GNSS receiver interference sensitivity threshold calibration method and device based on loss function
By constructing a joint loss function for GNSS receiver interference sensitivity threshold calibration, the problem of inaccurate performance evaluation of GNSS receivers in electromagnetic environments is solved, and high-precision interference sensitivity threshold calibration and performance evaluation are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGSU QIYUN FLYING TECHNOLOGY CO LTD
- Filing Date
- 2026-01-30
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies struggle to quantify the interference response mechanism of GNSS receivers in complex electromagnetic environments and lack flexible interference sensitivity threshold calibration schemes, leading to inaccurate GNSS receiver performance evaluations.
A loss function-based approach is adopted, which constructs a joint loss function that combines classification loss, monotonicity loss, and threshold alignment loss to iteratively calibrate the interference sensitivity threshold of the GNSS receiver. The receiver performance is evaluated using power-sensitive feature index and spectral similarity.
It achieves high-precision interference sensitivity threshold calibration, improving the accuracy of GNSS receiver performance evaluation and anti-interference capability under electromagnetic interference environment.
Smart Images
Figure CN122017895A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of GNSS signal reception, and in particular to a method and apparatus for calibrating the interference sensitivity threshold of a GNSS receiver based on a loss function. Background Technology
[0002] Global Navigation Satellite Systems (GNSS), represented by the Global Positioning System (GPS), have become a core component of modern infrastructure, playing an irreplaceable role in critical areas such as aerospace navigation, autonomous driving, precision agriculture, and emergency response. However, the sensitivity of satellite navigation receivers to electromagnetic interference (EMI) poses a significant challenge to their performance. In particular, with the proliferation of wireless devices and increasing human-caused interference, GPS L1 band signals are threatened by various types of interference, including continuous wave (CW), pulse modulation, and BPSK modulation. These interferences can significantly affect receiver performance; therefore, evaluating the performance of GNSS receivers is crucial.
[0003] However, there are still some key issues with the response mechanism of the receiver's RF front-end in complex electromagnetic environments. First, the traditional signal-to-interference-plus-noise ratio (SINR) evaluation framework is difficult to quantify the dynamic characteristics of the intermediate frequency output related to navigation faults. The spectral threshold distribution patterns of different modulation types lack a universal theoretical framework, which limits the general design of anti-interference receivers. Furthermore, the interference sensitivity threshold is generally a fixed value, lacking a flexible calibration scheme, which is not conducive to the high-precision performance evaluation of GNSS receivers. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a GNSS receiver interference sensitivity threshold calibration method and apparatus based on a loss function. This method can iteratively calibrate the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtain the final interference sensitivity threshold when the joint loss function converges, thereby achieving interference sensitivity threshold calibration and obtaining a high-precision interference sensitivity threshold.
[0005] The objective of this invention is achieved through the following technical solution: a GNSS receiver interference sensitivity threshold calibration method based on a loss function, comprising the following steps:
[0006] Step S1. Given the set of interference signal types, frequencies, and power values to be tested;
[0007] Step S2. Generate the interference signal to be tested. When the interference generator is turned off and the interference signal to be tested is transmitted, the signal is received and the spectrum is measured by the GNSS receiver to obtain the reference spectrum and the measured spectrum.
[0008] Step S3. Normalize the interference spectrum and the measurement spectrum, and calculate the single-frequency cosine similarity between the measurement spectrum and the interference-free reference spectrum based on the normalization result; introduce the concept of weighted similarity, combine frequency weights, and give the power sensitive feature index that integrates multi-frequency information;
[0009] Step S4. Based on the power-sensitive feature index, construct a continuous exponential model to evaluate the performance of the GNSS receiver;
[0010] Step S5. Construct a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss;
[0011] Step S6. Iteratively calibrate the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtain the final interference sensitivity threshold when the joint loss function converges.
[0012] A loss function-based GNSS receiver interference sensitivity threshold calibration device includes:
[0013] The spectrum measurement module is used to generate the interference signal to be tested, given a set of interference signal types, a set of frequencies, and a set of power values. When the interference generator is turned off and the interference signal to be tested is transmitted, the GNSS receiver receives the signal and performs spectrum measurement to obtain the reference spectrum and the measured spectrum.
[0014] The power-sensitive feature index calculation module normalizes the interference spectrum and the measurement spectrum, and calculates the single-frequency cosine similarity between the measurement spectrum and the interference-free reference spectrum based on the normalization result; it introduces the concept of weighted similarity and combines it with frequency weights to give a power-sensitive feature index that integrates multi-frequency information.
[0015] The performance evaluation module uses a continuous exponential model based on power-sensitive feature indexing to evaluate the performance of GNSS receivers.
[0016] The loss construction module builds a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss;
[0017] The interference sensitivity threshold calibration module iteratively calibrates the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtains the final interference sensitivity threshold when the joint loss function converges.
[0018] The beneficial effects of the present invention are: (1) The present invention uses the L2 norm to normalize the spectrum vector, eliminating the influence of amplitude difference on subsequent calculations; (2) In order to reflect the device's sensitivity to power, the present invention first introduces an exponential penalty term in the single-frequency cosine similarity to obtain the single-frequency weighted similarity. Then, in order to reduce the influence of frequency fluctuations, the average value of the single-frequency weighted similarity is calculated to obtain the final weighted similarity of the method power level. By using frequency weighting to integrate multi-frequency information, the power sensitive feature index is finally obtained. Then, a "feature index-power" model is constructed, and after fitting, it can complete the performance evaluation of GNSS receivers with arbitrary measured power; (3) The present invention can iteratively calibrate the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtain the final interference sensitivity threshold when the joint loss function converges, thereby realizing the calibration of the interference sensitivity threshold and obtaining a high-precision interference sensitivity threshold for GNSS receiver performance judgment. Attached Figure Description
[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.
[0021] like Figure 1 As shown, the GNSS receiver interference sensitivity threshold calibration method based on loss function includes the following steps:
[0022] Step S1. Given the set of interference signal types, frequencies, and power values to be tested;
[0023] Given a set of interference signal types to be tested, M= The frequency set is: The power value set is ;
[0024] Where, N m N f N P These represent the total number of types, frequencies, and power values of the interference signals to be tested, respectively.
[0025] Step S2. Generate the interference signal to be tested. When the interference generator is turned off and the interference signal to be tested is transmitted, the signal is received and the spectrum is measured by the GNSS receiver to obtain the reference spectrum and the measured spectrum.
[0026] Given the interference signal type m∈M, interference signal power p∈P, and interference signal frequency f∈F, an interference signal to be tested is generated and transmitted via an interference generator. The GNSS receiver then receives the GNSS signal under the interference signal and performs spectrum measurements to obtain the measured spectrum. ;
[0027] When the interference generator is off, the GNSS receiver receives signals at multiple consecutive moments and performs spectrum measurements to obtain the spectrum vector under the "interference-free baseline," which is denoted as the reference spectrum. .
[0028] Step S3. Normalize the interference spectrum and the measurement spectrum, and calculate the single-frequency cosine similarity between the measurement spectrum and the interference-free reference spectrum based on the normalization result; introduce the concept of weighted similarity, combine frequency weights, and give the power sensitive feature index that integrates multi-frequency information;
[0029] S301. Measuring the spectrum using L2 norm pairs Reference spectrum After normalization, we get:
[0030] ;
[0031] in, Indicates the measurement spectrum The result obtained after normalization Indicates the reference spectrum The result obtained after normalization; Represents the L2 norm;
[0032] S302. Calculate the single-frequency cosine similarity between the measured spectrum and the interference-free reference spectrum based on the normalization processing results. :
[0033] ;
[0034] in, The inner product is expressed as follows: at frequency f, the single-frequency cosine similarity between the normalized measured spectrum and the reference spectrum is equivalent to their inner product.
[0035] S303. Introducing the concept of weighted similarity and combining it with frequency weights, a power-sensitive feature index that integrates multi-frequency information is given:
[0036] To reflect the device's sensitivity to power, an exponential penalty term is first introduced into the single-frequency cosine similarity to obtain the single-frequency weighted similarity. :
[0037] ;
[0038] make Represents the power penalty term, where The attenuation factor represents the factor that controls the attenuation rate.
[0039] When the power approaches the reference power p0, the penalty term approaches 1, and the weighted similarity approaches the single-frequency cosine similarity. As the power increases, the penalty term grows exponentially, and the weighted similarity decays exponentially. This indicates that for electromagnetically sensitive devices, once the power exceeds the threshold, the performance will drop sharply.
[0040] The power-sensitive feature index is obtained by integrating multi-frequency information using frequency weighting. :
[0041] ;
[0042] in, These are weighting coefficients, with values ranging from [0,1], representing frequency weights. Initialize to , To measure the Euclidean distance difference between the measured spectrum and the reference spectrum under the conditions of interference signal type m, interference power p, and interference frequency f:
[0043] .
[0044] Step S4. Based on the power-sensitive feature index, construct a continuous exponential model to evaluate the performance of the GNSS receiver;
[0045] S401. First, when traversing the power set P for power p, repeat steps S1~S3 to obtain the corresponding values for multiple power values. ;
[0046] S402. Constructing a continuous exponential model By utilizing different powers p and corresponding Scaling factor fitted by least squares method and attenuation coefficient ;
[0047] S403. For the measured interference signal power p, input it into the fitted continuous exponential model to obtain... Then it is compared with the interference sensitivity threshold. If it exceeds the threshold, the receiver is sensitive.
[0048] Step S5. Construct a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss;
[0049] Step S5 includes:
[0050] Construct a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss.
[0051] ;
[0052] in, , , These represent the classification loss, monotonicity loss, and threshold alignment loss, respectively. The weighting coefficients for the monotonicity loss;
[0053] Among them, classification loss for:
[0054] ;
[0055] The score is the GNSS performance evaluation result in step S4. When score=1, it means that the evaluation result is that the GNSS receiver is normal. When score=0, it means that the evaluation result is that the GNSS receiver is sensitive / faulty. The label is the tag, which is obtained by actual measurement of the GNSS receiver. When score=1, it means that the actual measurement result is that the GNSS receiver is normal. When score=0, it means that the actual measurement result is that the GNSS receiver is sensitive / faulty.
[0056] When traversing the sets of interference signal types, frequencies, and power values respectively, the interference signal type m, interference frequency f, and interference power p are combined according to the formula in the classification loss based on the GNSS performance evaluation results and the measured results, thereby completing the classification loss calculation.
[0057] Monotonicity loss for:
[0058] ;
[0059] In the formula, This represents the expected decrease in similarity for every 1 dBm increase in power.
[0060] Threshold alignment loss for:
[0061]
[0062] in, Similarity at key power points , The similarity between the left and right sides of the key point power point; The sensitivity threshold is defined as the critical power point, where the critical power point refers to the point at which the GNSS receiver becomes sensitive or fails due to interference.
[0063] Step S6. Iteratively calibrate the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtain the final interference sensitivity threshold when the joint loss function converges.
[0064] In step S6, before each iteration, steps S1 to S5 need to be repeated to obtain the joint loss function, classification loss, and threshold alignment loss. The update process for the t-th iteration is as follows:
[0065] In step S6, before each iteration, steps S1 to S5 need to be repeated to obtain the joint loss function, classification loss, and threshold alignment loss. The update process for the t-th iteration is as follows:
[0066] S601. Let the joint loss obtained in the t-th iteration be... In the (t-1)th iteration, the joint loss is obtained as First, determine whether the joint loss converges in the t-th iteration. The convergence condition is:
[0067] ;
[0068] Indicates the convergence threshold;
[0069] If the joint loss converges, the current interference sensitivity threshold is used as the optimized interference sensitivity threshold. If the joint loss does not meet the convergence condition, proceed to step S602.
[0070] S602. Based on classification loss and threshold alignment loss, determine the updated value of the interference sensitivity threshold. :
[0071]
[0072] in, This represents the updated value of the interference sensitivity threshold at the (t-1)th iteration. When t=1, Set the initial value to 0; The damping coefficient; The learning rate is used to control the step size of single-step updates and ensure convergence stability;
[0073] S603. Based on the current interference sensitivity threshold, add... Then proceed to the next iteration until the joint loss converges.
[0074] A loss function-based GNSS receiver interference sensitivity threshold calibration device includes:
[0075] The spectrum measurement module is used to generate the interference signal to be tested, given a set of interference signal types, a set of frequencies, and a set of power values. When the interference generator is turned off and the interference signal to be tested is transmitted, the GNSS receiver receives the signal and performs spectrum measurement to obtain the reference spectrum and the measured spectrum.
[0076] The power-sensitive feature index calculation module normalizes the interference spectrum and the measurement spectrum, and calculates the single-frequency cosine similarity between the measurement spectrum and the interference-free reference spectrum based on the normalization result; it introduces the concept of weighted similarity and combines it with frequency weights to give a power-sensitive feature index that integrates multi-frequency information.
[0077] The performance evaluation module uses a continuous exponential model based on power-sensitive feature indexing to evaluate the performance of GNSS receivers.
[0078] The loss construction module builds a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss;
[0079] The interference sensitivity threshold calibration module iteratively calibrates the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtains the final interference sensitivity threshold when the joint loss function converges.
[0080] The foregoing description illustrates and describes a preferred embodiment of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.
Claims
1. A GNSS receiver interference sensitivity threshold calibration method based on loss function, characterized in that: Includes the following steps: Step S1. Given the set of interference signal types, frequencies, and power values to be tested; Step S2. Generate the interference signal to be tested. When the interference generator is turned off and the interference signal to be tested is transmitted, the signal is received and the spectrum is measured by the GNSS receiver to obtain the reference spectrum and the measured spectrum. Step S3. Normalize the interference spectrum and the measurement spectrum, and calculate the single-frequency cosine similarity between the measurement spectrum and the interference-free reference spectrum based on the normalization result; introduce the concept of weighted similarity, combine frequency weights, and give the power sensitive feature index that integrates multi-frequency information; Step S4. Based on the power-sensitive feature index, construct a continuous exponential model to evaluate the performance of the GNSS receiver; Step S5. Construct a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss; Step S6. Iteratively calibrate the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtain the final interference sensitivity threshold when the joint loss function converges.
2. The GNSS receiver performance evaluation method based on joint loss and frequency weight according to claim 1, characterized in that: Step S1 includes: Given a set of interference signal types to be tested, M= The frequency set is: The power value set is In the set of frequency values, each frequency increases sequentially from front to back with the same frequency increment; in the set of power values, each power increases sequentially from front to back with the same power increment. Where, N m N f N P These represent the total number of types, frequencies, and power values of the interference signals to be tested, respectively.
3. The GNSS receiver performance evaluation method based on joint loss and frequency weight according to claim 1, characterized in that: Step S2 includes: Given the interference signal type m∈M, interference signal power p∈P, and interference signal frequency f∈F, an interference signal to be tested is generated and transmitted via an interference generator. The GNSS receiver then receives the GNSS signal under the interference signal and performs spectrum measurements to obtain the measured spectrum. ; When the interference generator is off, the GNSS receiver receives signals at multiple consecutive time points and performs spectrum measurements to obtain the spectrum vector under the "interference-free baseline," which is denoted as the reference spectrum. .
4. The GNSS receiver performance evaluation method based on joint loss and frequency weight according to claim 1, characterized in that: Step S3 includes: S301. Measuring the spectrum using L2 norm pairs Reference spectrum After normalization, we get: ； in, Indicates the measurement spectrum The result obtained after normalization Indicates the reference spectrum The result obtained after normalization; Represents the L2 norm; S302. Calculate the single-frequency cosine similarity between the measured spectrum and the interference-free reference spectrum based on the normalization processing results. : ； in, The inner product is expressed as follows: at frequency f, the single-frequency cosine similarity between the normalized measured spectrum and the reference spectrum is equivalent to their inner product. S303. Introducing the concept of weighted similarity and combining it with frequency weights, a power-sensitive feature index that integrates multi-frequency information is given: To reflect the device's sensitivity to power, an exponential penalty term is first introduced into the single-frequency cosine similarity to obtain the single-frequency weighted similarity. : ； make Represents the power penalty term, where The attenuation factor represents the factor that controls the attenuation rate. When the power approaches the reference power p0, the penalty term approaches 1, and the weighted similarity approaches the single-frequency cosine similarity. As the power increases, the penalty term grows exponentially, and the weighted similarity decays exponentially. This indicates that for electromagnetically sensitive devices, once the power exceeds the threshold, the performance will drop sharply. The power-sensitive feature index is obtained by integrating multi-frequency information using frequency weighting. : ； in, These are weighting coefficients, with values ranging from [0,1], representing frequency weights. for , To measure the Euclidean distance difference between the measured spectrum and the reference spectrum under the conditions of interference signal type m, interference power p, and interference frequency f: 。 5. The GNSS receiver performance evaluation method based on joint loss and frequency weight according to claim 1, characterized in that: Step S4 includes: S401. First, when traversing the power set P for power p, repeat steps S1~S3 to obtain the corresponding values for multiple power values. ; S402. Constructing a continuous exponential model By utilizing different powers p and corresponding Scaling factor fitted by least squares method and attenuation coefficient ; S403. For the measured interference signal power p, input it into the fitted continuous exponential model to obtain... Then it is compared with the interference sensitivity threshold. If it exceeds the threshold, the receiver is sensitive.
6. The GNSS receiver performance evaluation method based on joint loss and frequency weight according to claim 1, characterized in that: Step S5 includes: Construct a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss. ； in, , , These represent the classification loss, monotonicity loss, and threshold alignment loss, respectively. The weighting coefficients for the monotonicity loss; Among them, classification loss for: ； The score is the GNSS performance evaluation result in step S4. When score=1, it means that the evaluation result is that the GNSS receiver is normal. When score=0, it means that the evaluation result is that the GNSS receiver is sensitive / faulty. The label is the tag, which is obtained by actual measurement of the GNSS receiver. When score=1, it means that the actual measurement result is that the GNSS receiver is normal. When score=0, it means that the actual measurement result is that the GNSS receiver is sensitive / faulty. When traversing the sets of interference signal types, frequencies, and power values respectively, the interference signal type m, interference frequency f, and interference power p are combined according to the formula in the classification loss based on the GNSS performance evaluation results and the measured results, thereby completing the classification loss calculation. Monotonicity loss for: ； In the formula, This represents the expected decrease in similarity for every 1 dBm increase in power. Threshold alignment loss for: in, Similarity at key power points , The similarity between the left and right sides of the key point power point; The sensitivity threshold is defined as the critical power point, where the critical power point refers to the point at which the GNSS receiver becomes sensitive or fails due to interference.
7. The GNSS receiver performance evaluation method based on joint loss and frequency weight according to claim 6, characterized in that: In step S6, before each iteration, steps S1 to S5 need to be repeated to obtain the joint loss function, classification loss, and threshold alignment loss. The update process for the t-th iteration is as follows: S601. Let the joint loss obtained in the t-th iteration be... In the (t-1)th iteration, the joint loss is obtained as First, determine whether the joint loss converges in the t-th iteration. The convergence condition is: ； Indicates the convergence threshold; If the joint loss converges, the current interference sensitivity threshold is used as the optimized interference sensitivity threshold. If the joint loss does not meet the convergence condition, proceed to step S602. S602. Based on classification loss and threshold alignment loss, determine the updated value of the interference sensitivity threshold. : in, This represents the updated value of the interference sensitivity threshold at the (t-1)th iteration. When t=1, Set the initial value to 0; The damping coefficient; The learning rate is used to control the step size of single-step updates and ensure convergence stability; S603. Based on the current interference sensitivity threshold, add... Then proceed to the next iteration until the joint loss converges.
8. A GNSS receiver interference sensitivity threshold calibration device based on a loss function, comprising the method described in any one of claims 1 to 7, characterized in that: include: The spectrum measurement module is used to provide a set of interference signal types, frequencies, and power values to be tested. An interference signal to be tested is generated. When the interference generator is turned off and the interference signal to be tested is transmitted, the signal is received and the spectrum is measured by a GNSS receiver to obtain the reference spectrum and the measured spectrum. The power-sensitive feature index calculation module normalizes the interference spectrum and the measurement spectrum, and calculates the single-frequency cosine similarity between the measurement spectrum and the interference-free reference spectrum based on the normalization result; it introduces the concept of weighted similarity and combines it with frequency weights to give a power-sensitive feature index that integrates multi-frequency information. The performance evaluation module uses a continuous exponential model based on power-sensitive feature indexing to evaluate the performance of GNSS receivers. The loss construction module builds a joint loss function based on classification loss, monotonicity loss, and threshold alignment loss; The interference sensitivity threshold calibration module iteratively calibrates the interference sensitivity threshold based on classification loss and threshold alignment loss, and obtains the final interference sensitivity threshold when the joint loss function converges.