted graph model to constrain residual propagation, uses graph Laplace smoothing to calculate the residuals of missing areas, and stitches together the superimposed compensation to generate a complete radio map. Only a small amount of measured data is needed to adapt to large areas of continuous missing areas, which conforms to the actual wireless propagation distribution pattern.

[0123] Example 2 describes in detail a radio map extrapolation system based on simulation correction and residual propagation, as follows: Figure 5 As shown, it includes:

[0124] The data acquisition module acquires building structure information, transmitter location information, and local measured data of the target area, and performs ray tracing wireless propagation simulation based on the building structure information and transmitter location information, outputting simulation data covering all reference points in the entire area;

[0125] The region division module divides the measured region and the extrapolation region, which consists of several consecutive missing regions, based on the distribution of local measured data at the reference point.

[0126] The simulation correction module extracts measured data and simulation data of the same reference point within the measured area, constructs and solves a global linear correction model, and uniformly corrects the simulation data across the entire domain to obtain the corrected simulation data.

[0127] The residual calculation module iterates through all reference points in the measured area, calculates the simulation-measured residuals for each point based on the measured data and the corrected simulation data, and generates a residual dataset.

[0128] The feature extraction module combines the coordinates of the building walls and the geometric positions of the transmitter and reference points to extract the structural visibility status, number of straight lines passing through the walls, and spatial zoning relationship propagation status features of each reference point.

[0129] The residual propagation module uses the residual dataset as a constraint and combines spatial proximity, similarity of propagation state, and difference of power after correction to construct a weighted graph model. The estimated residuals of each reference point in the region to be extrapolated are solved by the graph Laplace smoothing algorithm.

[0130] The map generation module overlays the estimated residuals of each reference point in the area to be extrapolated onto the corresponding corrected simulation data, combines the predicted data of the extrapolated area with the measured data of the measured area, and outputs a complete radio map of the target area.

[0131] Furthermore, the data acquisition module, as the system's data input, uniformly collects architectural structural information such as building outlines, wall coordinates, and room partitions required by the scene, synchronously inputs the spatial coordinate parameters of the wireless transmitting equipment, and simultaneously receives local wireless signal measurement data collected on-site. The module is equipped with a ray tracing simulation computing unit, which reads all scene parameters, traverses all grid reference points in the target area, batch calculates the wireless signal simulation values ​​corresponding to each point, and outputs a complete simulation dataset covering the entire target space, providing full-domain basic signal data for all subsequent computing stages.

[0132] Furthermore, the region division module receives the simulation data and local measured data output by the data acquisition module. Using a preset uniform grid reference point as the division benchmark, it searches for whether there is a valid measured signal value at each reference point. All reference points with measured data are aggregated to form the measured region. For reference points without any measured values ​​and whose spatial locations are adjacent, clustering is performed. Each clustering result is defined as a continuous missing region. All continuous missing regions are integrated to form the region to be extrapolated, completing the boundary division and point marking of the two types of regions in the target space. The division results are synchronously transmitted to the simulation correction module.

[0133] Furthermore, the simulation correction module reads the point information of the divided measured area, extracts the simulation values ​​and measured values ​​of the same reference point one by one, constructs a linear transformation model containing scale coefficient and global bias term based on the paired data, solves the two parameters in the model through least squares error optimization, applies the solved linear model to the original simulation data of all reference points in the target area, uniformly completes the scale and offset correction of the global simulation values, and outputs the corrected simulation dataset after eliminating global system bias, which is pushed to the residual calculation module and feature extraction module respectively.

[0134] Furthermore, the residual calculation module only retrieves all reference point data within the measured area. For each point, it matches and reads the measured signal value with the simulated value after linear correction, calculates the difference between the two as the simulated measured residual corresponding to that point, and summarizes all the residual values ​​of the measured points to form a standardized residual dataset. The dataset completely stores the residual values ​​of each point and the corresponding spatial coordinates, which are transmitted to the residual propagation module as known constraints for the residual propagation operation.

[0135] Furthermore, the feature extraction module reads the coordinates of the building walls, the coordinates of the transmitter, and the spatial coordinates of all reference points. Through the geometric line segment intersection judgment logic, it determines whether the line connecting the transmitter and each reference point passes through the wall to obtain the visible state of the structure. It counts the number of independent walls that the line passes through to obtain the number of walls passed through. It matches the building room partition labels to determine the spatial partition relationship of the transmitter and receiver points. The three types of information are integrated into a propagation state feature vector unique to each reference point. All feature vectors are simultaneously sent to the residual propagation module for weight calculation.

[0136] Furthermore, the residual propagation module uses all reference points as independent nodes in the graph structure. It calculates the connection weights between nodes based on three indicators: spatial distance between nodes, similarity calculated from the propagation state feature vector, and the difference between the corrected simulation values. This forms a graph model with weighted constraints. Using the residual dataset of measured points as fixed observation constraints, it constructs a graph Laplacian smoothing optimization function. Through matrix closed-form solution operations, it solves for the residual estimates corresponding to all reference points to be extrapolated. Finally, it transmits the residual results of all missing regions to the map generation module.

[0137] Furthermore, the map generation module receives the residual estimates of each point in the area to be extrapolated and the global correction simulation data. For each reference point to be extrapolated, the residual estimate is superimposed on the correction simulation data of the same point to obtain the final predicted signal data of that point. All the predicted data of the area to be extrapolated and the original measured data of the measured area are stitched together according to spatial coordinates to generate a radio map covering the entire target area and output it to the outside.

[0138] This application provides a computer device comprising two core hardware units: a processor and a memory. The memory stores a complete radio map extrapolation computer program, and the processor establishes a data read / write communication link with the memory. During device operation, the processor retrieves the program from memory and executes the entire computation process sequentially, completing all steps including scene data acquisition, ray tracing simulation, region division, linear correction, residual calculation, propagation feature extraction, weighted graph model construction, graph Laplacian residual estimation, and map stitching output. The device can be mounted on indoor testing terminals, base station backend servers, and other hardware carriers. It can automatically output a complete radio map by inputting only building structure, transmission point location, and local measured data. The hardware architecture is adaptable to the offline computation needs of various wireless communication scenarios.

[0139] This application provides a computer-readable storage medium, which is a standardized storage carrier capable of storing computer programs. The medium internally stores the complete program code implementing a radio map extrapolation method. The medium can be stored independently or installed in various computers and embedded devices. Any hardware with program reading capabilities can read the code within the medium and execute the entire set of computational logic, including global simulation, region classification, simulation linear correction, residual dataset generation, multidimensional propagation feature extraction, multi-constraint graph residual propagation, and map fusion output. The medium is not limited by a single hardware device and can be deployed and used across terminals, facilitating solution reuse, portability, and offline deployment.

[0140] This embodiment realizes the entire process of map extrapolation. Each module works in sequence to complete data simulation, region division, simulation correction, residual calculation, feature extraction, residual propagation and map generation. The entire set of calculations is completed automatically without manual step-by-step processing. The modules are independent and easy to split and debug. It can be adapted to various indoor and park wireless scenarios to quickly output complete radio maps.

[0141] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A radio map extrapolation method based on simulation correction and residual propagation, characterized in that, Includes the following steps: The system acquires building structure information, transmitter location information, and local measured data for the target area, and performs wireless propagation simulation based on the building structure information and transmitter location information to obtain simulation data covering the entire area. Based on the spatial distribution of the local measured data, the target area is divided into a measured area and an area to be extrapolated, wherein the area to be extrapolated consists of one or more consecutive missing areas. A linear correction model is constructed using the measured data within the measured area and the simulation data at the corresponding location. The simulation data for the entire area is then corrected to obtain the corrected simulation data. Calculate the difference between the measured value and the corresponding corrected simulation value at each measured point in the measured area to obtain the simulation-measured residual; Based on the building structure information and the transmitter location information, propagation state features characterizing the propagation characteristics of wireless signals are extracted. The propagation state features include at least one of the following: structural visibility status, number of direct-connection wires passing through walls, and spatial partitioning relationship. Using the simulation-measured residuals within the measured area as constraints, a graph model is constructed by integrating spatial proximity relationships, the similarity of the propagation state features, and the corrected power differences. The simulation-measured residuals are then propagated from the measured area to the extrapolation area using a residual propagation algorithm to obtain the estimated residuals of the extrapolation area. The estimated residuals of the region to be extrapolated are superimposed onto the corresponding corrected simulation data and combined with the measured data of the measured region to generate a radio map of the target region.

2. The radio map extrapolation method based on simulation correction and residual propagation according to claim 1, characterized in that, Several reference points are pre-distributed in a uniform grid pattern within the target area, and all reference points together constitute the complete sampling grid of the target area; only some reference points collect local measured data, corresponding to the measured area. Reference points for which no local measured data was collected are clustered according to spatial adjacency to form continuous missing regions. All continuous missing regions together constitute the region to be extrapolated. ,satisfy , ; A single contiguous missing region consists of at least two spatially adjacent reference points that are to be extrapolated and have no valid local measured data. The execution process of the wireless propagation simulation includes: building a simulation scene based on building structure information, recording the transmitter location, frequency band, transmission power, antenna parameters, and propagation model parameters in the simulation scene, and using a ray tracing algorithm to traverse all reference points to calculate the simulation data corresponding to each reference point.

3. The radio map extrapolation method based on simulation correction and residual propagation according to claim 2, characterized in that, The linear correction model is a globally unified linear transformation model, and its expression is: In the formula, Reference point numbering, For the first Simulation data at one reference point, For the first Simulation data corrected for each reference point The scaling factor. This is a global bias term; The scaling coefficient is solved with the objective of minimizing the squared error between the corrected simulation data and the measured data for all reference points in the measured area. Bias terms The optimization objective formula is: In the formula, For the first Measured data collected from several reference points Representing the Each reference point belongs to the measured area; The scale coefficient obtained by the solution Bias terms Substitute the linear correction model and traverse all reference points in the target area to complete the global simulation data correction.

4. The radio map extrapolation method based on simulation correction and residual propagation according to claim 3, characterized in that, The simulation-measured residual is calculated only for reference points within the measured area that have measured data. The residual calculation formula is as follows: In the formula, For the first Simulation-measured residuals corresponding to each measured reference point; The simulation-measured residual is used to characterize the local numerical deviation between the simulation data after linear correction and the real environment caused by wall materials, local occlusion, equipment response differences, and multipath disturbances. The residuals corresponding to all measured reference points together form the residual dataset.

5. The radio map extrapolation method based on simulation correction and residual propagation according to claim 4, characterized in that, Similarity of propagation states between pairwise reference points Calculated using the following formula: Differences in propagation status ; , From the transmitter to the reference point , Spatial distance, , The structure is visible. , This refers to the number of straight-through cables passing through walls. , Spatial partitioning relationship, For indicator functions, , For normalization parameters, , , , These are the weighting coefficients.

6. The radio map extrapolation method based on simulation correction and residual propagation according to claim 5, characterized in that, The graph model uses each reference point as an independent graph node, establishing connections only between graph nodes within their spatial neighborhood, with connection weights between two nodes. The calculation formula is: ； In the formula, , For reference point , spatial coordinates, For spatial distance scale parameters, This is the power difference scale parameter.

7. The radio map extrapolation method based on simulation correction and residual propagation according to claim 6, characterized in that, The graph Laplace smoothing algorithm is used to solve for the estimated residuals of all graph nodes in the region to be extrapolated, and a residual smoothing optimization objective is constructed: ； In the formula, For observation constraint weights, Given the known residuals at the measured reference point, To smooth the constraint strength, The target is the entire domain; After converting the optimization objective into matrix form, solve for the closed-form solution: , To observe the constraint weight diagonal matrix, For the graph Laplace matrix, For the graph adjacency matrix The corresponding degree matrix.

8. A radio map extrapolation system based on simulation correction and residual propagation, characterized in that, A method for performing radio map extrapolation according to any one of claims 1 to 7 includes: The data acquisition module acquires building structure information, transmitter location information, and local measured data of the target area, and performs ray tracing wireless propagation simulation based on the building structure information and transmitter location information, outputting simulation data covering all reference points in the entire area; The region division module divides the measured region and the extrapolation region, which consists of several consecutive missing regions, based on the distribution of local measured data at the reference point. The simulation correction module extracts measured data and simulation data of the same reference point within the measured area, constructs and solves a global linear correction model, and uniformly corrects the simulation data across the entire domain to obtain the corrected simulation data. The residual calculation module iterates through all reference points in the measured area, calculates the simulation-measured residuals for each point based on the measured data and the corrected simulation data, and generates a residual dataset. The feature extraction module combines the coordinates of the building walls and the geometric positions of the transmitter and reference points to extract the structural visibility status, number of straight lines passing through the walls, and spatial zoning relationship propagation status features of each reference point. The residual propagation module uses the residual dataset as a constraint and combines spatial proximity, similarity of propagation state, and difference of power after correction to construct a weighted graph model. The estimated residuals of each reference point in the region to be extrapolated are solved by the graph Laplace smoothing algorithm. The map generation module overlays the estimated residuals of each reference point in the area to be extrapolated onto the corresponding corrected simulation data, combines the predicted data of the extrapolated area with the measured data of the measured area, and outputs a complete radio map of the target area.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program, which, when executed by the processor, implements the radio map extrapolation method based on simulation correction and residual propagation as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the radio map extrapolation method based on simulation correction and residual propagation as described in any one of claims 1 to 7.

Images