Method and apparatus for constructing an electromagnetic spectrum map based on multi-precision monitoring data

Abstract

The application relates to a method and device for constructing an electromagnetic spectrum map based on multi-precision monitoring data, which comprises the following steps: firstly, collecting monitoring data with different precisions through different types of sensors, and fitting a trend surface through a generalized regression neural network to obtain an estimation of a path attenuation component of an electromagnetic wave propagation; then, subtracting the monitoring data from the path attenuation component estimation value corresponding to the position to obtain residual monitoring data; then, fitting a variation function according to a small amount of high-precision residual monitoring data, and only using the high-precision variation function when constructing a co-Kriging system; then, using a large amount of low-precision residual monitoring data to supplement the small amount of high-precision residual monitoring data to obtain an estimation of a shadow attenuation component through co-Kriging interpolation; and finally, adding the two component estimations to obtain the overall electromagnetic spectrum map construction result, so that the construction precision is improved.

Description

Technical Field

[0001] This application relates to the field of electromagnetic spectrum map construction technology, and in particular to a method and apparatus for constructing electromagnetic spectrum maps based on multi-precision monitoring data. Background Technology

[0002] An electromagnetic spectrum map, also known as a radio environment map or radio map, quantitatively describes the electromagnetic environment from multiple dimensions, including time, space, frequency, and energy, and visualizes it using a geographic information system (GIS). Electromagnetic spectrum maps provide fundamental information for numerous military and civilian applications, such as electromagnetic space cognition, control, and utilization. Their main uses are fourfold: first, to obtain information on the distribution of electromagnetic spectrum resources, supporting refined and dynamic allocation of spectrum resources and achieving efficient utilization; second, to obtain information on the distribution of electromagnetic radiation, providing quantitative data for electromagnetic pollution monitoring and prevention; third, to obtain information on electromagnetic interference behavior patterns, supporting the investigation and handling of illegal activities such as unauthorized radio broadcasts and fake base stations, and maintaining frequency usage order; and fourth, in military applications, electromagnetic spectrum maps can provide auxiliary decision-making information for electromagnetic situational awareness generation, equipment deployment and operation, operational frequency planning, and frequency conflict resolution. Therefore, accurately constructing electromagnetic spectrum maps is of paramount importance.

[0003] The accuracy of electromagnetic spectrum mapping is highly dependent on the quantity and quality of the measurement dataset, which typically includes RSS measurements. Therefore, more efficient and accurate measurement techniques are needed to improve the accuracy of electromagnetic spectrum mapping. Fortunately, the development of wireless sensor networks offers an effective solution to this challenge, as it enables multi-precision measurements of electromagnetic physical quantities in space using ubiquitous sensing technologies. Ubiquitous sensing provides a seamless communication channel between distributed wireless sensors, allowing for real-time data collection and analysis.

[0004] While a wealth of measurement data has been acquired through wireless sensors, high-quality data remains a challenge for electromagnetic spectrum mapping. Non-dedicated sensors result in a lack of accuracy, while expensive high-precision sensors can only be deployed in a limited number of locations. Therefore, a large portion of the dataset consists of low-precision measurements from low-precision sensors, failing to provide the same accuracy as specialized equipment. The term "low-precision" refers to two main characteristics: 1) amplitude reduction due to incompatible antennas, and 2) white noise caused by thermal effects. However, the substantial amount of low-precision data can help compensate for the scarcity of high-precision data. Therefore, a key issue in achieving high-precision electromagnetic spectrum mapping under ubiquitous sensing conditions is integrating data with varying levels of accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and apparatus for constructing an electromagnetic spectrum map based on multi-precision monitoring data to address the aforementioned technical problems.

[0006] A method for constructing an electromagnetic spectrum map based on multi-precision monitoring data, the method comprising:

[0007] Acquire multi-precision monitoring data of the electromagnetic environment within the region of interest and the location of the point to be estimated; the multi-precision monitoring data includes high-precision and low-precision monitoring data, and each set of data includes: location information and the received power at the corresponding location.

[0008] The data formats of high-precision and low-precision monitoring data will be standardized.

[0009] Based on the standardized high-precision and low-precision monitoring data and the location of the point to be estimated, a trend surface is fitted to obtain the path attenuation component estimate of the location of the point to be estimated.

[0010] Based on the standardized high-precision monitoring data and low-precision monitoring data, and the path attenuation component estimation value, high-precision residual data and low-precision residual data are obtained.

[0011] A co-kriging system is constructed based on the variation function fitted using the high-precision residual data.

[0012] By using the low-precision residual data and the high-precision residual data through co-kriging interpolation, the estimated value of the shadow fading component at the location of the point to be estimated is obtained.

[0013] The path attenuation component estimate and the shadow fading component estimate are added together to obtain the received power estimate for the location of the point to be estimated.

[0014] In one embodiment, based on the standardized high-precision monitoring data and low-precision monitoring data, and the location of the point to be estimated, trend surface fitting is performed to obtain the path attenuation component estimate of the location of the point to be estimated, including:

[0015] The location information and the corresponding received power in the high-precision monitoring data and low-precision monitoring data after the format is unified are used as training samples.

[0016] Construct a generalized regression neural network; train the generalized regression neural network based on the training samples to obtain a path attenuation component estimation model for the location of the point to be estimated.

[0017] The location of the point to be estimated is input into the path attenuation component estimation model to obtain the path attenuation component estimate of the location of the point to be estimated.

[0018] In one embodiment, the generalized regression neural network includes an input layer, a pattern layer, a summation layer, and an output layer.

[0019] The transfer function of the input layer is a linear function.

[0020] The pattern layer is a radial basis layer, and the activation function used inside each neuron is a radial basis function.

[0021] The summation layer includes two types of neurons: numerator units and denominator units. The numerator units are used to calculate the weighted sum of the neurons in the pattern layer, and the denominator units are used to calculate the algebraic sum.

[0022] The output layer is used to calculate the path decay component estimate based on the trend surface fitting result; the trend surface fitting result is expressed as the weighted sum divided by the algebraic sum.

[0023] In one embodiment, based on the standardized high-precision monitoring data and low-precision monitoring data, and the location of the point to be estimated, trend surface fitting is performed to obtain the estimated path attenuation component value of the location of the point to be estimated:

[0024]

[0025] Among them, P t For the radiated power of the signal source, l t Let η be the location of the signal source, and η be the path attenuation exponent. t -l r ||For l t With l r The distance between them, d0 is the far-field reference distance of the antenna, G dB =20log 10 (4πd0 / λ) is the path attenuation factor, which is a function of the operating wavelength λ of the radiated signal.

[0026] In one embodiment, based on the spatial correlation between the shadow fading model and the shadow fading components, the high-precision residual data is used to fit a variation function, including: the variation function is:

[0027]

[0028] Where, γ ij Let C0 be the value of the covariance function, C0 be the nugget value, C be the arch height, and l be the value of the nugget. i l j These are the position vectors for the i-th and j-th positions, respectively.

[0029] In one embodiment, the spatial correlation of the shadow fading component is:

[0030]

[0031] Where, σ c d is the standard deviation. c To solve for the correlation distance.

[0032] In one embodiment, the co-kriging system is:

[0033]

[0034] in, and For Lagrange multipliers, These are the Kriging weighting coefficients. These are high-precision and low-precision monitoring data, respectively. This represents the variation function value between two data sampling points of different precisions. This represents the variation function value between high-precision data sampling points and unknown points. This represents the variation function values ​​between high-precision data sampling points. Let N be the variation function value between the low-precision data sampling points, N be the number of high-precision monitoring data, and M be the number of low-precision monitoring data.

[0035] The matrix form of the co-kriging system is as follows:

[0036]

[0037] in,

[0038] In one embodiment, the shadow fading component estimate of the point location to be estimated is obtained by using the low-precision residual data and the high-precision residual data through co-kriging interpolation:

[0039]

[0040]

[0041]

[0042] in, Γ * This is a matrix form of the variogram values ​​for high-precision data where multi-precision data are used as independent variables. It is represented as the variation function value between the unknown point and the multi-precision sensor.

[0043] The co-kriging variance of the shadow fading component estimation can be expressed as:

[0044]

[0045] in, Co-kriging variance estimated for shadow fading components.

[0046] In one embodiment, the low-precision monitoring data has an additional measurement error compared to the high-precision monitoring data, and the measurement error follows a Gaussian distribution:

[0047]

[0048] Where, μ L The attenuation is due to antenna mismatch. White noise is caused by the thermal effect of inexpensive electronic components.

[0049] An electromagnetic spectrum map construction device based on multi-precision monitoring data, the device comprising:

[0050] The monitoring data acquisition module is used to acquire multi-precision monitoring data of the electromagnetic environment within the region of interest and the location of the point to be estimated; the multi-precision monitoring data includes high-precision and low-precision monitoring data, and each set of data includes: location information and the received power at the corresponding location.

[0051] The path attenuation component estimation module is used to unify the data format of high-precision and low-precision monitoring data; based on the unified high-precision and low-precision monitoring data and the location of the point to be estimated, a trend surface is fitted to obtain the path attenuation component estimate of the location of the point to be estimated.

[0052] The shadow fading component estimation module is used to obtain high-precision residual data and low-precision residual data based on the high-precision monitoring data and low-precision monitoring data after format unification and the path attenuation component estimation value; to construct a co-kriging system based on the variation function fitted by the high-precision residual data; and to obtain the shadow fading component estimation value of the point to be estimated by using the low-precision residual data and the high-precision residual data through co-kriging interpolation.

[0053] The module for determining the received power estimate of the location to be estimated is used to add the path attenuation component estimate and the shadow fading component estimate to obtain the received power estimate of the location to be estimated.

[0054] The aforementioned method and apparatus for constructing an electromagnetic spectrum map based on multi-precision monitoring data includes the following steps: First, monitoring data of varying precision is collected using different types of sensors. This data is then fitted to a trend surface using a generalized regression neural network to estimate the path attenuation component of electromagnetic wave propagation. Next, the difference between the monitoring data and the path attenuation component estimates at corresponding locations is calculated to obtain residual monitoring data. Then, a variogram function is fitted based on a smaller amount of high-precision residual monitoring data. Only the high-precision variogram function is used when constructing the co-kriging system. Finally, the smaller amount of high-precision monitoring residual data is supplemented by a larger amount of low-precision monitoring residual data, and co-kriging interpolation is used to obtain an estimate of the shadow attenuation component. Finally, the two component estimates are added together to obtain the overall electromagnetic spectrum map construction result, thus improving the construction accuracy. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating a method for constructing an electromagnetic spectrum map based on multi-precision monitoring data in one embodiment.

[0056] Figure 2 This is a flowchart of an electromagnetic spectrum map construction method based on multi-precision monitoring data in another embodiment;

[0057] Figure 3 In another embodiment, the generalized regression neural network (GRNN) model is used.

[0058] Figure 4 Comparison of root mean square error under different low-precision monitoring data volumes in another embodiment;

[0059] Figure 5 A comparison of root mean square error under different amounts of high-precision monitoring data in another embodiment;

[0060] Figure 6 This is a structural block diagram of an electromagnetic spectrum map construction device based on multi-precision monitoring data in one embodiment. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0062] This application proposes a method for constructing electromagnetic spectrum maps based on multi-precision monitoring data, employing a co-Kriging scheme with high-precision variogram (HV-CK) as a solution for building high-precision electromagnetic spectrum maps from multi-precision data. This method involves a comprehensive analysis of the multi-precision data. On one hand, it fits the variogram to the high-precision data; on the other hand, it constructs a co-Kriging system from the multi-precision data to ensure accuracy. Furthermore, trend surface fitting is used to represent the deterministic components of the propagation model, thus satisfying the inherent assumptions of Kriging-like methods. By using HV-CK, the proposed method can construct high-precision electromagnetic spectrum maps.

[0063] In one embodiment, such as Figure 1 , Figure 2 As shown, a method for constructing an electromagnetic spectrum map based on multi-precision monitoring data is provided. The method includes the following steps:

[0064] Step 100: Obtain multi-precision monitoring data of the electromagnetic environment within the region of interest and the location of the point to be estimated; the multi-precision monitoring data includes high-precision and low-precision monitoring data, and each set of data includes: location information and the received power at the corresponding location.

[0065] Specifically, multi-precision monitoring data sources are based on ubiquitous sensing technology and have broad application prospects in urban areas. In particular, the increase in the deployment of 5G base stations makes the collection of multi-precision monitoring data in the radio spectrum increasingly convenient.

[0066] High-precision monitoring data is acquired through high-precision sensors, while low-precision monitoring data is acquired through low-precision sensors.

[0067] In mathematical notation, this means utilizing a finite amount of monitoring data within the region of interest (l), including the receiving power of a high-precision sensor. and its location information Received power of low-precision sensors and its location information Estimated area Received power at any location without a high-precision sensor deployed

[0068] Step 102: Unify the data format of high-precision and low-precision monitoring data.

[0069] Step 104: Based on the high-precision monitoring data and low-precision monitoring data after standardization and the location of the point to be estimated, perform trend surface fitting to obtain the estimated value of the path attenuation component of the location of the point to be estimated.

[0070] Specifically, the electromagnetic spectrum map construction is decomposed into path decay and shadow fading components. The path decay component is fitted with a trend surface using a generalized regression neural network, thereby improving the construction accuracy of Kriging-like methods.

[0071] Step 106: Based on the high-precision monitoring data and low-precision monitoring data after standardization and the path attenuation component estimation, obtain high-precision residual data and low-precision residual data.

[0072] Specifically, after solving for the path decay component by fitting the trend surface, the multi-precision residual monitoring data can be obtained as follows:

[0073]

[0074]

[0075]

[0076] Step 108: Construct a co-kriging system based on the variation function fitted with high-precision residual data.

[0077] Specifically, by utilizing a limited amount of high-precision monitoring data, a high-precision variation function model is fitted, and a co-kriging system is constructed based on this variation function.

[0078] By leveraging the characteristics of multi-precision monitoring data, a variogram is constructed using high-precision monitoring data, and a high-precision variogram model is fitted. Using this variogram as the core, a co-kriging system is constructed using low-precision monitoring data, thus compensating for the lack of quantity in high-precision monitoring data and improving the overall construction accuracy.

[0079] Step 110: Using low-precision residual data and high-precision residual data, co-kriging interpolation is used to obtain the estimated value of the shadow fading component at the location of the point to be estimated.

[0080] Specifically, a variation function is fitted based on high-precision residual monitoring data, and only the high-precision variation function is used when constructing the co-kriging system. Then, the shadow attenuation component is estimated by co-kriging interpolation using high- and low-precision residual monitoring data.

[0081] The estimated value of the shadow fading component at the location to be estimated can be expressed as:

[0082]

[0083] Among them, parameters It is obtained through co-kriging interpolation.

[0084] Step 112: Add the path attenuation component estimate and the shadow fading component estimate to obtain the received power estimate of the location to be estimated.

[0085] Specifically, the estimated received power at the location to be estimated is:

[0086]

[0087] The above-mentioned method for constructing an electromagnetic spectrum map based on multi-precision monitoring data includes the following steps: First, monitoring data of different precisions are collected using different types of sensors. This data is then fitted to a trend surface using a generalized regression neural network to obtain an estimate of the path attenuation component of radio wave propagation. Next, the difference between the monitoring data and the path attenuation component estimate at the corresponding location is calculated to obtain residual monitoring data. Then, a variogram function is fitted based on a small amount of high-precision residual monitoring data. Only the high-precision variogram function is used when constructing the co-kriging system. Finally, the small amount of high-precision monitoring residual data is supplemented by a larger amount of low-precision monitoring residual data, and co-kriging interpolation is used to obtain an estimate of the shadow attenuation component. Finally, the two component estimates are added together to obtain the overall electromagnetic spectrum map construction result, thus improving the construction accuracy.

[0088] In one embodiment, step 104 includes: using the location information and corresponding received power at the location in the high-precision monitoring data and low-precision monitoring data after format unification as training samples; constructing a generalized regression neural network; training the generalized regression neural network according to the training samples to obtain a path attenuation component estimation model for the location of the point to be estimated; and inputting the location of the point to be estimated into the path attenuation component estimation model to obtain the path attenuation component estimate value of the location of the point to be estimated.

[0089] Specifically, fitting the trend surface of path loss and providing accurate measurement residuals to influence fading prediction are key steps. If the modeling results are too similar to the measurements, the measurement residuals may negatively impact the accuracy of the kriging prediction. However, if the modeling results are dissimilar to the measurements, they may fail to accurately fit the components of path loss. Therefore, to properly model path loss, this paper utilizes the GRNN algorithm for trend surface fitting.

[0090] In one embodiment, the generalized regression neural network includes an input layer, a pattern layer, a summation layer, and an output layer; the transfer function of the input layer is a linear function; the pattern layer is a radial basis layer, and the activation function used inside each neuron is a radial basis function; the summation layer includes two types of neurons, namely numerator units and denominator units; the numerator units are used to calculate the weighted sum of the neurons in the pattern layer, and the denominator units are used to calculate the algebraic sum; the output layer is used to calculate the path decay component estimate based on the trend surface fitting result; the trend surface fitting result is represented as a weighted sum divided by the algebraic sum.

[0091] Specifically, a generalized regressive neural network (GRNN) consists of four layers: an input layer, a pattern layer, a summation layer, and an output layer, such as... Figure 3 As shown. The input layer contains neurons whose transfer functions are linear functions, and their number is equal to the dimension of the input vector. The pattern layer is the radial base layer, where each neuron uses a radial basis function as its activation function, typically a Gaussian function:

[0092]

[0093] Among them, c i It is the center vector, σ g This is the smoothness parameter.

[0094] The summation layer contains two types of neurons: numerator units and denominator units. The numerator units calculate the weighted sum of the neurons in the pattern layer, while the denominator units calculate the algebraic sum. The trend surface fitting result is represented as the weighted sum divided by the algebraic sum. Finally, the output layer is used to calculate the predicted value based on the trend surface fitting result.

[0095] In one embodiment, based on the standardized high-precision monitoring data and low-precision monitoring data, and the location of the point to be estimated, a trend surface fitting is performed to obtain the estimated path attenuation component value of the location of the point to be estimated:

[0096]

[0097] Among them, P t For the radiated power of the signal source, l t Let η be the location of the signal source, and η be the path attenuation exponent. t -l r ||For l t With l r The distance between them, d0 is the far-field reference distance of the antenna, G dB =20log 10 (4πd0 / λ) is the path attenuation factor, which is a function of the operating wavelength λ of the radiated signal.

[0098] In one embodiment, based on the spatial correlation between the shadow fading model and the shadow fading components, a variation function is fitted using high-precision residual data, including: the variation function is an exponential model, and the variation function is:

[0099]

[0100] Where, γ ij Let C0 be the value of the covariance function, C0 be the nugget value, C be the arch height, and l be the value of the nugget. i l j These are the position vectors for the i-th and j-th positions, respectively.

[0101] In one embodiment, the spatial correlation of the shadow fading component is:

[0102]

[0103] Where, σ c d is the standard deviation. c To solve for the correlation distance.

[0104] In one embodiment, the co-kriging system is:

[0105]

[0106] in, and For Lagrange multipliers, These are the Kriging weighting coefficients. These are high-precision and low-precision monitoring data, respectively. This represents the variation function value between two data sampling points of different precisions. This represents the variation function value between high-precision data sampling points and unknown points. This represents the variation function values ​​between high-precision data sampling points. Let N be the variation function value between the low-precision data sampling points, N be the number of high-precision monitoring data, and M be the number of low-precision monitoring data.

[0107] The matrix form of the co-kriging system is:

[0108]

[0109] in,

[0110] In one embodiment, step 110 includes: using low-precision residual data and high-precision residual data through co-kriging interpolation, obtaining an estimate of the shadow fading component at the location of the point to be estimated as follows:

[0111]

[0112]

[0113]

[0114] in, Γ * This is a matrix form of the variogram values ​​for high-precision data where multi-precision data are used as independent variables. It is represented as the variation function value between the unknown point and the multi-precision sensor.

[0115] The co-kriging variance of the shadow fading component estimation can be expressed as:

[0116]

[0117] in, Co-kriging variance estimated for shadow fading components.

[0118] Specifically, the estimated values ​​of the shadow fading components in the unknown region are shown in Equation (4). The parameters in Equation (4) are... The parameters satisfy the co-kriging system constraints shown in equation (10).

[0119] The co-kriging system shown in equation (10) can also be expressed in matrix form as shown in equation (11). Equation (4) is rewritten in matrix form as shown in equation (12) for calculation.

[0120] In one embodiment, the low-precision monitoring data has an additional measurement error compared to the high-precision monitoring data. This measurement error follows a Gaussian distribution.

[0121]

[0122] Where, μ L The attenuation is due to antenna mismatch. White noise is caused by the thermal effect of inexpensive electronic components.

[0123] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0124] In a summary of verification embodiments, considering that the accuracy of electromagnetic spectrum map construction is crucial in specific applications such as electromagnetic situation generation, the root mean square error (RMSE) is selected as the evaluation index. Co-Kriging, Ordinary Kriging-H based on high-precision data, Ordinary Kriging-L based on low-precision data, Ordinary Kriging-H based on multi-precision data, inverse distance-weighted (IDW), and Nearest Neighbors (NN) are used as reference methods for performance comparison analysis with the method of this invention (HV-CK).

[0125] Figure 4 The root mean square error (RMSE) of the proposed method and the reference method is compared under different low monitoring data conditions. The high-precision method uses 5 sensors, the low-precision method has an attenuation of 5 dB, and the low-precision method has a standard deviation of 4 dB. Figure 4 It can be seen that: (1) The overall trend of RMSE for all methods decreases as the amount of data increases. This indicates that the more low-precision monitoring data obtained, the higher the overall accuracy of electromagnetic spectrum map construction. (2) Under the same data volume conditions, the RMSE of the method of the present invention is smaller than that of the reference method, indicating that when constructing electromagnetic spectrum maps using the same amount of multi-precision monitoring data, the method of the present invention can achieve higher construction accuracy than the reference method, and when the amount of low-precision monitoring data is sufficient, the advantage of the method of the present invention in terms of construction accuracy is more significant.

[0126] Figure 5 The root mean square error (RMSE) of the proposed method and the reference method is compared under different high-precision monitoring data volumes. The low-precision method has 65 sensors, a low-precision attenuation of 5 dB, and a low-precision standard deviation of 4 dB. Figure 5 It can be seen that: (1) The overall trend of RMSE for all methods decreases as the amount of data increases. This indicates that the more high-precision monitoring data obtained, the higher the overall accuracy of electromagnetic spectrum map construction. (2) After the amount of high-precision data is increased, the RMSE of OK-H increases significantly, but it is still higher than that of the proposed method, which shows the performance advantage of the proposed algorithm in terms of construction accuracy.

[0127] The above results demonstrate that decomposing the electromagnetic spectrum map construction into path attenuation and shadow fading components, and using a co-kriging system of a generalized regression neural network and a high-precision data variation function to estimate the two components, can significantly improve the construction accuracy compared to using a mixture of multi-precision monitoring data.

[0128] In one embodiment, such as Figure 6 As shown, an electromagnetic spectrum map construction device based on multi-precision monitoring data is provided, including: a monitoring data acquisition module, a path attenuation component estimation module, a shadow fading component estimation module, and a received power estimate determination module for the location of the point to be estimated, wherein:

[0129] The monitoring data acquisition module is used to acquire multi-precision monitoring data of the electromagnetic environment within the region of interest and the location of the point to be estimated. The multi-precision monitoring data includes high-precision and low-precision monitoring data, and each set of data includes: location information and the received power at the corresponding location.

[0130] The path attenuation component estimation module is used to unify the data format of high-precision and low-precision monitoring data; based on the unified high-precision and low-precision monitoring data and the location of the point to be estimated, trend surface fitting is performed to obtain the path attenuation component estimate of the location of the point to be estimated.

[0131] The shadow fading component estimation module is used to obtain high-precision residual data and low-precision residual data based on the high-precision monitoring data and path attenuation component estimation values ​​after the format is unified; to construct a co-kriging system based on the variation function fitted by the high-precision residual data; and to obtain the shadow fading component estimation value of the point to be estimated by using low-precision residual data and high-precision residual data through co-kriging interpolation.

[0132] The module for determining the received power estimate of the location to be estimated is used to add the path attenuation component estimate and the shadow fading component estimate to obtain the received power estimate of the location to be estimated.

[0133] In one embodiment, the path attenuation component estimation module is further configured to use the location information and corresponding received power at the location in the high-precision monitoring data and low-precision monitoring data after format unification as training samples; construct a generalized regression neural network; train the generalized regression neural network according to the training samples to obtain a path attenuation component estimation model for the location of the point to be estimated; input the location of the point to be estimated into the path attenuation component estimation model to obtain the path attenuation component estimate value of the location of the point to be estimated.

[0134] In one embodiment, the generalized regression neural network in the path decay component estimation module includes an input layer, a pattern layer, a summation layer, and an output layer; the transfer function of the input layer is a linear function; the pattern layer is a radial basis layer, and the activation function used inside each neuron is a radial basis function; the summation layer includes two types of neurons, namely numerator units and denominator units; the numerator units are used to calculate the weighted sum of the neurons in the pattern layer, and the denominator units are used to calculate the algebraic sum; the output layer is used to calculate the path decay component estimate based on the trend surface fitting result; the trend surface fitting result is represented as a weighted sum divided by the algebraic sum.

[0135] In one embodiment, the path attenuation component estimation module is also used to perform trend surface fitting based on the high-precision monitoring data and low-precision monitoring data after format unification and the location of the point to be estimated, so as to obtain the path attenuation component estimate of the location of the point to be estimated as shown in Equation (7).

[0136] In one embodiment, the shadow fading component estimation module is further configured to fit a variation function using high-precision residual data based on the spatial correlation between the shadow fading model and the shadow fading component, as shown in Equation (8).

[0137] In one embodiment, the spatial correlation of the shadow fading component is shown in Equation (9).

[0138] In one embodiment, the co-kriging system is shown in Equation (10); the matrix form of the co-kriging system is shown in Equation (11).

[0139] In one embodiment, the shadow fading component estimation module is further configured to use low-precision residual data and high-precision residual data to obtain the shadow fading component estimate of the location of the point to be estimated by co-kriging interpolation as shown in Equation (12); the co-kriging variance of the shadow fading component estimate is shown in Equation (15).

[0140] In one embodiment, the low-precision monitoring data has an additional measurement error compared to the high-precision monitoring data, and the measurement error follows a Gaussian distribution as shown in equation (16).

[0141] Specific limitations regarding the electromagnetic spectrum map construction device based on multi-precision monitoring data can be found in the limitations of the electromagnetic spectrum map construction method based on multi-precision monitoring data described above, and will not be repeated here. Each module in the aforementioned electromagnetic spectrum map construction device based on multi-precision monitoring data can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0142] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0143] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for constructing an electromagnetic spectrum map based on multi-precision monitoring data, characterized in that, The method includes: Acquire multi-precision monitoring data of the electromagnetic environment within the region of interest and the location of the point to be estimated; the multi-precision monitoring data includes high-precision and low-precision monitoring data, and each set of data includes: location information and the received power at the corresponding location; Unify the data formats of high-precision and low-precision monitoring data; Based on the high-precision monitoring data and low-precision monitoring data after standardization and the location of the point to be estimated, a trend surface is fitted to obtain the path attenuation component estimate of the location of the point to be estimated. Based on the standardized high-precision monitoring data and low-precision monitoring data, and the path attenuation component estimation value, high-precision residual data and low-precision residual data are obtained. A co-kriging system is constructed based on the variation function fitted using the high-precision residual data; and only the high-precision variation function is used when constructing the co-kriging system, wherein the co-kriging system satisfies: , The constraints, where, , These are the Kriging weighting coefficients; N To increase the amount of high-precision monitoring data, M The amount of low-precision monitoring data is used; the low-precision residual data and the high-precision residual data are used to obtain the estimated value of the shadow fading component of the point to be estimated through co-kriging interpolation; The path attenuation component estimate and the shadow fading component estimate are added together to obtain the received power estimate for the location of the point to be estimated.

2. The method according to claim 1, characterized in that, Based on the standardized high-precision and low-precision monitoring data and the location of the point to be estimated, a trend surface is fitted to obtain the path attenuation component estimate of the point location, including: The location information and the corresponding received power in the high-precision monitoring data and low-precision monitoring data after the format is unified are used as training samples. Construct a generalized regression neural network; train the generalized regression neural network based on the training samples to obtain a path decay component estimation model for the location of the point to be estimated; The location of the point to be estimated is input into the path attenuation component estimation model to obtain the path attenuation component estimate of the location of the point to be estimated.

3. The method according to claim 2, characterized in that, The generalized regression neural network includes an input layer, a pattern layer, a summation layer, and an output layer; The transfer function of the input layer is a linear function; The pattern layer is a radial base layer, and the activation function used inside each neuron is a radial basis function. The summation layer includes two types of neurons: numerator units and denominator units; the numerator units are used to calculate the weighted sum of the neurons in the pattern layer, and the denominator units are used to calculate the algebraic sum. The output layer is used to calculate the estimated value of the path attenuation component based on the trend surface fitting result; The trend surface fitting result is expressed as the weighted sum divided by the algebraic sum.

4. The method according to claim 1, characterized in that, Based on the standardized high-precision and low-precision monitoring data and the location of the point to be estimated, trend surface fitting is performed to obtain the estimated path attenuation component value of the point location: in, The radiated power of the signal source, Location of the signal source The path decay exponent, for and The distance between them For the far-field reference distance of the antenna, The path attenuation factor is a factor relating to the operating wavelength of the radiated signal. The function.

5. The method according to claim 1, characterized in that, Based on the spatial correlation of the shadow fading model and shadow fading components, the high-precision residual data is used to fit the variogram function, including: the variogram function is: in, The value of the covariance function. Value of a nugget. For the height of the arch, The first i The position and the first j Position vectors at each location For variable range.

6. The method according to claim 1, characterized in that, The spatial correlation of the shadow fading component is: in, The standard deviation is... To solve for the correlation distance.

7. The method according to claim 1, characterized in that, The collaborative kriging system is as follows: in, For Lagrange multipliers, , These are the Kriging weighting coefficients. These are high-precision and low-precision monitoring data, respectively. This represents the variation function value between two data sampling points of different precisions. This represents the variation function value between high-precision data sampling points and unknown points. This represents the variation function values ​​between high-precision data sampling points. The values ​​of the variation function within the low-precision data sampling points. N To increase the amount of high-precision monitoring data, M The amount of low-precision monitoring data; The matrix form of the co-kriging system is as follows: in, , , , , .

8. The method according to claim 7, characterized in that, Using the low-precision residual data and the high-precision residual data, through co-kriging interpolation, the shadow fading component estimate of the location of the point to be estimated is obtained as follows: in, , , This is a matrix form of the variogram values ​​for high-precision data where multi-precision data are used as independent variables. It is expressed as the variation function value between the unknown point and the multi-precision sensor; The co-kriging variance of the shadow fading component estimation can be expressed as: in, Co-kriging variance estimated for shadow fading components.

9. The method according to any one of claims 1-8, characterized in that, The low-precision monitoring data has an additional measurement error compared to the high-precision monitoring data. This measurement error follows a Gaussian distribution. in, The attenuation is due to antenna mismatch. White noise is caused by the thermal effect of inexpensive electronic components.

10. An electromagnetic spectrum map construction device based on multi-precision monitoring data, characterized in that, The device includes: The monitoring data acquisition module is used to acquire multi-precision monitoring data of the electromagnetic environment within the region of interest and the location of the point to be estimated; the multi-precision monitoring data includes high-precision and low-precision monitoring data, and each set of data includes: location information and the received power at the corresponding location; The path attenuation component estimation module is used to unify the data format of high-precision and low-precision monitoring data; based on the high-precision and low-precision monitoring data after format unification and the location of the point to be estimated, a trend surface is fitted to obtain the path attenuation component estimate of the location of the point to be estimated. The shadow fading component estimation module is used to obtain high-precision residual data and low-precision residual data based on the high-precision monitoring data and low-precision monitoring data after format unification and the path attenuation component estimation value; and to construct a co-kriging system based on the variogram fitted using the high-precision residual data, using only the high-precision variogram function when constructing the co-kriging system, wherein the co-kriging system satisfies: , The constraints, where, , These are the Kriging weighting coefficients; N To increase the amount of high-precision monitoring data, M The amount of low-precision monitoring data; using the low-precision residual data and the high-precision residual data through co-kriging interpolation, the estimated value of the shadow fading component of the point to be estimated is obtained; The module for determining the received power estimate of the location to be estimated is used to add the path attenuation component estimate and the shadow fading component estimate to obtain the received power estimate of the location to be estimated.

Images