A method and device for identifying a speed control loop model of an unmanned aerial vehicle

Abstract

The present application belongs to the field of drone control technology, and discloses a method and device for identifying a drone speed control loop model, the method comprising: preprocessing the original data of the drone speed control to obtain identification data; based on the identification data, identifying a first identification model through the MATLAB identification toolbox, and identifying a second identification model through the CIFER identification toolbox; complementary fusion of the first identification model and the second identification model to obtain a third identification model; integrating the first identification model, the second identification model and the third identification model to obtain a drone speed control loop model. The present application can achieve the effect of obtaining a more accurate drone speed control loop model.

Description

Technical Field

[0001] The present application relates to the field of unmanned aerial vehicle control technology, and in particular to a method and device for identifying a speed control loop model of an unmanned aerial vehicle. Background Art

[0002] With the rapid development of mechatronics, computer and communication technology, the application scenarios of rotary-wing UAVs are becoming more and more extensive, such as performing aerial photography and cruising tasks within the region, complex formation performances, etc. Waypoint tracking and fixed-point hovering are typical outer-loop position control problems. High-precision and high-robust position control algorithms often require the use of model-based design (MBD). In recent years, many commercial small rotary-wing UAVs have been transformed or redeveloped by colleges and universities or research institutes to serve the flight verification of new concepts and new algorithms in the field of single UAVs or multi-UAV clusters. However, the speed control loop of most rotary-wing UAVs is a black box mode, that is, the secondary developer does not know the specific speed control algorithm, so it is necessary to obtain a simplified approximate speed control loop model through identification methods. With the help of the identified approximate model, the designer can understand and analyze the time scale of the speed control loop and the characteristics of the speed control algorithm, so as to carry out the design and parameter optimization of the UAV position controller based on the model.

[0003] However, due to the large number of complex uncertainties, it is difficult to construct a relatively accurate underlying control loop model based on the transfer function models of the drone's gyroscopic effect, aerodynamics, control efficiency, power unit, control distribution, and the parameters in the model; and when the speed command of the drone is in different frequency bands, the two identification models identified by using the two current mainstream system identification toolboxes, Matlab and CIFER, have different fitting effects, and when the speed command is at the junction of low frequency and high frequency, the correlation coefficient, determination coefficient, and mean square error between the two identification models and the output signal of the original actual system of the drone are all less than the fitting standard, that is, the two identification models cannot accurately simulate the actual speed control loop model of the drone. It can be seen that the existing technology has the problem of poor accuracy of the drone speed control loop model obtained by identification. Summary of the invention

[0004] The present application provides a method and device for identifying a UAV speed control loop model, which can obtain a more accurate UAV speed control loop model.

[0005] In a first aspect, an embodiment of the present application provides a method for identifying a speed control loop model of a drone, the method comprising:

[0006] Preprocess the raw data of UAV speed control to obtain identification data;

[0007] Based on the identification data, a first identification model is obtained by using the MATLAB identification toolbox, and a second identification model is obtained by using the CIFER identification toolbox;

[0008] Complementarily fusing the first identification model and the second identification model to obtain a third identification model;

[0009] The first identification model, the second identification model and the third identification model are integrated to obtain the speed control loop model of the UAV.

[0010] In one embodiment, the identification data includes input data and output data, the input data includes speed command data, and the output data includes measured position data;

[0011] The first identification model is obtained by using the MATLAB identification toolbox, including:

[0012] Import the speed command data and measured position data into the MATLAB identification toolbox;

[0013] The imported measured position data is filtered, and the speed command data and the filtered measured position data are identified by using the MATLAB identification toolbox to obtain multiple MATLAB identification models;

[0014] A MATLAB identification model is selected from a plurality of MATLAB identification models as a first identification model, and a fitting degree of the first identification model is greater than a first threshold.

[0015] In one embodiment, the output data also includes measured speed data;

[0016] The second identification model is obtained by using the CIFER identification toolbox, including:

[0017] Import the speed command data and measured speed data into the CIFER identification toolbox;

[0018] The speed command data and measured speed data are identified through the CIFER identification toolbox to obtain multiple CIFER identification models;

[0019] A CIFER identification model is selected from the multiple CIFER identification models as a second identification model, and a cost function of the second identification model is less than a second threshold.

[0020] In one embodiment, the first recognition model and the second recognition model are complementary fused to obtain a third recognition model, including:

[0021] Based on the multiple weight coefficient values, a fusion calculation is performed on the first identification model and the second identification model to obtain a fusion model corresponding to each weight coefficient value;

[0022] Selecting a fusion model that meets a preset fitting condition from each fusion model to obtain a plurality of optional fusion models;

[0023] The preset fitting condition is that all fitting index values ​​of a fusion model in the mid-frequency band are within a preset threshold range.

[0024] A fusion model with the best fitting effect is selected from multiple optional fusion models as the third identification model.

[0025] In one embodiment, performing fusion calculation on the first identification model and the second identification model based on multiple weight coefficient values ​​includes:

[0026] Based on multiple weight coefficient values, a weighted average method and / or a geometric average method are used to perform a fusion calculation on the first identification model and the second identification model.

[0027] In one embodiment, the first identification model, the second identification model and the third identification model are integrated to obtain a speed control loop model of the drone, including:

[0028] Obtain the low frequency range, medium frequency range and high frequency range of the UAV speed command;

[0029] The first identification model is set as a low-frequency band model corresponding to the low-frequency range, the third identification model is set as a mid-frequency band model corresponding to the mid-frequency range, and the second identification model is set as a high-frequency band model corresponding to the high-frequency range;

[0030] According to the low-frequency band model, the medium-frequency band model and the high-frequency band model, the speed control loop model of the UAV is obtained; the speed control loop model is a full-frequency band model that uses the low-frequency band model in the low-frequency band, the medium-frequency band model in the medium-frequency band, and the high-frequency band model in the high-frequency band.

[0031] In one embodiment, the raw data of the drone speed control is preprocessed, including:

[0032] The raw data of UAV speed control is preprocessed by identifying, eliminating, repairing and filtering outliers.

[0033] In a second aspect, an embodiment of the present application provides a device for identifying a speed control loop model of a drone, the device comprising:

[0034] The data preprocessing module is used to preprocess the raw data of the UAV speed control to obtain identification data;

[0035] A model identification module, used for obtaining a first identification model through a MATLAB identification toolbox based on identification data, and obtaining a second identification model through a CIFER identification toolbox;

[0036] A model fusion module, used for complementary fusion of the first identification model and the second identification model to obtain a third identification model;

[0037] The model integration module is used to integrate the first identification model, the second identification model and the third identification model to obtain the speed control loop model of the UAV.

[0038] In a third aspect, an embodiment of the present application provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the processor executes the steps of the method for identifying the drone speed control loop model of any of the above embodiments.

[0039] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for identifying a speed control loop model of a drone as in any of the above embodiments.

[0040] In summary, compared with the prior art, the technical solution provided by this application has at least the following beneficial effects:

[0041] The present application provides an identification method for a speed control loop model of a UAV, which can preprocess the original data of the speed control of the UAV to obtain identification data; based on the identification data, a first identification model is obtained by identifying through a MATLAB identification toolbox, and a second identification model is obtained by identifying through a CIFER identification toolbox, wherein the first identification model is applicable to low-frequency speed instructions, and the second identification model is applicable to high-frequency speed instructions; the first identification model and the second identification model are complementary and integrated to obtain a third identification model, and the third identification model is applicable to medium-frequency speed instructions; the first identification model, the second identification model and the third identification model are integrated to obtain a speed control loop model of the UAV, and the above-mentioned speed control loop model may include three identification models applicable to full-band speed instructions of the UAV. The above method can be used to identify the first identification model with good fitting effect when the speed instruction is low frequency and the second identification model with good fitting effect when the speed instruction is high frequency respectively through the MATLAB identification toolbox and the CIFER identification toolbox. By complementary fusion of the first identification model and the second identification model, a third identification model with good fitting effect when the speed instruction is medium frequency is obtained; the full-band identification model obtained by integrating the above three identification models is used as the final speed control loop model of the UAV, which can obtain a model that is closer to the actual speed control state of the UAV, that is, a more accurate UAV speed control loop model. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flowchart of a method for identifying a UAV speed control loop model provided as an exemplary embodiment of the present application.

[0043] Figure 2 An example of a full-band model of a small rotorcraft is provided as an exemplary embodiment of the present application.

[0044] Figure 3 A flowchart of the model fusion steps provided for an exemplary embodiment of the present application.

[0045] Figure 4 A flowchart of the model integration steps provided for an exemplary embodiment of the present application.

[0046] Figure 5 A structural diagram of a device for identifying a drone speed control loop model provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] See also Figure 1 The embodiment of the present application provides a method for identifying a speed control loop model of a drone, which is described by taking a terminal as an example. The method specifically includes the following steps:

[0049] Step S1, preprocessing the original data of the drone speed control to obtain identification data.

[0050] The raw data of the UAV speed control may include the input data and output data of the UAV speed control. The input data is the speed command, i.e., the sweep frequency signal; the output data includes the measured speed and measured position data of the UAV. In the actual implementation process, a sweep frequency signal with a time history of 0 to 100 s and a frequency of 0.09 to 1.21 Hz uniformly distributed can be selected as the speed command.

[0051] Specifically, the terminal can preprocess the raw data using common data preprocessing methods, for example, preprocessing the raw data by identifying, removing and repairing outliers, or preprocessing the raw data by filtering. Data preprocessing can improve the accuracy of data that needs to be identified later, and by improving the quality of identified data, the accuracy of subsequent identification can be improved.

[0052] Step S2: Based on the identification data, a first identification model is obtained by identifying with a MATLAB identification toolbox, and a second identification model is obtained by identifying with a CIFER identification toolbox.

[0053] Among them, the MATLAB identification toolbox is a common tool for providing system model identification. It can identify the first identification model of the drone based on the speed command and actual position received by the drone. The CIFER identification toolbox is a comprehensive software package that uses swept frequency input for frequency domain system identification. It includes all modules from data processing to model identification, and can quickly and accurately identify mathematical models. The CIFER identification toolbox can identify the second identification model of the drone based on the speed command and actual speed received by the drone.

[0054] Since speed instructions of different frequencies correspond to different acceleration requirements, that is, different voltage values ​​or PWM waves (used to control the motor speed), the motor will produce different speeds at different speed instruction frequencies; and since the motor model itself is nonlinear, the linearization models at different speed points will be different. It can be seen that speed instructions of different frequencies will correspond to different speed control loop models. In the specific implementation process, when the speed instruction is low frequency, the fitting effect of the first identification model obtained by the MATLAB identification toolbox is better; when the speed instruction is high frequency, the identification result of the second identification model obtained by the CIFER toolbox is better. When the speed instruction is at the junction of low frequency and high frequency, the correlation coefficient, determination coefficient and mean square error between the above two identification models and the output signal of the original actual system are all less than the fitting standard, that is, the fitting effect of the above two identification models is relatively poor. Therefore, the present embodiment divides the identification model into three frequency bands: low, medium and high. In step S2, the first identification model applicable to the low frequency band and the second identification model applicable to the high frequency band are first determined, and then the identification model applicable to the medium frequency band is determined by the subsequent step S3.

[0055] Specifically, for the MATLAB identification toolbox, the speed command in the identification data can be used as input data, and the data of the measured position can be used as output data, and the first identification model can be obtained by the MATLAB identification toolbox; for the CIFER identification toolbox, the speed command in the identification data can be used as input data, and the data of the measured speed can be used as output data, and the second identification model can be obtained by the CIFER identification toolbox. In the actual implementation process, the first identification model and the second identification model can both be first-order inertia plus pure delay (a first-order system with a time delay, FOTD) identification models, referred to as FOTD identification models.

[0056] Step S3, complementarily fusing the first identification model and the second identification model to obtain a third identification model.

[0057] The third identification model is obtained by complementary fusion of the first identification model and the second identification model and has a good fitting effect when the speed instruction is a medium frequency, that is, an identification model suitable for the medium frequency band.

[0058] Specifically, the weighted average method or the geometric mean method is used to complementarily fuse the first identification model and the second identification model, and the fusion models corresponding to the weight coefficient values ​​are obtained by setting different weight coefficient values. From the above fusion models, a fusion model with the best fitting effect when the speed instruction is a medium frequency is selected as the third identification model.

[0059] Step S4, integrating the first identification model, the second identification model and the third identification model to obtain a speed control loop model of the UAV.

[0060] Among them, when the speed command is in the low frequency band, the first identification model obtained by the MATLAB identification toolbox is used, when the speed command is in the high frequency band, the second identification model obtained by the CIFER identification toolbox is used, and when the speed command is in the medium frequency band, the fusion model is used. The full-band splicing model with the best fitting effect can be obtained. The full-band splicing model can be expressed by a piecewise transfer function, see Figure 2 An example of a full-band model of a small rotorcraft is shown.

[0061] Specifically, a full-band splicing model is obtained according to the first identification model, the second identification model and the third identification model, and the full-band splicing model is used as the speed control loop model of the UAV.

[0062] In this embodiment, the terminal may be a desktop terminal or a mobile terminal, and the mobile terminal may be at least one of a mobile phone, a tablet computer, a laptop computer, etc. Since the frequency band division theory is based on the nonlinearity of the motor model, the above method is applicable to electric propulsion aircraft using a motor, an electronic speed regulator and a propeller structure, such as a small rotor drone.

[0063] The above embodiment provides an identification method for a UAV speed control loop model, which can pre-process the original data of the UAV speed control to obtain identification data; based on the identification data, a first identification model is obtained by identifying through the MATLAB identification toolbox, and a second identification model is obtained by identifying through the CIFER identification toolbox, wherein the first identification model is suitable for low-frequency speed instructions, and the second identification model is suitable for high-frequency speed instructions; the first identification model and the second identification model are complementary and integrated to obtain a third identification model, and the third identification model is suitable for medium-frequency speed instructions; the first identification model, the second identification model and the third identification model are integrated to obtain a speed control loop model of the UAV, and the above speed control loop model may include three identification models suitable for full-band speed instructions of the UAV. The above method can be used to identify the first identification model with good fitting effect when the speed instruction is low frequency and the second identification model with good fitting effect when the speed instruction is high frequency respectively through the MATLAB identification toolbox and the CIFER identification toolbox. By complementary fusion of the first identification model and the second identification model, a third identification model with good fitting effect when the speed instruction is medium frequency is obtained; the full-band identification model obtained by integrating the above three identification models is used as the final speed control loop model of the UAV, which can obtain a model that is closer to the actual speed control state of the UAV, that is, a more accurate UAV speed control loop model.

[0064] In some embodiments, the identification data includes input data and output data, the input data includes speed instruction data, and the output data includes measured position data.

[0065] In step S2, the step of obtaining the first identification model by using the MATLAB identification toolbox specifically includes the following steps:

[0066] Import the speed command data and measured position data into the MATLAB identification toolbox.

[0067] Among them, the speed command data and the measured position data are the input signal and output signal of the UAV speed control, and the above data are imported into the MATLAB identification toolbox.

[0068] In order to improve the quality of identified data and ensure the accuracy of MATLAB identification, the imported measured position data is filtered, and the speed command data and the filtered measured position data are identified through the MATLAB identification toolbox to obtain multiple MATLAB identification models.

[0069] Specifically, the imported data can be de-biased and filtered, and the de-biased data can be divided into two parts, one part as the estimated model data and the other part as the verification model data. The estimated model data is identified using the MATLAB identification toolbox to obtain multiple MATLAB identification models, and the degree of fit of each MATLAB identification model is calculated using the verification model data. In the actual implementation process, since the low-order model is easy to use, it can be preferentially identified as a continuous transfer function model (Transfer Function Model) with a denominator of 1 order, a numerator of 0 order and a delay link.

[0070] A MATLAB identification model is selected from a plurality of MATLAB identification models as a first identification model, and a fitting degree of the first identification model is greater than a first threshold.

[0071] The first threshold is set according to actual needs and may be 70% to 90%, for example, 80%.

[0072] Specifically, the fitting degree of each MATLAB identification model among the multiple MATLAB identification models is calculated, and a MATLAB identification model with a fitting degree greater than 80% is selected as the first identification model.

[0073] In the above embodiment, the method can identify the speed command data and the measured position data through the MATLAB identification toolbox to obtain multiple MATLAB identification models, and select the first identification model according to the degree of fitting, and pre-process the data before identification to improve the quality of the identified data and ensure the accuracy of MATLAB identification.

[0074] In some embodiments, the output data also includes measured speed data; in step S2, the step of obtaining the second identification model by using the CIFER identification toolbox specifically includes the following steps:

[0075] Import the speed command data and measured speed data into the CIFER identification toolbox.

[0076] The speed command data and measured speed data are identified through the CIFER identification toolbox to obtain multiple CIFER identification models.

[0077] Specifically, when the input speed command data and the measured speed data are time domain data, the time domain data are converted into frequency domain data, that is, the frequency domain response data of the input / output data; and the identification range is selected according to the correlation between the input data and the output data, and the number of iterations is input. The continuous transfer function model with a denominator of 1st order, a numerator of 0th order and a delay link is preferentially identified, and multiple CIFER identification models can be obtained.

[0078] A CIFER identification model is selected from the multiple CIFER identification models as a second identification model, and a cost function of the second identification model is less than a second threshold.

[0079] The second threshold is set according to actual needs, for example, 100. Specifically, the cost function of each CIFER identification model in the multiple CIFER identification models is calculated, and a CIFER identification model with a cost function less than 100 is selected as the second identification model.

[0080] In the above embodiment, the method can obtain multiple CIFER identification models by identifying the speed command data and the measured position data through the CIFER identification toolbox, and select the second identification model according to the cost function; wherein, since it is mentioned that compared with MATLAB using time domain identification, CIFER using frequency domain identification can better reduce the errors caused by the deviation effect and measurement noise, especially in the high frequency band, the CIFER toolbox has a more obvious noise elimination effect on the measured speed, therefore, the second identification model obtained by the CIFER identification toolbox has a better fitting degree in the high frequency band.

[0081] In some embodiments, see Figure 3 , step S3 specifically includes the following steps:

[0082] Step S31 , performing fusion calculation on the first identification model and the second identification model based on a plurality of weight coefficient values ​​to obtain fusion models corresponding to the respective weight coefficient values.

[0083] Wherein, the first identification model and the second identification model may be fused and calculated by using a weighted average method and / or a geometric average method based on a plurality of weight coefficient values.

[0084] Specifically, the weighted average method and the geometric mean method are respectively used to perform fusion calculation on the first identification model and the second identification model, and the fitting index of the fusion model in the mid-frequency band under different weight coefficient values ​​is calculated.

[0085] For example: The standard FOTD model can be expressed as The model of toolbox identification is often expressed as:

[0086]

[0087] Among them, K is the gain coefficient, T is the time constant of the inertia link, and τ is the pure delay time.

[0088] Fusion model of weighted average method:

[0089]

[0090] Fusion model of geometric mean method:

[0091]

[0092] Among them, a is the weight coefficient, which ranges from 0 to 1. The closer a is to 1, the greater the proportion of the first identification model of the MATLAB identification toolbox in the fusion model; the closer a is to 0, the greater the proportion of the second identification model of the CIFER identification toolbox in the fusion model.

[0093] Step S32, selecting a fusion model that meets a preset fitting condition from each fusion model to obtain a plurality of optional fusion models.

[0094] Step S33: Select a fusion model with the best fitting effect from multiple optional fusion models as the third identification model.

[0095] The preset fitting condition is that the values ​​of each fitting index of a fusion model in the mid-frequency band are within the preset threshold range; each fitting index may include the correlation coefficient ρ(X, Y), the determination coefficient R 2 and mean square error MSE.

[0096] The correlation coefficient ρ(X,Y) is the amount of linear correlation between the research variables X and Y, and its expression is:

[0097]

[0098] in, are the means of variables X and Y, respectively, and n is the sample size. When two variables are positively correlated, ρ(X, Y) is between 0 and 1. The closer it is to 1, the stronger the positive correlation between the two variables. When ρ(X, Y) < 0.95, the identification model is considered to not meet the fitting criteria.

[0099] Coefficient of determination R 2 It is a measure of the goodness of fit between the research variables X and Y, and its expression is:

[0100]

[0101] in, is the mean of variable X. In general, R 2 It is between 0 and 1. The closer it is to 1, the better the goodness of fit between the two variables. 2 When <0.9, the identification model is considered not to meet the fitting criteria.

[0102] The mean square error (MSE) is a measure of the degree of deviation between the study variables X and Y, and its expression is:

[0103]

[0104] Where n is the sample size. As can be seen from the expression, MSE is greater than 0. The closer it is to 0, the smaller the deviation between the two variables. When MSE>0.01, it is considered that the identification model does not meet the fitting standard.

[0105] For example, we traverse a from 0.1 to 0.9 at intervals of 0.1, compare the fitting indexes of the mid-frequency model after the weighted average method and the geometric mean method are combined, and retain the weight coefficient value a that satisfies the following three conditions at the same time: ρ(X,Y)>0.95, R 2 >0.9, MSE <0.01, and finally select ρ(X,Y) and R 2 The case where the value is closer to 1 and the MSE is closer to 0 is taken as the optimal model, that is, the third identification model obtained by the complementary fusion method. At this time, the preset fitting conditions are ρ(X,Y)>0.95, R 2 >0.9, MSE <0.01, the above preset fitting conditions are met and ρ(X,Y) and R 2 The fusion model whose MSE is closer to 1 and closer to 0 is the fusion model with the best fitting effect.

[0106] In the above embodiment, the method can adopt a complementary fusion method to obtain a third identification model that is accurately applicable to the mid-frequency band, wherein the fusion model with the best fitting effect can be selected as the third identification model through the weighted average method, the geometric mean method combined with the optimization of the weight coefficient value, thereby improving the accuracy of the speed control loop model of the UAV.

[0107] In some embodiments, see Figure 4 , step S4 specifically includes the following steps:

[0108] Step S41, obtaining the low frequency range, medium frequency range and high frequency range of the speed instruction of the drone.

[0109] Among them, each frequency range can be set by the technicians according to actual needs. For example, when a sweep signal with a time history of 0 to 100 s and a uniformly distributed frequency of 0.09 to 1.21 Hz is selected as the speed instruction, 0.09 to 0.2636 Hz can be divided into a low frequency band, 0.2636 to 0.3196 Hz can be divided into a medium frequency band, and 0.3196 to 1.21 Hz can be divided into a high frequency band.

[0110] Step S42, setting the first identification model to a low frequency band model corresponding to the low frequency range, setting the third identification model to a mid-frequency band model corresponding to the mid-frequency range, and setting the second identification model to a high frequency band model corresponding to the high frequency range.

[0111] Specifically, the first identification model is determined to be a low-frequency band model corresponding to the low-frequency range, the third identification model is a mid-frequency band model corresponding to the mid-frequency range, and the second identification model is a high-frequency band model corresponding to the high-frequency range. Step S42 determines the identification models corresponding to different frequency ranges.

[0112] Step S43, obtaining a speed control loop model of the UAV according to the low-frequency band model, the mid-frequency band model and the high-frequency band model.

[0113] The speed control loop model is a full-band model that uses a low-frequency model in the low-frequency band, a medium-frequency model in the medium-frequency band, and a high-frequency model in the high-frequency band.

[0114] In the above embodiment, the method can divide the simulated approximate speed control loop model into three frequency bands: low, medium and high according to the models identified by the two toolboxes. The obtained multi-band model can more accurately simulate the actual speed control loop model of the drone.

[0115] In some embodiments, step S1 specifically includes the following steps: pre-processing the raw data of the drone speed control by identifying, eliminating and repairing outliers, and filtering to obtain identification data.

[0116] In actual flight tests, external interference or accidental instrument jumps may cause unreasonable jumps in the collected data. These jumps are called outliers. In order to avoid the influence of outliers on the identification results, outliers must be identified, eliminated and repaired first. Specifically, a low-order polynomial sliding fitting method with a simple algorithm and less calculation can be used to identify and eliminate outliers. Here, the forward difference form is used to prevent the current value from being misjudged as an outlier due to the influence of subsequent outliers. Because using more data points is conducive to more accurate identification of outliers, the seven-point second-order formula is selected to identify outliers:

[0117]

[0118] Among them, y i (i=1,…,7) represents a node, Yes i The improved value.

[0119] Calculate point by point in time series and If satisfied It is considered as an outlier. Outlier identification is the premise of outlier removal. In order to ensure the consistency of data length, it is necessary to perform numerical interpolation on outlier points. This paper uses Lagrange interpolation. For a polynomial function, given k+1 value points (x0, y0),…, (x k ,y k ), xj ,y j Represent the position of the independent variable and the corresponding function value respectively, then the Lagrange interpolation polynomial is:

[0120]

[0121] Among them, l i (x) is called the Lagrange basic polynomial, and its expression is:

[0122]

[0123] Substituting the independent variable corresponding to the position to be interpolated into L(x) can obtain the required function value. During the flight, the data collection may have continuous jump points. If point k is an outlier point and point k+i satisfies Then the k+i point is also an outlier point. In order to prevent the step signal from being regarded as an outlier, m=3 can be set. When |y i -yi2>0.81k=i-3iyi-yi2 and i>3, yk, yk+1,…, yk+m are considered to be normal values.

[0124] If we judge y k ,y k+1 ,…,y k+m is an outlier, using y k-3 ,y k-2 ,y k-1 ,y k+m+1 ,y k+m+2 ,y k+m+3 To perform Lagrange interpolation to obtain the correction value of the outlier point, the calculation formula is as follows (where: l = k, k+1, ..., k+m and i≠j≠k, k+1, ..., k+m):

[0125]

[0126] There is often noise in actual flight test data, which is mainly the measurement noise of the sensor. This paper uses moving average filtering to reduce the impact of noise. Assuming the input is x and the output is y, the calculation formula of the moving average filter is:

[0127]

[0128] Wherein, y(n) refers to the correction value of the nth item, N is the length of the queue, and N=10.

[0129] In the above embodiment, the method can pre-process the original data by identifying, removing, repairing and filtering outliers in the original data, thereby avoiding the influence of outliers and noise on the recognition results and improving the accuracy of the recognition results.

[0130] See also Figure 5 Another embodiment of the present application provides a device for identifying a speed control loop model of a drone, the device comprising:

[0131] The data preprocessing module 101 is used to preprocess the original data of the drone speed control to obtain identification data.

[0132] The model identification module 102 is used to obtain a first identification model through the MATLAB identification toolbox based on the identification data, and to obtain a second identification model through the CIFER identification toolbox.

[0133] The model fusion module 103 is used to complementarily fuse the first identification model and the second identification model to obtain a third identification model.

[0134] The model integration module 104 is used to integrate the first identification model, the second identification model and the third identification model to obtain a speed control loop model of the UAV.

[0135] In some embodiments, the identification data includes input data and output data, the input data includes speed command data, and the output data includes measured position data; the model identification module 102 is specifically used to import the speed command data and the measured position data into the MATLAB identification toolbox; filter the imported measured position data, and identify the speed command data and the filtered measured position data through the MATLAB identification toolbox to obtain multiple MATLAB identification models; select a MATLAB identification model from the multiple MATLAB identification models as a first identification model, and the fitting degree of the first identification model is greater than a first threshold.

[0136] In some embodiments, the output data also includes measured speed data; the model identification module 102 is specifically used to import the speed command data and the measured speed data into the CIFER identification toolbox; the speed command data and the measured speed data are identified by the CIFER identification toolbox to obtain multiple CIFER identification models; a CIFER identification model is selected from the multiple CIFER identification models as the second identification model, and the cost function of the second identification model is less than the second threshold.

[0137] In some embodiments, the model fusion module 103 is specifically used to perform a fusion calculation on the first identification model and the second identification model based on multiple weight coefficient values ​​to obtain fusion models corresponding to each weight coefficient value; select a fusion model that meets the preset fitting conditions from each fusion model to obtain multiple optional fusion models; select the fusion model with the best fitting effect from multiple optional fusion models as the third identification model.

[0138] In some embodiments, the model fusion module 103 is specifically used to perform a fusion calculation on the first identification model and the second identification model based on multiple weight coefficient values ​​using a weighted average method and / or a geometric average method to obtain fusion models corresponding to each weight coefficient value; select a fusion model that meets a preset fitting condition from each fusion model to obtain multiple optional fusion models; select a fusion model with the best fitting effect from multiple optional fusion models as the third identification model.

[0139] In some embodiments, the model integration module 104 is specifically used to obtain the low frequency range, medium frequency range and high frequency range of the speed command of the UAV; the first identification model is set to the low frequency band model corresponding to the low frequency range, the third identification model is set to the medium frequency band model corresponding to the medium frequency range, and the second identification model is set to the high frequency band model corresponding to the high frequency range; according to the low frequency band model, the medium frequency band model and the high frequency band model, the speed control loop model of the UAV is obtained; the speed control loop model is a full-band model that uses a low frequency band model in the low frequency band, a medium frequency band model in the medium frequency band, and a high frequency band model in the high frequency band.

[0140] In one embodiment, the data preprocessing module 101 is specifically used to preprocess the raw data of the drone speed control by identifying, eliminating and repairing outliers, and filtering to obtain identification data.

[0141] For the specific definition of the identification device of the UAV speed control loop model provided in this embodiment, please refer to the embodiment of the identification method of the UAV speed control loop model mentioned above, which will not be repeated here. Each module in the above-mentioned identification device of the UAV speed control loop model can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0142] An embodiment of the present application provides a computer device, which may include a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the processor executes the steps of the method for identifying the speed control loop model of a drone as in any of the above embodiments.

[0143] The working process, working details and technical effects of the computer device provided in this embodiment can be found in the above embodiment of the identification method of the UAV speed control loop model, which will not be repeated here.

[0144] The embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the identification method of the UAV speed control loop model in any of the above embodiments are implemented. The computer-readable storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive and / or a memory stick, etc., and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0145] The working process, working details and technical effects of the computer-readable storage medium provided in this embodiment can be found in the above embodiment of the identification method of the UAV speed control loop model, and will not be repeated here.

[0146] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0147] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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.

[0148] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for identifying a speed control loop model of an unmanned aerial vehicle, characterized in that: The method comprises: Preprocess the raw data of UAV speed control to obtain identification data; Based on the identification data, a first identification model is obtained by identifying with a MATLAB identification toolbox, and a second identification model is obtained by identifying with a CIFER identification toolbox; Complementarily fusing the first identification model and the second identification model to obtain a third identification model; The first identification model, the second identification model and the third identification model are integrated to obtain the speed control loop model of the UAV; specifically, the low frequency range, the medium frequency range and the high frequency range of the speed command of the UAV are obtained; the first identification model is set as the low frequency band model corresponding to the low frequency range, the third identification model is set as the medium frequency band model corresponding to the medium frequency range, and the second identification model is set as the high frequency band model corresponding to the high frequency range; according to the low frequency band model, the medium frequency band model and the high frequency band model, the speed control loop model of the UAV is obtained; the speed control loop model is a full frequency band model which adopts the low frequency band model in the low frequency band, the medium frequency band model in the medium frequency band and the high frequency band model in the high frequency band.

2. The method according to claim 1, characterized in that The identification data includes input data and output data, the input data includes speed instruction data, and the output data includes measured position data; The first identification model is obtained by identifying the first identification model through the MATLAB identification toolbox, including: Importing the speed command data and the measured position data into a MATLAB identification toolbox; The imported measured position data is filtered, and the speed command data and the filtered measured position data are identified by using the MATLAB identification toolbox to obtain a plurality of MATLAB identification models; A MATLAB identification model is selected from the multiple MATLAB identification models as a first identification model, and a fitting degree of the first identification model is greater than a first threshold.

3. The method according to claim 2, characterized in that The output data also includes measured speed data; The second identification model obtained by identifying the second identification model through the CIFER identification toolbox includes: Importing the speed command data and the measured speed data into a CIFER identification toolbox; Identifying the speed command data and the measured speed data by using the CIFER identification toolbox to obtain a plurality of CIFER identification models; A CIFER identification model is selected from the multiple CIFER identification models as a second identification model, and a cost function of the second identification model is less than a second threshold.

4. The method according to claim 1, characterized in that The method of performing complementary fusion on the first identification model and the second identification model to obtain a third identification model includes: Based on a plurality of weight coefficient values, a fusion calculation is performed on the first identification model and the second identification model to obtain fusion models corresponding to the weight coefficient values ​​respectively; Selecting a fusion model that meets a preset fitting condition from the fusion models to obtain a plurality of optional fusion models; A fusion model with the best fitting effect is selected from the multiple optional fusion models as the third identification model.

5. The method according to claim 4, characterized in that The performing fusion calculation on the first identification model and the second identification model based on multiple weight coefficient values ​​includes: Based on multiple weight coefficient values, a weighted average method and / or a geometric average method are used to perform a fusion calculation on the first identification model and the second identification model.

6. The method according to any one of claims 1 to 5, characterized in that: The preprocessing of the raw data of the UAV speed control includes: The raw data of UAV speed control is preprocessed by identifying, eliminating, repairing and filtering outliers.

7. An identification device for a speed control loop model of an unmanned aerial vehicle, characterized in that: The device comprises: The data preprocessing module is used to preprocess the raw data of the UAV speed control to obtain identification data; A model identification module, used to obtain a first identification model through a MATLAB identification toolbox based on the identification data, and to obtain a second identification model through a CIFER identification toolbox; A model fusion module, used for complementary fusion of the first recognition model and the second recognition model to obtain a third recognition model; A model integration module is used to integrate the first identification model, the second identification model and the third identification model to obtain a speed control loop model of the UAV; specifically, to obtain a low frequency range, a medium frequency range and a high frequency range of the speed command of the UAV; the first identification model is set as a low frequency band model corresponding to the low frequency range, the third identification model is set as a medium frequency band model corresponding to the medium frequency range, and the second identification model is set as a high frequency band model corresponding to the high frequency range; the speed control loop model of the UAV is obtained according to the low frequency band model, the medium frequency band model and the high frequency band model; the speed control loop model is a full frequency band model that uses the low frequency band model in the low frequency band, the medium frequency band model in the medium frequency band, and the high frequency band model in the high frequency band.

8. A computer device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images