Small sample remote lightning space calibration method and system based on diversified rule optimization

Abstract

The invention provides a small sample remote thunder space calibration method and system based on diversified rule optimization, and is suitable for space calibration of telemetering and radar in trajectory measurement. The method comprises the following steps: firstly, establishing a remote lightning space calibration model; then, aiming at a small sample problem, Gaussian white noise is added to generate diversified data, and a differential innovation search algorithm of an adaptive learning strategy is adopted to promote solving of a space calibration optimization problem; secondly, averaging different results to obtain a stable parameter estimation result; the method provided by the invention is suitable for telemetering and radar space calibration in an environment with few measured data, ill-conditioned data and large noise interference.

Description

Technical Field

[0001] The invention belongs to the field of guided projectile trajectory measurement in a high dynamic environment, and relates to a small sample remote mine space calibration method and system based on diversified rule optimization. Background Art

[0002] In modern information warfare, precision-guided artillery shells have become a key weapon for achieving precision strikes in modern warfare due to their excellent combat performance. The key to achieving high-precision guidance of guided artillery shells lies in obtaining accurate ballistic measurement data. Since the missile-borne telemetry equipment (IMU and GNSS combined navigation) and the ground radar spatial coordinate system are not unified, it is necessary to perform spatial calibration between the telemetry data and radar measurement data in different spatial coordinate systems before fusion of measurement data.

[0003] Spatial calibration is a necessary step before multi-sensor data fusion, and its core lies in solving the external parameters between different three-dimensional rectangular coordinate systems. So far, there have been many studies on the solution of spatial calibration problems, such as analytical methods such as least squares and weighted least squares, numerical iterative methods such as Gauss-Newton and Levenberg-Marquardt, and intelligent optimization methods such as particle swarm optimization, ant colony optimization, and simulated annealing. Conventional analytical algorithms are prone to divergence in parameter estimation results when faced with measurement data containing large noise, while numerical iterative methods require good initial values ​​for iteration. If the given initial values ​​are inappropriate, it is very easy for the algorithm to fall into a local optimal solution, and even cause distortion of the results. Intelligent optimization algorithms have good robustness and global search capabilities under complex conditions, and have been widely used in multiple optimization fields. The differentiated creative search algorithm is a novel meta-heuristic optimization algorithm proposed by Duankhan et al., which has a good ability to solve nonlinear optimization problems.

[0004] Unlike general carriers, guided artillery shells are in a harsh environment with high overload and high dynamics, and are easily disturbed by external uncertain factors such as airflow, resulting in large measurement errors. Within the measurement range of the radar, the flight trajectory of the shell is approximately a straight line motion, with little effective data and high linearity, and there is a serious morbidity. The problem of ill-conditioned matrix will lead to the instability of the solution of the equation. Even a very small disturbance will have a huge impact on the estimation result. In response to the ill-conditioned matrix problem, Eitan et al. proposed a method for determining the general Tikhonov regularization parameter of linear ill-posed problems to separate unknown noise from the signal, thereby obtaining reliable parameter estimation results. Li et al. used TSVD and TR methods to improve the degree of morbidity and obtain a stable numerical solution. DK et al. proposed an ill-conditioned total least squares load identification method for the morbid problems and unknown noise interference problems in ship load identification. However, further research is still needed in the face of problems with large noise and morbid data. Summary of the invention

[0005] The purpose of the present invention is to provide a small sample remote mine space calibration method and system based on diversified rule optimization, so as to improve the remote mine space calibration accuracy under the environment of little measurement data, pathological data and large noise interference.

[0006] The technical solution to achieve the purpose of the present invention is:

[0007] A small sample remote mine space calibration method based on diversified rule optimization, comprising:

[0008] Constructing spatial calibration model and objective function;

[0009] Collect data and add Gaussian white noise to generate diversified sample data;

[0010] The diversified sample data are used as the input of the remote mine space calibration model, and the differential search algorithm with adaptive learning strategy is used to solve the constructed objective function, and the parameter estimation values ​​of the space calibration model under different sample data are obtained.

[0011] The parameter estimates obtained from different sample data are averaged and used as the final parameter estimates of the spatial calibration model.

[0012] Furthermore, the remote mine space calibration model is:

[0013]

[0014] Where L and I represent the radar coordinate system and the telemetry coordinate system respectively. is a 3×3 rotation matrix, which represents the rotation relationship between the radar coordinate system and the telemetry coordinate system. z The three rotation angles of the axis are determined; is a 3×1 translation vector, which represents the translation relationship between the radar coordinate system and the telemetry coordinate system; They are the measurement data in radar coordinate system and telemetry coordinate system respectively; The position of the shell under the radar's perspective at time zero when the telemetry equipment is powered on and starts working.

[0015] Furthermore, assuming that the radar measurement data X L With translation parameters is fixed, then the spatial calibration model is equivalent to:

[0016]

[0017] In the formula,

[0018] Furthermore, the objective function is:

[0019]

[0020] Where: Y i I is the i-th telemetry data, is the i-th radar measurement data, i=1,2,3...n is the total number of measurement data.

[0021] Furthermore, the learning rate of the differential search algorithm of the adaptive learning strategy is:

[0022]

[0023] In the formula, fit i is the fitness of the i-th member, fit best is the fitness of the best member in the current iteration, fit worst is the fitness of the worst member in the current iteration.

[0024] A small sample remote mine space calibration system based on diversified rule optimization, comprising:

[0025] A model building unit, used to build a spatial calibration model and an objective function;

[0026] A diversification unit adds Gaussian white noise to the collected data to generate diversified sample data;

[0027] The optimization solution unit uses the diversified sample data as the input of the remote mine space calibration model, and uses the differentiated search algorithm of the adaptive learning strategy to solve the constructed objective function, and obtains the parameter estimation value of the space calibration model under different sample data;

[0028] The regularization unit averages the parameter estimates obtained from different sample data as the final parameter estimates of the spatial calibration model.

[0029] A small sample remote mine space calibration device based on diversified rule optimization includes: a memory, a processor and a computer program stored in the memory. When the processor executes the computer program, the steps of the small sample remote mine space calibration method are implemented.

[0030] A computer storage medium stores an executable program, and the executable program is executed by a processor to implement the steps of the small sample remote mine space calibration method.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1) The present invention diversifies the original small data samples by adding Gaussian white noise, and averages the parameter estimation results to obtain the optimal estimation results;

[0033] 2) The calibration framework of the present invention can improve the stability and accuracy of parameter estimation and has robustness and portability;

[0034] 3) Compared with the traditional algorithm, the present invention designs the learning rate and adopts the differentiated search algorithm of the adaptive learning strategy to solve the constructed objective function, which can improve the accuracy of parameter estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of a small sample remote mine space calibration method based on diverse rule optimization.

[0036] Figure 2 Schematic diagram of a small sample remote mine space calibration framework based on optimization of diverse rules.

[0037] Figure 3 Schematic diagram of the diversification process. DETAILED DESCRIPTION

[0038] Next, the implementation of the present invention will be described in detail with reference to the accompanying drawings.

[0039] Combination Figure 1 , a small sample remote mine spatial calibration method based on diversified rule optimization, suitable for the spatial calibration of telemetry and radar in ballistic measurement, including:

[0040] Firstly, a remote mine space calibration model is established, the objective function is constructed, the influence of noise on the spatial extrinsic parameter solution of pathological data is analyzed, and the objective function is constructed;

[0041] This patent uses the telemetry coordinate system as a reference to transform the measurement data in the radar coordinate system into the telemetry coordinate system. The translation parameters between the telemetry coordinate system and the radar coordinate system can be approximately determined by the position of the shell at the zero-time radar perspective. Therefore, the spatial calibration model can be established as:

[0042]

[0043]

[0044] Where: L and I represent the radar coordinate system and the telemetry coordinate system respectively. is a 3×3 rotation matrix, which represents the rotation relationship between the radar coordinate system and the telemetry coordinate system. z The three rotation angles of the axis are determined; is a 3×1 translation vector, which represents the translation relationship between the radar coordinate system and the telemetry coordinate system; They are the measurement data in radar coordinate system and telemetry coordinate system respectively; The position of the shell under the radar's perspective at time zero when the telemetry equipment is powered on and starts working.

[0045] matrix Represents rotation and translation respectively, with 6 degrees of freedom, and with three or more common points, the spatial extrinsic parameters between two different coordinate systems can be solved. Therefore, the objective function can be constructed as:

[0046]

[0047] Where: Y i I is the i-th telemetry data, is the i-th radar measurement data, i=1,2,3...n is the total number of measurement data.

[0048] By optimizing the objective function, the optimal spatial extrinsic parameters can be obtained. and

[0049] In the early stage of launch, the flight trajectory of the projectile is close to linear motion, the data is highly linear, and there is a serious pathological problem. The projectile is in a highly dynamic and harsh environment, and the measurement data contains a lot of noise. Assume that the radar measurement data X L With translation parameters If is fixed, the spatial calibration model can be equivalent to:

[0050]

[0051] In the formula,

[0052] If there is a small disturbance δY′ (i.e. error) in the measurement data of the telemetry equipment, the rotation parameter The corresponding error have:

[0053]

[0054] So we have:

[0055]

[0056] According to the definition of norm:

[0057]

[0058] The relative error of the solution can be obtained as:

[0059]

[0060] For the spatial calibration model When Y′ produces a disturbance δY′, the rotation parameter The corresponding error Will be magnified Therefore, the problem of ill-conditioned matrix will lead to unstable solution of spatial external parameters. Even very small error disturbance will have a huge impact on the estimation results, and it is impossible to obtain accurate and reliable parameter estimation results.

[0061] Then, in order to solve the problem of small data samples, Gaussian white noise is added to the collected telemetry and radar small sample data to generate diversified data, so as to reduce the influence of noise on the subsequent estimation of spatial extrinsic parameters;

[0062] In order to solve the spatial calibration problems of large noise, highly pathological data and small samples, this patent adds uniformly distributed white noise to the original small sample data multiple times based on the diversity principle and the characteristic that the mean of white noise is zero to generate more data. At the same time, the artificially added noise can cover up the noise of the data itself, such as the attached Figure 3 shown.

[0063] Secondly, the diversified sample data are used as the input of the remote mine space calibration model, and the differential innovation search algorithm with adaptive learning strategy is used to solve the objective function, and the parameter estimation values ​​of the space calibration model under different sample data are obtained to improve the accuracy of parameter estimation.

[0064] Diversity can bring rich variability and adaptability, and will not change the original commonalities and laws (i.e., real spatial external parameters). Rich data will give the "system" greater flexibility and stability. Therefore, it is easier for us to find its unchanging commonalities, i.e., spatial external parameters. and

[0065] This patent transforms the remote mine space calibration problem into a nonlinear optimization problem of formula (3). The DCS algorithm uses differential knowledge acquisition and creative realism strategies to solve complex optimization problems, but the fixed learning rate cannot fully inspire team members to explore unknown areas, affecting the accuracy of the solution.

[0066] Since the performance of each member in the population, i.e., the fitness, is constantly changing during the iterative update of the population, and the gap between each member and the optimal member is also constantly changing, in order to more reasonably stimulate the learning enthusiasm of each member and promote the development of the population, this patent proposes to use formula (11) to update the learning rate of each member:

[0067]

[0068] In the formula, fit i is the fitness of the i-th member, fit best is the fitness of the best member in the current iteration, fit worst is the fitness of the worst member in the current iteration.

[0069] Finally, the parameter estimates obtained from different sample data are averaged as the final parameter estimates of the spatial calibration model, thereby obtaining stable parameter estimation results.

[0070] There is a spatial transformation relationship between the data generated by adding uniformly distributed Gaussian white noise, which contains real external parameters. Figure 3 As shown in the figure, it can be seen that the generated large sample data still maintains the original spatial transformation relationship. This shows that when the same phenomenon shows diversity under different conditions, it often still follows a certain specific law. By revealing the commonality behind the phenomenon, it can help us discover the law from the complex information.

[0071] In order to obtain stable parameter estimation results, we perform parameter estimation on data with different noise levels respectively, and then take the mean of the first m parameter estimation results as the representative of the m parameter estimations, among which the set of mean results with the smallest objective function loss is taken as the optimal result.

[0072] This embodiment provides a framework for optimizing spatial calibration algorithms based on various rules. Figure 2 As shown, including:

[0073] A model building unit, used to build a spatial calibration model and an objective function;

[0074] A diversification unit adds Gaussian white noise to the collected data to generate diversified sample data;

[0075] The optimization solution unit uses the diversified sample data as the input of the remote mine space calibration model, and uses the differentiated search algorithm of the adaptive learning strategy to solve the constructed objective function, and obtains the parameter estimation value of the space calibration model under different sample data;

[0076] The regularization unit averages the parameter estimates obtained from different sample data as the final parameter estimates of the spatial calibration model.

[0077] This embodiment also provides a small sample remote mine space calibration device based on diverse rule optimization, including: a memory, a processor and a computer program stored in the memory, and the processor implements the steps of the small sample remote mine space calibration method when executing the computer program.

[0078] This embodiment also provides a computer storage medium, which stores an executable program. The executable program is executed by a processor to implement the steps of the small sample remote mine spatial calibration method.

[0079] The method of the invention is suitable for the space calibration of telemetry and radar in an environment with little measurement data, morbid data and large noise interference.

[0080] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0081] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A small sample remote mine space calibration method based on diversified rule optimization, characterized in that: include: Constructing spatial calibration models and objective functions; Collect data and add Gaussian white noise to generate diversified sample data; The diversified sample data are used as the input of the remote mine space calibration model, and the differential search algorithm with adaptive learning strategy is used to solve the constructed objective function, and the parameter estimation values ​​of the space calibration model under different sample data are obtained. The parameter estimates obtained from different sample data are averaged and used as the final parameter estimates of the spatial calibration model.

2. According to the method of claim 1, a small sample remote mine space calibration method based on diversified rule optimization is characterized in that: The remote mine space calibration model is: Where L and I represent the radar coordinate system and the telemetry coordinate system respectively. is a 3×3 rotation matrix, which represents the rotation relationship between the radar coordinate system and the telemetry coordinate system. z The three rotation angles of the axis are determined; is a 3×1 translation vector, which represents the translation relationship between the radar coordinate system and the telemetry coordinate system; They are the measurement data in radar coordinate system and telemetry coordinate system respectively; The position of the shell under the radar's perspective at time zero when the telemetry equipment is powered on and starts working.

3. According to the method of claim 2, a small sample remote mine space calibration method based on diversified rule optimization is characterized in that: Assume that the radar measurement data X L With translation parameters is fixed, then the spatial calibration model is equivalent to: In the formula, 4. According to the method of claim 2, a small sample remote mine space calibration method based on diversified rule optimization is characterized in that: The objective function is: Where: Y i I is the i-th telemetry data, is the i-th radar measurement data, i=1,2,3...n is the total number of measurement data.

5. The method for spatial calibration of remote mines with small samples based on optimization of diverse rules according to claim 1 is characterized in that: The learning rate of the differential search algorithm of the adaptive learning strategy is: In the formula, fit i is the fitness of the i-th member, fit best is the fitness of the best member in the current iteration, fit worst is the fitness of the worst member in the current iteration.

6. A small sample remote mine space calibration system based on diversified rule optimization for implementing any of the methods described in claims 1-5, characterized in that: include: A model building unit, used to build a spatial calibration model and an objective function; A diversification unit adds Gaussian white noise to the collected data to generate diversified sample data; The optimization solution unit uses the diversified sample data as the input of the remote mine space calibration model, and uses the differentiated search algorithm of the adaptive learning strategy to solve the constructed objective function, and obtains the parameter estimation value of the space calibration model under different sample data; The regularization unit averages the parameter estimates obtained from different sample data as the final parameter estimates of the spatial calibration model.

7. A small sample remote mine space calibration device based on diversified rule optimization, characterized in that: include: A memory, a processor and a computer program stored in the memory, wherein when the processor executes the computer program, the steps of the small sample remote mine space calibration method according to any one of claims 1 to 5 are implemented.

8. A computer storage medium, characterized in that: The computer storage medium stores an executable program, and the executable program is executed by a processor to implement the steps of the small sample remote mine spatial calibration method according to any one of claims 1 to 5.

Images