Adaptive vibration suppression method and system for integrated gimbal

Abstract

The application relates to the technical field of gimbal vibration control, and discloses a self-adaptive vibration suppression method and system for an integrated gimbal, the method comprising the following steps: acquiring a gimbal vibration signal, performing segmented frequency domain analysis, obtaining a vibration frequency spectrum and a frequency band energy distribution; determining a dominant frequency band according to the frequency band energy distribution, and extracting a dominant frequency from the vibration frequency spectrum; obtaining a dominant energy sequence according to the frequency band energy distribution of a continuous analysis period, performing state estimation on the dominant energy sequence, and obtaining an energy state; determining a vibration suppression filter node according to the dominant frequency band, obtaining a gimbal suppression parameter according to the dominant frequency and the energy state; and performing instruction encapsulation according to the gimbal suppression parameter and the filter node identifier, and obtaining a vibration suppression control instruction. The method can improve the adaptability of the integrated gimbal vibration suppression parameter to vibration state changes.

Description

Technical Field

[0001] This invention relates to the field of gimbal vibration control technology, and in particular to an adaptive vibration suppression method and system for integrated gimbals. Background Technology

[0002] Currently, integrated pan-tilt units are used in equipment for mobile imaging, visual inspection, precision measurement, and target observation. By adjusting the attitude of the supporting components through a drive mechanism, the imaging or inspection components mounted on the pan-tilt unit remain relatively stable. With the increasing complexity of equipment operating states and environments in industrial control systems, the mechanical vibrations experienced by the pan-tilt unit's supporting structure typically contain different frequency components, and the intensity of each frequency component varies with the drive state, the load-bearing state, and external disturbances.

[0003] In a current technology, attitude or vibration data of a gimbal during operation is typically acquired using inertial or vibration sensors, and filters and control loops are used to reduce the impact of vibration signals on the gimbal's attitude. Some solutions employ pre-configured filtering parameters to suppress vibrations within a specific frequency range, while others adjust parameters such as the center frequency of the notch filter based on the detected resonant frequency. These methods can suppress vibrations at predetermined frequencies or relatively stable spectral states. However, when the gimbal's operating state changes continuously, the energy proportions of different vibration frequency bands also change accordingly. Vibration components that previously dominated energy may weaken, while vibration components at other frequencies may become new major disturbances. When using pre-configured filtering parameters, a deviation can easily occur between the filter's operating frequency and the current dominant vibration frequency. When using methods that adjust parameters locally around the current resonant frequency, the parameter adjustments are mainly based on local frequency information, offering limited reflection of the changes in the dominant vibration energy over time. Therefore, when the dominant vibration frequency and intensity continuously change, the vibration suppression parameters tend to lag behind the actual disturbance state.

[0004] Existing technologies suffer from the problem that the vibration suppression parameters of integrated gimbals are not matched in a timely manner with the current major disturbances. Summary of the Invention

[0005] This invention provides an adaptive vibration suppression method and system for integrated gimbals to solve the problem in the prior art where the vibration suppression parameters of integrated gimbals are not matched with the current main disturbances in a timely manner.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an adaptive vibration suppression method for an integrated gimbal, comprising: The vibration signal of the integrated gimbal is acquired by the sensor during operation, and the vibration signal is analyzed in segmented frequency domain to obtain the vibration spectrum and frequency band energy distribution corresponding to each analysis period. The dominant frequency band is determined based on the frequency band energy distribution of the current analysis period, and the dominant frequency corresponding to the dominant frequency band is extracted from the vibration spectrum of the current analysis period. The energy percentage corresponding to the dominant frequency band in each analysis period is extracted from the frequency band energy distribution of the continuous analysis period. The energy percentages are arranged in chronological order to obtain the dominant energy sequence. The state of the dominant energy sequence is estimated to obtain the energy state. Based on the dominant frequency band of the current analysis period, a corresponding vibration suppression filter node is determined in the gimbal control loop of the integrated gimbal. The target center frequency is determined based on the dominant frequency, and the target attenuation depth is determined based on the energy state. The target center frequency and the target attenuation depth are combined to obtain the gimbal suppression parameters. The vibration suppression control command is obtained by encapsulating the command based on the gimbal suppression parameters and the filter node identifier of the vibration suppression filter node.

[0007] In a second aspect, the present invention provides an adaptive vibration suppression system for an integrated gimbal, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention performs segmented frequency domain analysis on the vibration signal of the integrated gimbal during operation to obtain the vibration spectrum and frequency band energy distribution corresponding to each analysis period. Based on the frequency band energy distribution, the dominant frequency band is determined, and the dominant frequency corresponding to the dominant frequency band is extracted from the vibration spectrum. This process associates the vibration energy distribution in different frequency ranges with specific frequency positions, so that the frequency positions of the main vibration components during gimbal operation can be continuously determined with each analysis period, providing a frequency basis corresponding to the current main disturbance for subsequent adjustment of vibration suppression parameters.

[0009] (2) This invention extracts the energy proportion of the dominant frequency band for each analysis period from the frequency band energy distribution of a continuous analysis period, constructs a dominant energy sequence in chronological order, and estimates the energy state of the dominant energy sequence. This process uses the energy change information of the main vibration components within a continuous analysis period to characterize the relative energy state of the current main vibration components, reduces the influence of energy fluctuations within a single analysis period on the vibration state judgment, and enables subsequent attenuation parameters to be adjusted according to the continuous energy state of the main vibration components.

[0010] (3) This invention determines the corresponding vibration suppression filter node based on the dominant frequency band of the current analysis period, determines the target center frequency based on the dominant frequency, and determines the target attenuation depth based on the energy state. The target center frequency and the target attenuation depth are combined to form the gimbal suppression parameters and generate vibration suppression control commands. This process makes the operating frequency of the vibration suppression filter node correspond to the frequency position of the current main vibration, and adjusts the attenuation degree according to the energy state of the main vibration, thereby improving the matching degree between the integrated gimbal vibration suppression parameters and the current main disturbance. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the adaptive vibration suppression method for an integrated gimbal provided in the first embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides an adaptive vibration suppression method for an integrated gimbal, comprising the following steps: S1, acquire the gimbal vibration signal collected by the sensor during the operation of the integrated gimbal, perform segmented frequency domain analysis on the gimbal vibration signal, and obtain the vibration spectrum and frequency band energy distribution corresponding to each analysis period; S2, determine the dominant frequency band based on the frequency band energy distribution of the current analysis period, and extract the dominant frequency corresponding to the dominant frequency band from the vibration spectrum of the current analysis period; S3, extract the energy proportion corresponding to the dominant frequency band of each analysis period from the frequency band energy distribution of the continuous analysis period, arrange the energy proportions in chronological order to obtain the dominant energy sequence, and perform state estimation on the dominant energy sequence to obtain the energy state; S4. Based on the dominant frequency band of the current analysis cycle, determine the corresponding vibration suppression filter node in the gimbal control loop of the integrated gimbal, determine the target center frequency based on the dominant frequency, determine the target attenuation depth based on the energy state, and combine the target center frequency and the target attenuation depth to obtain the gimbal suppression parameters. S5. The vibration suppression control command is obtained by encapsulating the command based on the gimbal suppression parameters and the filter node identifier of the vibration suppression filter node.

[0014] In step S1, the vibration signal of the gimbal is acquired by the sensor during the operation of the integrated gimbal, and the gimbal vibration signal is subjected to segmented frequency domain analysis to obtain the vibration spectrum and frequency band energy distribution corresponding to each analysis period.

[0015] The analysis involves segmenting the frequency domain of the gimbal vibration signal to obtain the vibration spectrum and frequency band energy distribution for each analysis period, including: The gimbal vibration signal is segmented and windowed to obtain vibration data segments corresponding to each analysis period; Perform a fast Fourier transform on the vibration data segment to obtain the vibration spectrum; The vibration spectrum is divided into multiple vibration frequency bands according to the frequency range, and the spectral amplitude in each vibration frequency band is squared and accumulated to obtain the corresponding frequency band energy. Calculate the proportion of each frequency band energy in the total frequency band energy to obtain the frequency band energy distribution.

[0016] Specifically, the gimbal vibration signal is acceleration time-series data reflecting the mechanical vibration of the gimbal's supporting structure during operation. In one implementation, a piezoelectric accelerometer is fixed to the connection between the gimbal's supporting component and the drive mechanism. The sensor's sensing direction is consistent with the detection direction of the corresponding supporting structure, and the acquisition module records the sensor output data at each sampling time in chronological order.

[0017] After the sensor is installed, its corresponding sensitivity parameters are read. These sensitivity parameters characterize the change in the sensor's output electrical signal corresponding to a unit change in acceleration. When determining the sensitivity parameters using a calibration method, a reference acceleration of known amplitude is applied to the sensor, the amplitude of the corresponding sensor output electrical signal is recorded, and the ratio between the output electrical signal amplitude and the reference acceleration is calculated. This ratio is recorded as the sensitivity parameter. Subsequently, the sensor's output electrical signal is divided by the sensitivity parameter to obtain the corresponding acceleration data.

[0018] After sensitivity calibration, a vibration data segment is continuously acquired while the gimbal drive mechanism remains stationary. The arithmetic mean of all sampled values ​​in the vibration data segment is calculated and recorded as the zero-point output. The subsequent acceleration data is subtracted from the zero-point output to form the gimbal vibration signal. If the sensor installation position, acquisition channel, or sensor body changes, the above zero-point calibration is repeated.

[0019] The upper limit of the frequency domain analysis is determined based on the effective frequency range of the sensor and the adjustable center frequency range of the vibration suppression filter nodes in the gimbal control loop. Specifically, it is the upper boundary of the intersection of the frequency range in which the sensor can effectively respond and the center frequency range that the filter nodes can be configured with. This boundary is used to limit the frequency coverage range of subsequent frequency domain analysis. Specifically, the upper boundary of the sensor's effective frequency range and the total upper boundary of the adjustable center frequency ranges of all vibration suppression filter nodes are read, and the smaller of the two upper boundaries is determined as the upper limit of the frequency domain analysis.

[0020] For example, if the effective frequency range of the sensor covers more than 500 Hz, and the vibration suppression filter node in the gimbal control loop can configure the center frequency in the range of 0 Hz to 500 Hz, then 500 Hz is determined as the upper limit of frequency domain analysis under the current configuration.

[0021] The sampling frequency is selected from the sampling frequency ranges supported by the acquisition module. During selection, the Nyquist frequency corresponding to the sampling frequency is higher than the upper limit of the frequency domain analysis, and the passband of the anti-aliasing low-pass filter of the acquisition module covers the frequency domain analysis range, while the stopband start frequency is not higher than the Nyquist frequency. Among the sampling ranges that meet the above conditions, the one with the lower numerical value is selected.

[0022] For example, this embodiment selects a sampling frequency of 2000 Hz, which corresponds to a Nyquist frequency of 1000 Hz. The frequency domain analysis range from 0 Hz to 500 Hz is within the Nyquist frequency. The passband of the anti-aliasing low-pass filter corresponding to the acquisition module covers 0 Hz to 500 Hz, and enters the stopband before 1000 Hz. This attenuates frequency components above the analysis range before completing discrete sampling.

[0023] In one implementation, each vibration data segment contains 1024 consecutive sampling points. At a sampling frequency of 2000 Hz, the 1024 sampling points correspond to 0.512 seconds of continuous vibration data, and the discrete frequency interval corresponding to the 1024 points of Fast Fourier Transform is approximately 1.953125 Hz.

[0024] The length of the vibration data segment is determined jointly based on the frequency resolution and the data length of the Fast Fourier Transform. First, the initial bandwidth parameters of each vibration suppression filter node in the pan-tilt control loop are read. The effective bandwidth with the smallest value is determined from all vibration suppression filter nodes. The frequency interval not greater than half of the minimum effective bandwidth is determined as the spectral resolution condition. Then, from the data lengths corresponding to integer powers of two, the smaller data length that satisfies the spectral resolution condition when the sampling frequency is divided by the data length is selected.

[0025] For example, when the minimum effective bandwidth of each vibration suppression filter node is 4 Hz, the spectral frequency interval is no greater than 2 Hz. At a sampling frequency of 2000 Hz, 1024 points correspond to a frequency interval of approximately 1.953125 Hz, therefore 1024 sampling points are selected as the vibration data segment length.

[0026] The frame shift between adjacent vibration data segments is determined based on the scheduling period of the filter parameter update task in the gimbal control program. The sampling frequency is multiplied by the scheduling period to obtain the number of new sampling points within one scheduling period, and this number of sampling points is determined as the frame shift.

[0027] For example, the scheduling period for the filter parameter update task is 0.128 seconds. At a sampling frequency of 2000 Hz, one scheduling period corresponds to 256 new sampling points. Therefore, the frame shift is set to 256 sampling points. Each vibration data segment contains 1024 sampling points, and adjacent vibration data segments have 768 overlapping sampling points, corresponding to a 75% overlap. The 75% is obtained by combining the 1024-point data length and the 256-point frame shift, and is not set separately as a data segmentation parameter.

[0028] After segmenting the gimbal vibration signal, the arithmetic mean of all acceleration values ​​in each vibration data segment is calculated, and the acceleration value corresponding to each sampling point is subtracted from the arithmetic mean to obtain the vibration data segment after removing the DC bias.

[0029] Subsequently, a Hanning window is applied to the vibration data segment. Each sampled value in the vibration data segment is multiplied point-by-point by the Hanning window coefficient corresponding to the same sampling position to obtain the windowed vibration data segment. The Hanning window coefficient is calculated directly based on the sampling point number and the current vibration data segment length, without setting additional weighting parameters.

[0030] Perform a 1024-point Fast Fourier Transform on the vibration data segments corresponding to each analysis period to obtain frequency domain data in complex form. Read the frequency domain data of the non-negative frequency part, calculate the corresponding spectral amplitude based on the real and imaginary parts of each complex frequency component, and associate the spectral amplitude with the corresponding discrete frequency to obtain the vibration spectrum of the current analysis period.

[0031] The same sensor calibration relationship, sampling frequency, data length, Hanning window, and spectral amplitude calculation method are used for each analysis period, thereby ensuring that the vibration spectrum has the same data scale between different analysis periods.

[0032] The vibration frequency band is divided according to the adjustable center frequency range of the vibration suppression filter node in the pan-tilt control loop. The filter node identifier, lower boundary of the adjustable center frequency, and upper boundary of the adjustable center frequency of each vibration suppression filter node are read. The vibration suppression filter nodes are arranged from low to high frequency, and the continuous center frequency range corresponding to each filter node is determined as a vibration frequency band.

[0033] For example, the gimbal control loop includes three vibration suppression filter nodes, with corresponding adjustable center frequency ranges of 0 Hz to less than 50 Hz, 50 Hz to less than 200 Hz, and 200 Hz to 500 Hz, respectively, thus forming three vibration frequency bands corresponding to the three vibration suppression filter nodes.

[0034] The frequency band boundaries adopt a left-closed, right-open assignment relationship, with the last vibration band containing the upper limit of frequency domain analysis. Therefore, the frequency point corresponding to 50 Hz is assigned to the second vibration band, the frequency point corresponding to 200 Hz is assigned to the third vibration band, and the frequency point corresponding to 500 Hz is assigned to the third vibration band. A valid frequency point belongs to only one vibration band.

[0035] For each vibration frequency band, read all discrete frequency points that fall within the corresponding frequency range in the vibration spectrum of the current analysis period, square the spectral amplitude of each discrete frequency point, and then sum all the squared results. The summed value is determined as the frequency band energy of the corresponding vibration frequency band.

[0036] After obtaining the frequency band energy corresponding to all vibration frequency bands, the energy of each frequency band is added together to obtain the total frequency band energy. Then, the frequency band energy of each vibration frequency band is divided by the total frequency band energy to obtain the energy proportion of each vibration frequency band. The energy proportions are recorded according to the frequency arrangement of the vibration frequency bands to obtain the frequency band energy distribution of the current analysis period.

[0037] When the energy of all frequency bands is zero, the energy percentage calculation is not performed, and the current analysis period is recorded as a period without effective vibration data. The analysis period does not enter the subsequent process of determining the dominant frequency band. During the period without effective vibration data, the vibration suppression filter node keeps the gimbal suppression parameter corresponding to the previous effective analysis period unchanged. After the energy of all frequency bands in the next analysis period is greater than zero, the frequency band energy distribution calculation continues to be performed.

[0038] For example, if the frequency band energies corresponding to the three vibration frequency bands in a certain analysis cycle are 1200, 4800, and 2000, then the total frequency band energy is 8000. Dividing 1200, 4800, and 2000 by 8000 respectively, we obtain the energy percentages corresponding to the three vibration frequency bands as 15%, 60%, and 25%. Recording these three energy percentages according to the order of the vibration frequency bands yields the frequency band energy distribution for the current analysis cycle.

[0039] In step S2, the dominant frequency band is determined based on the frequency band energy distribution of the current analysis period, and the dominant frequency corresponding to the dominant frequency band is extracted from the vibration spectrum of the current analysis period.

[0040] The process of determining the dominant frequency band based on the frequency band energy distribution of the current analysis period includes: Extract the energy percentage corresponding to each vibration frequency band in the frequency band energy distribution of the current analysis period; Compare the energy percentages described above and determine the target energy percentage with the largest value. The vibration frequency band corresponding to the target energy percentage is determined as the dominant frequency band.

[0041] Specifically, the frequency band energy distribution is composed of the energy proportion corresponding to each vibration frequency band within the current analysis period. The energy proportion corresponding to each vibration frequency band in the current analysis period is read sequentially, the values ​​of each energy proportion are compared, the energy proportion with the largest value is determined as the target energy proportion, and the vibration frequency band corresponding to the target energy proportion is determined as the dominant frequency band of the current analysis period.

[0042] The energy percentage is directly taken from the data in the frequency band energy distribution formed in step S1, without further weighting, interval normalization or operating condition classification.

[0043] When the energy proportions of two vibration frequency bands are the same, the frequency point with the largest spectral amplitude in each vibration frequency band is read, and the spectral amplitudes corresponding to the two frequency points are compared. The vibration frequency band with the larger maximum spectral amplitude is determined as the dominant frequency band. When the two maximum spectral amplitudes are still the same, the vibration frequency band with the lower frequency position is determined as the dominant frequency band according to the arrangement order of the vibration frequency bands, thereby ensuring that the same set of input data corresponds to a unique dominant frequency band.

[0044] For example, if the energy proportions of the three vibration frequency bands in the current analysis period are 15%, 60%, and 25%, respectively, then 60% is determined as the target energy proportion, and the second vibration frequency band from 50 Hz to less than 200 Hz is determined as the dominant frequency band in the current analysis period.

[0045] Extracting the dominant frequency corresponding to the dominant frequency band from the vibration spectrum of the current analysis period includes: The target frequency range in the vibration spectrum is determined based on the frequency range of the dominant frequency band. Compare the spectral amplitudes corresponding to each frequency point within the target frequency range to determine the target frequency point with the largest spectral amplitude; The frequency corresponding to the target frequency point is determined as the dominant frequency.

[0046] Specifically, the vibration spectrum already formed in the current analysis period is read, and all discrete frequency points falling into the corresponding frequency range are selected according to the lower and upper frequency boundaries of the dominant frequency band. The selected discrete frequency points are then combined to form the target frequency interval without performing a fast Fourier transform again.

[0047] It should be noted that when the dominant frequency band is the first vibration frequency band including 0 Hz, the target frequency range does not include the DC frequency point corresponding to 0 Hz, and the target frequency point is determined from discrete frequency points greater than 0 Hz. The DC frequency point is not used as the dominant frequency of the vibration suppression filter node.

[0048] In one implementation, the current maximum spectral amplitude and corresponding frequency are recorded starting from the first valid frequency point in the target frequency range, and subsequent frequency points are read in frequency order. When the spectral amplitude of a subsequent frequency point is greater than the currently recorded maximum spectral amplitude, the spectral amplitude and frequency corresponding to the subsequent frequency point are updated to the currently recorded values. After comparing all frequency points within the target frequency range, the finally recorded frequency point is determined as the target frequency point, and the frequency corresponding to the target frequency point is determined as the dominant frequency.

[0049] When two or more frequency points have the same maximum spectral amplitude, the frequency point with the smallest frequency value is selected as the target frequency point.

[0050] For example, if the dominant frequency band of the current analysis period is 50 Hz to less than 200 Hz, and the frequency point corresponding to 156.25 Hz has the largest spectral amplitude in the vibration spectrum corresponding to the dominant frequency band, then the frequency point corresponding to 156.25 Hz is determined as the target frequency point, and 156.25 Hz is determined as the dominant frequency of the current analysis period.

[0051] According to the 2000 Hz sampling frequency and 1024-point fast Fourier transform settings in step S1, the frequency interval between adjacent discrete frequency points is approximately 1.953125 Hz. Therefore, the dominant frequency is directly adopted from the discrete frequency points in the vibration spectrum, without additional interpolation between adjacent frequency points.

[0052] The dominant frequency band records the frequency range where the current main vibration component is located, and the dominant frequency records the specific frequency position with the largest spectral amplitude within the frequency range. Subsequently, vibration suppression filter nodes are determined based on the dominant frequency band, and the target center frequency is determined based on the dominant frequency.

[0053] In step S3, the energy proportion corresponding to the dominant frequency band of each analysis period is extracted from the frequency band energy distribution of the continuous analysis period, the energy proportions are arranged in chronological order to obtain the dominant energy sequence, and the state of the dominant energy sequence is estimated to obtain the energy state.

[0054] Specifically, the energy percentage corresponding to the dominant frequency band for each analysis period is extracted from the frequency band energy distribution of the continuous analysis period, and the energy percentages are arranged in chronological order to obtain the dominant energy sequence, including: Determine the dominant frequency band for each analysis period in the continuous analysis cycle; Extract the energy percentage of the dominant frequency band from the frequency band energy distribution corresponding to each analysis period; The dominant energy sequence is obtained by arranging the corresponding energy percentages according to the time sequence of each analysis period.

[0055] Specifically, the continuous analysis cycle consists of multiple effective analysis cycles formed sequentially in time. Each effective analysis cycle has a corresponding vibration spectrum and frequency band energy distribution, and each determines its own dominant frequency band in the same manner as in step S2.

[0056] Each time a new effective analysis period is formed, the dominant frequency band corresponding to the current analysis period is determined from the frequency band energy distribution of the current analysis period, and the energy percentage corresponding to the dominant frequency band is read. The energy percentage is then appended to the end of the existing sequence according to the time sequence of the analysis period to form the dominant energy sequence.

[0057] Each analysis period extracts the energy percentage corresponding to its own dominant frequency band, without using the dominant frequency band of the current analysis period to read the energy percentage of the same frequency range in previous analysis periods. Therefore, when the dominant frequency band changes in consecutive analysis periods, adjacent data in the dominant energy sequence correspond to the dominant frequency band of the adjacent analysis period.

[0058] For example, the energy percentages of the three vibration frequency bands in three consecutive analysis cycles are 65%, 25%, 10%, 30%, 55%, 15%, and 20%, 60%, 20%, respectively. The dominant frequency band energy percentages for the three analysis cycles are 65%, 55%, and 60%, respectively. Arranged in chronological order according to the analysis cycles, a dominant energy sequence consisting of 65%, 55%, and 60% is obtained.

[0059] The dominant energy sequence records the proportion of the frequency band energy of the dominant frequency band to the total energy of the analysis frequency band in each analysis period. Therefore, it represents the relative energy state of the main vibration component in the entire analysis frequency band, and does not represent the absolute mechanical vibration energy collected by the sensor.

[0060] Specifically, the energy states are obtained by performing state estimation on the dominant energy sequence, including: The energy percentage corresponding to each analysis period in the dominant energy sequence is taken as the observation. Kalman filtering is used to predict the energy value of the current analysis period based on the estimated energy value of the previous analysis period, thus obtaining the predicted energy value. The deviation between the observed value and the predicted energy value in the current analysis period is calculated, the predicted energy value is corrected according to the deviation to obtain the estimated energy value, and the estimated energy value in the current analysis period is determined as the energy state.

[0061] Specifically, this embodiment employs a one-dimensional scalar Kalman filter, where both state variables and observations use dimensionless energy percentages within the range of 0 to 1. The estimated energy value corrected from the previous analysis period serves as the basis for state prediction in the current analysis period, with the state transition coefficient set to 1; the dominant frequency band energy percentage in the current analysis period is used as the observation, with the observation coefficient set to 1.

[0062] Kalman filtering is calculated using the following formula: ; ; ; ; ; in, This represents the observation corresponding to the i-th analysis period, i.e., the energy percentage corresponding to the dominant frequency band in the i-th analysis period; This represents the predicted energy value for the i-th analysis period; This represents the estimated energy value after correction in the i-th analysis period; This represents the variance of the prediction error in the i-th analysis period; This represents the variance of the estimation error after correction in the i-th analysis period; Indicates the process noise variance; Indicates the variance of observation noise; This represents the Kalman gain in the i-th analysis period.

[0063] The observation noise variance This is determined based on stable gimbal operation data. Within a continuous operating range where the gimbal drive commands and load status remain unchanged, the dominant frequency band energy percentages corresponding to multiple effective analysis cycles are formed according to steps S1 and S2. The arithmetic mean of all these energy percentages is calculated, and then the variance of each energy percentage relative to the arithmetic mean is calculated. The resulting variance is determined as the observation noise variance. .

[0064] The stable operating interval is directly adopted as the time interval during which the same set of drive commands in the gimbal control program remains unchanged, without setting a separate stable state threshold. When the value of the drive command in the gimbal control program changes, the current stable operating interval ends, and the new stable operating interval restarts from the moment the drive command last changed. A stable operating interval contains at least two valid analysis periods; when there are multiple consecutive stable operating intervals, the valid analysis periods in all stable operating intervals are included in the calculation of the observation noise variance.

[0065] The process noise variance The determination is based on the data of changes in the gimbal's operating status. During the process of the gimbal drive mechanism continuously changing its operating status according to the working range set in the control program, the energy percentage of the dominant frequency band corresponding to each effective analysis cycle is recorded. The difference between the energy percentages of adjacent analysis cycles is calculated in chronological order, and the variance of all such differences is calculated to obtain the variance of adjacent observation changes.

[0066] When vibration data segments from adjacent analysis periods do not overlap, adjacent observation noises are independent of each other, and twice the observation noise variance is subtracted from the variance of the adjacent observation changes. The obtained non-negative result is determined as the process noise variance. When the result is less than zero, the process noise variance is... Set to zero; when vibration data segments from adjacent analysis periods overlap, recalculate the variance of adjacent observations using non-overlapping data segments, and then perform the above calculation of subtracting twice the observation noise variance, or set the process noise variance to zero. Directly set as observation noise variance The value ranges from 0.1 to 1.0 times; the specific multiplier is determined based on the response speed of the gimbal drive mechanism. A larger multiplier is selected when the drive mechanism has a faster response speed, and a smaller multiplier is selected when the response speed is slower. A typical value is 0.5 times.

[0067] Both the observation noise variance and the process noise variance are calculated using data generated from the same sensors, sampling frequencies, vibration data segment lengths, vibration frequency band division methods, and dominant frequency band determination methods as during the online operation phase. If the sensor, sampling configuration, or vibration frequency band configuration changes, the above parameter determination process is re-executed; if the vibration frequency band configuration changes, the frequency band node correspondence and energy attenuation correspondence are synchronously re-established.

[0068] During Kalman filter initialization, the energy percentage corresponding to the first effective analysis period in the dominant energy sequence is determined as the initial estimated energy value, and the observation noise variance is... This is determined as the initial estimate of the error variance. Starting from the second effective analysis period, it is updated periodically according to the above recursive relationship.

[0069] When the observation noise variance When the value is zero, it indicates that the dominant frequency band energy proportion in the stable operating data has not shown any observed fluctuations under the current digital precision. In the current analysis cycle, the corresponding observation is directly determined as the energy state, and the Kalman gain division calculation is not performed.

[0070] For example, under the condition that adjacent vibration data segments do not overlap, the variance of the observed noise is calculated from a set of stable operating data. The variance of the change in adjacent observations is calculated to be 0.0054, given a set of operational status change data. Subtracting twice the value of 0.0025 from 0.0054 yields the process noise variance. It is 0.0004.

[0071] The estimated energy value for the previous analysis period was 0.55, and the estimated error variance for the previous analysis period was 0.0025. When the observed value for the current analysis period is 0.60, 0.55 is first used as the predicted energy value for the current analysis period. 0.0025 is added to 0.0004 to obtain the prediction error variance of 0.0029. Then, 0.0029 is divided by the sum of 0.0029 and 0.0025 to obtain a Kalman gain of approximately 0.5370. The difference between 0.60 and 0.55, 0.05, is multiplied by 0.5370 and then added to 0.55 to obtain an estimated energy value of approximately 0.5769. The value of 0.5769 is determined as the energy state for the current analysis period.

[0072] Each time a new effective analysis period is formed, only the estimated energy value and estimated error variance of the previous analysis period are read, and a recursion is completed by combining the observations corresponding to the current analysis period, without recalculating the entire historical dominant energy sequence.

[0073] In step S4, based on the dominant frequency band of the current analysis period, a corresponding vibration suppression filter node is determined in the gimbal control loop of the integrated gimbal. The target center frequency is determined based on the dominant frequency, and the target attenuation depth is determined based on the energy state. The target center frequency and the target attenuation depth are combined to obtain the gimbal suppression parameters.

[0074] Specifically, based on the dominant frequency band of the current analysis period, the corresponding vibration suppression filter node is determined in the gimbal control loop of the integrated gimbal, including: Based on the preset frequency band node correspondence, the dominant frequency band of the current analysis period is matched with nodes to obtain the corresponding filter node identifier; The filter node corresponding to the filter node identifier is identified as the vibration suppression filter node.

[0075] Specifically, the vibration suppression filter node is a digital notch filter configuration node in the PTZ control loop. Each vibration suppression filter node has a corresponding filter node identifier, adjustable center frequency lower boundary, adjustable center frequency upper boundary, minimum attenuation depth, maximum attenuation depth, and initialization filter parameters, which are recorded in the filter node configuration table of the PTZ control program.

[0076] The filter node configuration table is established based on the parameter interfaces of each digital notch filter node in the PTZ control loop. During system initialization, the node number, center frequency writeable range, attenuation depth writeable range, and initial quality factor corresponding to each digital notch filter node are read sequentially, and these parameters are written into the same node configuration record. All node configuration records constitute the filter node configuration table.

[0077] The preset frequency band node correspondence is established based on the filter node configuration table. The vibration suppression filter nodes are arranged from low to high according to their center frequency range, and the center frequency range corresponding to each vibration suppression filter node is determined as the corresponding vibration frequency band. The vibration frequency band and the corresponding filter node identifier are written into the same configuration record.

[0078] For example, the gimbal control loop is equipped with three vibration suppression filter nodes. The filter node of the first vibration suppression filter node is identified as FN01, and its center frequency range is 0 Hz to less than 50 Hz; the filter node of the second vibration suppression filter node is identified as FN02, and its center frequency range is 50 Hz to less than 200 Hz; the filter node of the third vibration suppression filter node is identified as FN03, and its center frequency range is 200 Hz to 500 Hz.

[0079] The vibration frequency band division in step S1 and the frequency band node correspondence in this step are read from the same filter node configuration table, therefore the same frequency boundaries and boundary assignment relationships are used. 50 Hz is assigned to the second vibration frequency band corresponding to FN02, 200 Hz is assigned to the third vibration frequency band corresponding to FN03, and 500 Hz is assigned to the third vibration frequency band corresponding to FN03.

[0080] After the dominant frequency band is formed in the current analysis period, the frequency range corresponding to the dominant frequency band is matched in the frequency band node correspondence relationship, the filter node identifier in the matching record is read, and then the digital notch filter configuration node corresponding to the filter node identifier is determined as the vibration suppression filter node of the current analysis period.

[0081] For example, if the dominant frequency band of the current analysis period is 50 Hz to less than 200 Hz, then the filter node identifier FN02 is matched and the digital notch filter configuration node corresponding to FN02 is determined as the vibration suppression filter node of the current analysis period.

[0082] The filtering parameters in the vibration suppression filter node, except for the center frequency and attenuation depth, adopt the initialization parameters in the filter node configuration table. In this embodiment, the quality factor is not dynamically modified; it remains unchanged while the current node is selected. The quality factor is determined based on the effective bandwidth of the vibration suppression filter node and is directly set by the initialization quality factor in the filter node configuration table. For example, the value ranges from 1 to 10. A larger quality factor results in a narrower filtering bandwidth, and a smaller quality factor results in a wider filtering bandwidth. The specific value is determined based on the vibration modal frequency distribution of the gimbal support structure.

[0083] Specifically, the target center frequency is determined based on the dominant frequency, the target attenuation depth is determined based on the energy state, and the gimbal suppression parameters are obtained by combining the target center frequency and the target attenuation depth, including: The dominant frequency is determined as the target center frequency of the vibration suppression filter node; Based on a preset energy attenuation correspondence, the energy state is attenuated and mapped to obtain the target attenuation depth of the vibration suppression filter node. By combining the target center frequency and the target attenuation depth, the gimbal suppression parameters are obtained.

[0084] Specifically, the dominant frequency obtained in step S2 is within the frequency range of the current dominant frequency band, and the adjustable center frequency range of the current vibration suppression filter node corresponds to the dominant frequency band. Therefore, the dominant frequency is directly determined as the target center frequency of the current vibration suppression filter node.

[0085] For example, if the current dominant frequency band corresponds to the filter node identifier FN02, and the dominant frequency obtained in step S2 is 156.25 Hz, then 156.25 Hz is determined as the target center frequency corresponding to FN02.

[0086] The preset energy attenuation correspondence is established based on the number of vibration frequency bands and the attenuation depth range that the current vibration suppression filter node is allowed to write.

[0087] Since the sum of the energy proportions corresponding to each vibration frequency band is 1, and the energy state is obtained by state estimation from the energy proportion with the largest value in each analysis period, under the condition that each vibration frequency band may carry the dominant vibration energy, when the number of vibration frequency bands is M, the theoretical lower boundary of the energy proportion of the dominant frequency band is 1 / M, and the theoretical upper boundary is 1.

[0088] During attenuation mapping, the current energy state is first read, then 1 / M is subtracted from it, and finally divided by the difference between 1 and 1 / M. The result is determined as the current dominance level. The current dominance level is between 0 and 1. When the current energy state equals 1 / M, the current dominance level is 0; when the current energy state equals 1, the current dominance level is 1.

[0089] Then, the minimum and maximum attenuation depths corresponding to the current vibration suppression filter node are read, the difference between the maximum and minimum attenuation depths is calculated, the difference is proportionally adjusted according to the current dominance level, and then added to the minimum attenuation depth. The resulting value is determined as the target attenuation depth.

[0090] The above calculation is a linear interpolation of the dimensionless dominance within the allowable attenuation parameter range of the filter node, without directly converting the frequency band energy or energy proportion into decibels.

[0091] For example, this embodiment includes three vibration frequency bands, so the theoretical lower boundary of the energy proportion of the dominant frequency band is one-third. Step S3 obtains the energy state of the current analysis cycle as 0.5769, then first calculates the difference between 0.5769 and one-third, and then divides it by the difference between 1 and one-third to obtain the current dominance level of approximately 0.365.

[0092] When the filter node configuration table corresponding to FN02 records a minimum attenuation depth of 0 dB and a maximum attenuation depth of 20 dB, the difference between 20 dB and 0 dB is calculated. The difference is then taken at a ratio of approximately 0.365 to obtain approximately 7.30 dB. This 7.30 dB is determined as the target attenuation depth corresponding to the current analysis period.

[0093] The 0 dB and 20 dB values ​​are derived from the filter node configuration record corresponding to the current FN02. Other vibration suppression filter nodes read the minimum and maximum attenuation depths from their own configuration records and determine the target attenuation depth according to the same mapping process.

[0094] When the current dominance level calculated numerically is less than 0, the current dominance level is set to 0; when the current dominance level is greater than 1, the current dominance level is set to 1. This processing only limits the floating-point rounding deviation caused by state estimation and does not introduce an additional state determination threshold.

[0095] After determining the target center frequency and target attenuation depth, the two are combined according to the field order of target center frequency and target attenuation depth to obtain the gimbal suppression parameters corresponding to the current analysis period.

[0096] For example, if the current vibration suppression filter node is FN02, the target center frequency is 156.25 Hz, and the target attenuation depth is approximately 7.30 dB, then 156.25 Hz and 7.30 dB are combined to form the gimbal suppression parameters corresponding to the current analysis cycle.

[0097] In step S5, the vibration suppression control command is obtained by encapsulating the command based on the gimbal suppression parameters and the filter node identifier of the vibration suppression filter node, including: The target center frequency and the target attenuation depth are associated with the filter node identifier to obtain filter configuration data; The filter configuration data is encapsulated according to the control command format of the pan-tilt control loop to obtain the vibration suppression control command.

[0098] Specifically, the filter configuration data consists of the filter node identifier, target center frequency, and target attenuation depth corresponding to the same analysis cycle. The filter node identifier comes from the vibration suppression filter node determined in step S4, and the target center frequency and target attenuation depth come from the gimbal suppression parameters formed in step S4.

[0099] In one implementation, the gimbal control program sets up a parameter writing interface for the vibration suppression filter node. The data structure of the parameter writing interface includes a node field, a center frequency field, and an attenuation depth field. The node field corresponds to the filter node identifier, the center frequency field corresponds to the target center frequency, and the attenuation depth field corresponds to the target attenuation depth.

[0100] The node field adopts the existing numbering format of the filter nodes in the PTZ control program; the center frequency field uses the same data type and Hertz unit as the center frequency parameter of the vibration suppression filter node; the attenuation depth field uses the same data type and parameter unit as the attenuation depth parameter of the vibration suppression filter node. This step does not establish additional node numbering rules, center frequency conversion relationships, or attenuation depth conversion relationships.

[0101] The filter node identifier, target center frequency, and target attenuation depth corresponding to the current analysis period are associated with the same parameter record to obtain filter configuration data. One effective analysis period corresponds to a set of filter configuration data for the current vibration suppression filter node, and parameters from different analysis periods are not cross-combined.

[0102] Subsequently, following the field order specified by the parameter writing interface, the filter node identifier is written into the node field, the target center frequency is written into the center frequency field, and the target attenuation depth is written into the attenuation depth field. After all fields are written, the resulting parameter writing structure is determined as the vibration suppression control command.

[0103] In one implementation, the PTZ control loop sets only the vibration suppression filter node corresponding to the node field as the currently active filter node in any analysis cycle, while vibration suppression filter nodes not selected by the node field are in a bypass state. Therefore, when the dominant frequency band changes from one vibration frequency band to another, the currently active filter node is switched by changing the filter node identifier, without retaining the attenuation effect of the previous filter node on the current vibration signal. As a switching transition process, during the control cycle of the filter node switching, the parameter configuration of the filter node before the switch remains unchanged and attenuation continues to be applied. After the parameters of the filter node after the switch are written and stabilized, the filter node before the switch is set to a bypass state, thereby avoiding interruption of the filtering effect at the moment of switching.

[0104] It should be noted that the selection relationship of the active filter nodes is written into the parameter interface definition during the initialization of the PTZ control loop. After reading the node field, the parameter interface sets the nodes with the same filter node identifier as the node field to the active state, and sets the remaining nodes to the bypass state, without performing additional frequency determination.

[0105] For example, if the filter node identifier corresponding to the current analysis period is FN02, the target center frequency is 156.25 Hz, and the target attenuation depth is 7.30 dB, then the current filter configuration data records FN02, 156.25 Hz, and 7.30 dB. After writing the fields in the order of node field, center frequency field, and attenuation depth field, the vibration suppression control command corresponding to the current analysis period is obtained. When the node field records FN02, the vibration suppression filter node corresponding to FN02 is in an active state, and the other vibration suppression filter nodes are in a bypass state.

[0106] After a new frequency band energy distribution is formed in the next effective analysis cycle, the corresponding dominant frequency band and dominant frequency are re-determined, and the energy state is updated according to step S3. When the filter node identifier, target center frequency, or target attenuation depth changes in the next analysis cycle, the new parameters formed in the next analysis cycle are rewritten into the parameter writing structure to obtain the updated vibration suppression control command.

[0107] The instruction encapsulation process only associates and writes the filter node identifier, target center frequency, and target attenuation depth already formed in step S4, without re-executing frequency domain analysis, dominant frequency band determination, dominant frequency extraction, state estimation, or attenuation mapping.

[0108] In summary, this invention performs segmented frequency domain analysis on the vibration signal of an integrated gimbal during operation, determines the dominant frequency band based on the frequency band energy distribution, extracts the corresponding dominant frequency, constructs a dominant energy sequence based on the energy proportion of the dominant frequency band in a continuous analysis period, and performs state estimation to obtain the energy state. Furthermore, it determines vibration suppression filter nodes based on the dominant frequency band, determines the target center frequency based on the dominant frequency, and determines the target attenuation depth based on the energy state. The resulting gimbal suppression parameters are encapsulated as vibration suppression control commands. This allows the center frequency and attenuation depth of the vibration suppression filter nodes to be adjusted according to the current main vibration frequency and relative energy state, improving the matching degree between the integrated gimbal vibration suppression parameters and the current main disturbance.

[0109] A second embodiment of the present invention provides an adaptive vibration suppression system for an integrated gimbal, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0110] It should be noted that the adaptive vibration suppression system for integrated gimbals provided in this embodiment of the invention is used to execute all the process steps of the adaptive vibration suppression method for integrated gimbals described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0111] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An adaptive vibration suppression method for integrated gimbals, characterized in that, include: The vibration signal of the integrated gimbal is acquired by the sensor during operation, and the vibration signal is analyzed in segmented frequency domain to obtain the vibration spectrum and frequency band energy distribution corresponding to each analysis period. The dominant frequency band is determined based on the frequency band energy distribution of the current analysis period, and the dominant frequency corresponding to the dominant frequency band is extracted from the vibration spectrum of the current analysis period. The energy percentage corresponding to the dominant frequency band in each analysis period is extracted from the frequency band energy distribution of the continuous analysis period. The energy percentages are arranged in chronological order to obtain the dominant energy sequence. The state of the dominant energy sequence is estimated to obtain the energy state. Based on the dominant frequency band of the current analysis period, a corresponding vibration suppression filter node is determined in the gimbal control loop of the integrated gimbal. The target center frequency is determined based on the dominant frequency, and the target attenuation depth is determined based on the energy state. The target center frequency and the target attenuation depth are combined to obtain the gimbal suppression parameters. The vibration suppression control command is obtained by encapsulating the command based on the gimbal suppression parameters and the filter node identifier of the vibration suppression filter node.

2. The adaptive vibration suppression method for integrated gimbals according to claim 1, characterized in that, The step of performing segmented frequency domain analysis on the gimbal vibration signal to obtain the vibration spectrum and frequency band energy distribution corresponding to each analysis period includes: The gimbal vibration signal is segmented and windowed to obtain vibration data segments corresponding to each analysis period; Perform a fast Fourier transform on the vibration data segment to obtain the vibration spectrum; The vibration spectrum is divided into multiple vibration frequency bands according to the frequency range, and the spectral amplitude in each vibration frequency band is squared and accumulated to obtain the corresponding frequency band energy. Calculate the proportion of each frequency band energy in the total frequency band energy to obtain the frequency band energy distribution.

3. The adaptive vibration suppression method for integrated gimbals according to claim 1, characterized in that, The process of determining the dominant frequency band based on the frequency band energy distribution of the current analysis period includes: Extract the energy percentage corresponding to each vibration frequency band in the frequency band energy distribution of the current analysis period; Compare the energy percentages described above and determine the target energy percentage with the largest value. The vibration frequency band corresponding to the target energy percentage is determined as the dominant frequency band.

4. The adaptive vibration suppression method for an integrated gimbal according to claim 3, characterized in that, Extracting the dominant frequency corresponding to the dominant frequency band from the vibration spectrum of the current analysis period includes: The target frequency range in the vibration spectrum is determined based on the frequency range of the dominant frequency band. By comparing the spectral amplitudes of each frequency point within the target frequency range, the target frequency point with the largest spectral amplitude is determined. The frequency corresponding to the target frequency point is determined as the dominant frequency.

5. The adaptive vibration suppression method for an integrated gimbal according to claim 3, characterized in that, The step of extracting the energy percentage corresponding to the dominant frequency band for each analysis period from the frequency band energy distribution of continuous analysis periods, and arranging the energy percentages in chronological order to obtain the dominant energy sequence includes: Determine the dominant frequency band for each analysis period in the continuous analysis cycle; Extract the energy percentage of the dominant frequency band from the frequency band energy distribution corresponding to each analysis period; The dominant energy sequence is obtained by arranging the corresponding energy percentages according to the time sequence of each analysis period.

6. The adaptive vibration suppression method for an integrated gimbal according to claim 5, characterized in that, The process of estimating the state of the dominant energy sequence to obtain the energy state includes: The energy percentage corresponding to each analysis period in the dominant energy sequence is taken as the observation. Kalman filtering is used to predict the energy value of the current analysis period based on the estimated energy value of the previous analysis period, thus obtaining the predicted energy value. The deviation between the observed value and the predicted energy value in the current analysis period is calculated, the predicted energy value is corrected according to the deviation to obtain the estimated energy value, and the estimated energy value in the current analysis period is determined as the energy state.

7. The adaptive vibration suppression method for an integrated gimbal according to claim 1, characterized in that, The step of determining the corresponding vibration suppression filter node in the gimbal control loop of the integrated gimbal based on the dominant frequency band of the current analysis period includes: Based on the preset frequency band node correspondence, the dominant frequency band in the current analysis period is matched with nodes to obtain the corresponding filter node identifier; The filter node corresponding to the filter node identifier is identified as the vibration suppression filter node.

8. The adaptive vibration suppression method for an integrated gimbal according to claim 1, characterized in that, The process of determining the target center frequency based on the dominant frequency, determining the target attenuation depth based on the energy state, and combining the target center frequency and the target attenuation depth to obtain gimbal suppression parameters includes: The dominant frequency is determined as the target center frequency of the vibration suppression filter node; Based on a preset energy attenuation correspondence, the energy state is attenuated and mapped to obtain the target attenuation depth of the vibration suppression filter node. By combining the target center frequency and the target attenuation depth, the gimbal suppression parameters are obtained.

9. The adaptive vibration suppression method for an integrated gimbal according to claim 1, characterized in that, The step of encapsulating instructions based on the gimbal suppression parameters and the filter node identifier of the vibration suppression filter node to obtain vibration suppression control instructions includes: The target center frequency and the target attenuation depth are associated with the filter node identifier to obtain filter configuration data; The filter configuration data is encapsulated according to the control command format of the pan-tilt control loop to obtain the vibration suppression control command.

10. An adaptive vibration suppression system for an integrated pan-tilt unit, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1 to 9.

Images