Vibration suppression method and device for high-altitude operation machine and high-altitude operation machine

Abstract

The embodiment of the invention provides a vibration suppression method and device for a high-altitude operation machine and the high-altitude operation machine, and relates to the technical field of mechanical control. The method comprises the following steps: acquiring a low-dimensional motion signal of a working platform, wherein the low-dimensional motion signal is obtained by compressing an original dynamic signal based on a self-adaptive observation matrix; obtaining a sparse activation vector and determining a reconstruction signal according to the low-dimensional motion signal, the adaptive observation matrix and a pre-trained somatosensory event physical dictionary; determining a target swing frequency in a frequency spectrum corresponding to the reconstruction signal according to a preset swing frequency library so as to generate an enhancement vector reflecting the target swing frequency, generating a suppression signal according to a working platform motion parameter determined by the reconstruction signal, and generating an exploration vector according to a somatosensory event physical dictionary and a sparse activation vector; and updating the adaptive observation matrix according to the enhancement vector and the exploration vector, and sending a suppression signal to a motion control unit of the working platform. According to the method, real-time and self-adaptive vibration suppression can be realized to optimize the somatosensory.

Description

Technical Field

[0001] This application relates to the field of mechanical control technology, specifically to a vibration suppression method, device, and aerial work platform for use in aerial work platforms. Background Technology

[0002] Aerial work platforms, as specialized engineering equipment that transports personnel, tools, and materials to high altitudes for operations, directly impact the smoothness, stability, and responsiveness of their operation—aspect-wise, "consistency of operation"—affecting the safety, comfort, and work efficiency of the workers. Current technologies for improving this consistency primarily rely on passive design and offline optimization. For example, increasing the rigidity of the boom structure can suppress vibration, but this significantly increases the equipment's weight and cost; or optimizing the valve characteristics of hydraulic valves and repeatedly, time-consumingly debugging PID control parameters on-site. However, these existing technologies suffer from parameter fixation, failing to adapt to the dynamic changes in the aerial work platform's characteristics under different boom lengths, loads, and operating speeds, and making it difficult to achieve vibration suppression under various working conditions. Summary of the Invention

[0003] The purpose of this application is to provide a vibration suppression method, device, and aerial work platform for use in aerial work platforms.

[0004] To achieve the above objectives, the first aspect of this application provides a vibration suppression method for aerial work platforms. The aerial work platforms include a work platform equipped with an inertial measurement unit (IMU) for acquiring the raw dynamic signals of the work platform. The vibration suppression method includes: The low-dimensional motion signal of the working platform is acquired. The low-dimensional motion signal is obtained by the inertial measurement unit compressing the original dynamic signal using the current adaptive observation matrix. Based on low-dimensional motion signals, adaptive observation matrices, and pre-trained physical dictionaries of somatosensory events, sparse activation vectors are obtained through a sparse reconstruction algorithm. The reconstructed signals are then determined based on the sparse activation vectors and the physical dictionaries of somatosensory events. The physical dictionaries of somatosensory events are trained using a dictionary learning algorithm based on simulation data of high-altitude work machinery, and their atomic representations are the basic dynamic event patterns. Based on the preset swing frequency library, the target swing frequency in the spectrum corresponding to the reconstructed signal is determined to generate an enhancement vector that reflects the target swing frequency. An inhibition signal is generated based on the motion parameters of the working platform determined by the reconstructed signal, and an exploration vector is generated based on the physical dictionary of haptic events and sparse activation vectors. The adaptive observation matrix is ​​updated based on the enhancement vector and the exploration vector, and the suppression signal is sent to the motion control unit of the working platform so that the motion control unit can suppress the swaying of the working platform based on the suppression signal.

[0005] In this embodiment, determining the target swing frequency in the spectrum corresponding to the reconstructed signal based on a preset swing frequency library to generate an enhancement vector reflecting the target swing frequency, generating a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal, and generating an exploration vector based on the physical dictionary of somatosensory events and sparse activation vectors include: determining the target swing frequency in the spectrum corresponding to the reconstructed signal based on the preset swing frequency library, and determining the confidence level of the target swing frequency; if the confidence level is greater than a preset confidence threshold, generating an enhancement vector reflecting the target swing frequency, and generating a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal; if the sparse activation vector does not exist in the historical dynamic pattern memory library, generating an exploration vector based on the physical dictionary of somatosensory events and sparse activation vectors.

[0006] In this embodiment, determining the target swing frequency in the spectrum corresponding to the reconstructed signal based on a preset swing frequency library to generate an enhancement vector reflecting the target swing frequency, generating a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal, and generating an exploration vector based on the physical dictionary of somatosensory events and sparse activation vectors include: determining the target swing frequency in the spectrum corresponding to the reconstructed signal based on the preset swing frequency library, and determining the confidence level of the target swing frequency; if the confidence level is greater than a preset confidence threshold, generating an enhancement vector reflecting the target swing frequency, and generating a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal; if the sparse activation vector does not exist in the historical dynamic pattern memory library, generating an exploration vector based on the physical dictionary of somatosensory events and sparse activation vectors.

[0007] In this embodiment, the historical dynamic pattern memory includes hash fingerprints of sparse activation vectors that have appeared historically. When the sparse activation vector does not exist in the historical dynamic pattern memory, generating an exploration vector based on the haptic event physical dictionary and the sparse activation vector includes: querying the sparse activation vector in the historical dynamic pattern memory; generating a first hash fingerprint of the sparse activation vector; determining the Hamming distance between the hash fingerprint and the first hash fingerprint for each hash fingerprint in the historical dynamic pattern memory; and determining that the sparse activation vector does not exist in the historical dynamic pattern memory when all Hamming distances are greater than a preset distance threshold, and generating an exploration vector based on the haptic event physical dictionary and the sparse activation vector.

[0008] In this embodiment of the application, generating an enhancement vector that reflects the target swing frequency includes: generating a Hanning window centered on the target swing frequency as an enhancement vector based on a preset bandwidth.

[0009] In this embodiment of the application, updating the adaptive observation matrix based on the enhancement vector and the exploration vector includes updating the adaptive observation matrix based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector.

[0010] In this embodiment, updating the adaptive observation matrix based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector includes: weighted fusion of the enhancement vector and the exploration vector to obtain the update vector; obtaining the update step size based on the gradient ascent strategy that maximizes the projection energy of the adaptive observation matrix in the direction of the update vector; and processing the sum of the adaptive observation matrix and the update step size based on the Gram-Schmidt orthogonalization algorithm to obtain the updated adaptive observation matrix.

[0011] In this embodiment, generating a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal includes: solving the reconstructed signal to obtain the motion speed of the working platform; obtaining a damping force opposite to the direction of the motion speed based on the product of the magnitude of the motion speed and a preset damping gain coefficient; generating a suppression signal for suppressing the swaying of the working platform based on the damping force; and sending the suppression signal to the motion control unit of the working platform so that the motion control unit suppresses the swaying of the working platform based on the suppression signal.

[0012] In this embodiment of the application, the vibration suppression method further includes: determining the sum of activation energies of impact event atoms activated by sparse activation vectors in the physical dictionary of haptic events; and reducing the response sensitivity of the work platform to motion control signals when the sum of activation energies is greater than a preset energy threshold.

[0013] In this embodiment of the application, determining the total activation energy of the impact event atoms activated by the sparse activation vector in the physical dictionary of haptic events includes: determining the activation coefficients corresponding to the activated impact event atoms in the sparse activation vector; and determining the sum of squares of each activation coefficient to obtain the total activation energy.

[0014] In this embodiment of the application, when the total activation energy is greater than a preset energy threshold, reducing the response sensitivity of the work platform to the motion control signal includes: acquiring the motion control signal; when the total activation energy is greater than the preset energy threshold, reducing the slope of the motion control signal in the time domain according to a preset inhibition factor to obtain a weakened motion control signal; and sending the weakened motion control signal to the motion control unit of the work platform so that the motion control unit controls the movement of the work platform according to the weakened motion control signal.

[0015] In this embodiment, the inertial measurement unit determines the low-dimensional motion signal based on the following steps: monitoring the motion behavior of the working platform to obtain the original motion signal; processing the original motion signal by removing the mean and normalizing the variance to obtain the motion signal; compressing the motion signal according to the adaptive observation matrix to obtain the low-dimensional motion signal.

[0016] In this embodiment, the physical dictionary for motion-sensing events is obtained based on the following steps: acquiring multiple training data segments, each representing a motion signal of the working platform under a specific motion behavior; iterating through the K-SVD algorithm based on the multiple training data segments to obtain the final dictionary and the sparsest activation vector of each training data segment; for each atom in the final dictionary, determining the training data segment with the maximum activation sparsity for the atom based on the sparsest activation vector of each training data segment, thus obtaining the activation data segment of the atom; if the proportion of data segments representing the same motion behavior in the activation data segments of the atom is greater than a preset proportion threshold, determining the atom as a valid atom representing a specific motion behavior; retaining each valid atom in the final dictionary to obtain the physical dictionary for motion-sensing events.

[0017] In this embodiment, obtaining the final dictionary and the sparsest activation vector of each training data segment through iterative K-SVD algorithm based on multiple training data segments includes: for each training data segment, determining the sparsest activation vector of the current dictionary using a sparse coding algorithm, wherein the sparsest activation vector minimizes the reconstruction error of the training data segment and the number of non-zero terms in the sparsest activation vector is less than a preset sparsity; updating each atom in the current dictionary based on the sparsest activation vector of each training data segment to obtain an updated dictionary, and re-determining the sparsest activation vector of each training data segment based on the updated dictionary; and determining the updated dictionary as the final dictionary if the updated dictionary meets a preset termination condition, and obtaining the sparsest activation vector of each training data segment determined based on the final dictionary.

[0018] In this embodiment of the application, obtaining multiple training data segments includes: obtaining a training database, wherein the training database includes multiple training motion signals, the training motion signals representing the motion signals of the working platform under specific motion behaviors; for each training motion signal, dividing the training motion signal into multiple original training data segments with overlapping signal segments in time according to a preset signal length; for each original training data segment, processing the original training data segment by removing the mean and normalizing the variance to obtain the training data segment.

[0019] In this embodiment of the application, obtaining the training database includes: constructing a dynamic model and an electro-hydraulic control model of the aerial work platform; performing joint simulation based on the joint simulation model of the dynamic model and the electro-hydraulic control model, wherein the joint simulation model includes a virtual working platform of the aerial work platform and a virtual inertial measurement unit on the virtual working platform; during the joint simulation, recording the motion signals obtained by the virtual inertial measurement unit monitoring the motion behavior of the working platform under different working conditions as training motion signals; and saving multiple training motion signals as a training database.

[0020] The second aspect of this application provides a vibration suppression device for aerial work machinery, the vibration suppression device for aerial work machinery including: a processor configured to execute the vibration suppression method for aerial work machinery provided according to the first aspect of this application.

[0021] The third aspect of this application provides an aerial work platform, which includes: a work platform equipped with an inertial measurement unit; a vibration suppression device for aerial work platforms as provided in the second aspect of this application; and a motion control unit for controlling the movement of the work platform and for receiving suppression signals from the vibration suppression device to suppress the swaying of the work platform according to the suppression signals.

[0022] A fourth aspect of this application provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the aforementioned vibration suppression method for aerial work machinery.

[0023] Through the above technical solution, the vibration suppression method for aerial work platforms provided in this application, after obtaining the low-dimensional motion signal of the working platform, obtains a sparse activation vector based on an adaptive observation matrix and a pre-trained somatosensory event physical dictionary through a sparse reconstruction algorithm. The reconstructed signal is then determined based on the sparse activation vector and the somatosensory event physical dictionary. This generates a suppression signal to suppress platform swaying, thereby achieving vibration suppression of the aerial work platform. Simultaneously, the adaptive observation matrix is ​​updated using enhancement and exploration vectors to obtain the adaptive observation matrix used in the next vibration suppression cycle. The enhancement vector reflects the target swaying frequency of the reconstructed signal, while the exploration vector generated based on the somatosensory event physical dictionary and sparse activation vector reflects the degree of uncertainty in the current motion mode of the working platform. This allows the low-dimensional motion signal required for the next round of vibration suppression to focus on the current motion mode and target swaying frequency of the working platform based on the adaptive observation matrix. Thus, the vibration suppression method focuses on potential vibration modes in a real-time feedback manner, adapting to the dynamic changes in the working platform's motion mode under different working conditions through a real-time adjusted adaptive observation matrix, maintaining optimal vibration suppression performance, and achieving real-time and adaptive somatosensory optimization.

[0024] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0025] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings: Figure 1A schematic diagram of a high-altitude work machinery according to an embodiment of this application is shown. Figure 2 The illustration shows a schematic flow diagram of a vibration suppression method for aerial work machinery according to an embodiment of this application; Figure 3 The schematic diagram illustrates a flow chart of another vibration suppression method for aerial work machinery according to an embodiment of this application; Figure 4 The schematic diagram illustrates a flow chart of another vibration suppression method for aerial work machinery according to an embodiment of this application; Figure 5 This illustration schematically shows a training process diagram of a motion-sensing event physical dictionary according to an embodiment of this application; Figure 6 An illustration shows a motion-sensing event physical dictionary according to an embodiment of this application; Figure 7 The diagram illustrates the internal structure of a computer device according to an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0027] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0028] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0029] The acquisition, transmission, storage, use, and processing of data in this application comply with relevant laws and regulations. Furthermore, it should be noted that certain software, components, models, and other existing industry solutions may be mentioned in the embodiments of this application. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0030] Because existing technologies only suppress boom vibration by optimizing the boom structure and using PID control algorithms with fixed parameters, they cannot adapt to the dynamic changes in boom characteristics under different boom lengths, loads, and operating speeds. Therefore, existing technologies for suppressing boom vibration have obvious limitations and cannot effectively suppress boom vibration when the aerial work platform is in different working conditions, thus failing to provide the best experience for the operator.

[0031] Based on the above analysis of the prior art, this application provides a vibration suppression method for aerial work machinery. The method adjusts the original dynamic signal of the aerial work machinery working platform by using an adaptive observation matrix, and obtains a reconstructed signal and sparse activation vector by combining a pre-trained physical dictionary of somatosensory events. Based on the reconstructed signal, a suppression signal to suppress the swaying of the working platform is generated, and the adaptive observation matrix is ​​updated based on the reconstructed signal and sparse activation vector. This allows the vibration suppression method to focus on possible vibration modes in a real-time feedback manner and suppress them, thereby achieving real-time, adaptive vibration suppression and somatosensory optimization.

[0032] The vibration suppression method for aerial work machinery provided in this application can be applied to, for example... Figure 1 The aerial work platform shown is a work platform equipped with an inertial measurement unit (IMU) for acquiring raw dynamic signals from the platform. Figure 1 As shown, the inertial measurement unit (IMU) may include a six-axis IMU, i.e., a six-axis inertial sensor, which may include orthogonally arranged three-axis accelerometers and orthogonally arranged three-axis gyroscopes. The IMU may be mounted at the center of the work platform. The vibration suppression method for aerial work machinery provided in this application can be used as... Figure 1 The VECU (Vehicle Electronic Control Unit) controller shown has a built-in algorithm engine that is executed by the VECU controller. Figure 2 A schematic flowchart illustrating a vibration suppression method for aerial work machinery according to an embodiment of this application is shown. Figure 2 As shown, in one embodiment of this application, a vibration suppression method for aerial work machinery is provided. This embodiment mainly applies this method to the above-mentioned... Figure 1 Taking the VECU controller in the example, the following steps are included: S202. Obtain the low-dimensional motion signal of the working platform. The low-dimensional motion signal is obtained by the inertial measurement unit compressing the original dynamic signal using the current adaptive observation matrix.

[0033] Specifically, the original dynamic signal For example, data can be collected based on a time window, and the number of sampling points can be, for example, The inertial measurement unit may include a built-in microprocessor that utilizes the current adaptive observation matrix. (dimension is) ,like Perform compression operation low-dimensional motion signals are obtained. Low-dimensional motion signals It can be sent to the VECU controller via the CAN bus.

[0034] S204. Based on the low-dimensional motion signal, the adaptive observation matrix, and the pre-trained somatosensory event physical dictionary, a sparse activation vector is obtained through a sparse reconstruction algorithm. The reconstructed signal is then determined based on the sparse activation vector and the somatosensory event physical dictionary. The somatosensory event physical dictionary is trained using a dictionary learning algorithm based on high-altitude work machinery simulation data, and its atomic representation is the basic dynamic event pattern.

[0035] Specifically, sparse reconstruction algorithms include, for example, high-speed sparse reconstruction engines based on neural networks (such as PC-OMPNet v2), which allow the VECU controller to utilize low-dimensional motion signals. and the pre-calculated dynamic response matrix High-dimensional sparse activation vectors can be solved in, for example, within 12 milliseconds. It can selectively reconstruct a pure reconstruction signal. Sparse activation vectors The non-zero terms and their magnitudes constitute a structured description of the current platform's haptic state, among which, A physical dictionary representing haptic events.

[0036] S206. Based on the preset swing frequency library, determine the target swing frequency in the spectrum corresponding to the reconstructed signal to generate an enhancement vector that reflects the target swing frequency, generate a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal, and generate an exploration vector based on the physical dictionary of haptic events and sparse activation vectors. S208. Update the adaptive observation matrix based on the enhancement vector and the exploration vector, and send the suppression signal to the motion control unit of the working platform so that the motion control unit can suppress the swaying of the working platform based on the suppression signal.

[0037] The vibration suppression method for aerial work platforms provided in this application, after obtaining the low-dimensional motion signal of the working platform, uses an adaptive observation matrix and a pre-trained somatosensory event physical dictionary to obtain a sparse activation vector through a sparse reconstruction algorithm. Based on the sparse activation vector and the somatosensory event physical dictionary, a reconstructed signal is determined. This generates a suppression signal to inhibit platform swaying, thus suppressing vibration in the aerial work platform. Simultaneously, the adaptive observation matrix is ​​updated using enhancement and exploration vectors to obtain the adaptive observation matrix used in the next vibration suppression cycle. The enhancement vector reflects the target swaying frequency of the reconstructed signal, while the exploration vector, generated based on the somatosensory event physical dictionary and sparse activation vector, reflects the degree of uncertainty in the current motion mode of the working platform. This allows the low-dimensional motion signal required for the next round of vibration suppression to focus on the current motion mode and target swaying frequency of the working platform based on the adaptive observation matrix. This enables the vibration suppression method to focus on potential vibration modes in real-time feedback, adapting to the dynamic changes in the working platform's motion mode under different working conditions through a real-time adjusted adaptive observation matrix, maintaining optimal vibration suppression performance, and achieving real-time and adaptive somatosensory optimization.

[0038] See Figure 3 In some embodiments of this application, step S206 may include: S302. Based on the preset swing frequency library, determine the target swing frequency in the spectrum corresponding to the reconstructed signal, and determine the confidence level of the target swing frequency. S304. When the confidence level is greater than the preset confidence threshold, an enhancement vector reflecting the target swing frequency is generated, and a suppression signal is generated based on the motion parameters of the working platform determined by the reconstructed signal. S306. If the sparse activation vector does not exist in the historical dynamic pattern memory, generate an exploration vector based on the haptic event physics dictionary and the sparse activation vector.

[0039] In the above embodiments, if the confidence level is greater than a preset confidence threshold, it indicates that the reconstructed signal is concentrated at the target swing frequency in the frequency domain, and the working platform mainly vibrates at the target swing frequency. Therefore, a suppression signal is generated based on the motion parameters of the working platform determined by the reconstructed signal, achieving the technical effect of suppressing the working platform's vibration at the target swing frequency. The target swing frequency is recorded through an enhancement vector. If the sparse activation vector does not exist in the historical dynamic pattern memory, it indicates that the motion pattern represented by the sparse activation vector has not appeared in the history. Therefore, an exploration vector is generated to update the adaptive observation matrix, enabling effective capture of the motion pattern in the low-dimensional motion signal during the next round of vibration suppression.

[0040] In some embodiments of this application, determining the target swing frequency in the spectrum corresponding to the reconstructed signal according to a preset swing frequency library includes: processing the reconstructed signal based on Fast Fourier Transform to obtain the frequency domain signal of the reconstructed signal; determining the swing frequency in the preset swing frequency library that best matches the frequency domain peak of the frequency domain signal as the target swing frequency; and using the ratio of the energy value of the target swing frequency in the frequency domain signal to the average energy of the frequency domain signal as the confidence level of the target swing frequency.

[0041] Based on the above steps, the frequency domain signal of the reconstructed signal can be determined by Fast Fourier Transform, and the swing frequency in the preset swing frequency library that best matches the frequency domain peak value of the signal can be used as the target swing frequency. Specifically, the preset swing frequency library may include multiple swing frequencies, and the swing frequency that best matches the frequency domain peak value may, for example, be the swing frequency that is closest to the frequency domain peak value.

[0042] In some embodiments of this application, generating an enhancement vector that reflects the target swing frequency may include: generating a Hanning window centered on the target swing frequency as an enhancement vector based on a preset bandwidth.

[0043] In some embodiments of this application, updating the adaptive observation matrix based on the enhancement vector and the exploration vector may include updating the adaptive observation matrix based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector.

[0044] In some embodiments of this application, the historical dynamic pattern memory includes hash fingerprints of sparse activation vectors that have appeared historically. When the sparse activation vector does not exist in the historical dynamic pattern memory, generating an exploration vector based on the haptic event physical dictionary and the sparse activation vector includes: querying the sparse activation vector in the historical dynamic pattern memory; generating a first hash fingerprint of the sparse activation vector; determining the Hamming distance between the hash fingerprint and the first hash fingerprint for each hash fingerprint in the historical dynamic pattern memory; and, if all Hamming distances are greater than a preset distance threshold, determining that the sparse activation vector does not exist in the historical dynamic pattern memory, and generating an exploration vector based on the haptic event physical dictionary and the sparse activation vector.

[0045] In some embodiments of this application, step S208, updating the adaptive observation matrix based on the enhancement vector and the exploration vector, includes updating the adaptive observation matrix based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector.

[0046] In some embodiments of this application, updating the adaptive observation matrix based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector includes: weighted fusion of the enhancement vector and the exploration vector to obtain the update vector; obtaining the update step size based on the gradient ascent strategy that maximizes the projection energy of the adaptive observation matrix in the direction of the update vector; and processing the sum of the adaptive observation matrix and the update step size based on the Gram-Schmidt orthogonalization algorithm to obtain the updated adaptive observation matrix.

[0047] In some embodiments of this application, generating a suppression signal based on the motion parameters of the working platform determined by the reconstructed signal includes: solving the reconstructed signal to obtain the motion speed of the working platform; obtaining a damping force opposite to the direction of the motion speed based on the product of the magnitude of the motion speed and a preset damping gain coefficient; generating a suppression signal for suppressing the swaying of the working platform based on the damping force; and sending the suppression signal to the motion control unit of the working platform so that the motion control unit suppresses the swaying of the working platform based on the suppression signal. This includes: sending the suppression signal to the motion control unit of the working platform so that the motion control unit controls the motion of the working platform and suppresses the swaying of the working platform based on the linear superposition signal of the suppression signal and the motion control signal.

[0048] Understandably, see Figure 1 The electro-hydraulic proportional valve and hydraulic cylinder can drive the working platform to move. These valves and cylinders are driven based on a PWM driver instruction synthesis unit. A linear superposition signal of the suppression signal and motion control signal can be processed by the VECU controller and sent to the PWM driver instruction synthesis unit. The PWM driver instruction synthesis unit executes this linear superposition signal to control the electro-hydraulic proportional valve and hydraulic cylinder, thereby suppressing the vibration of the working platform while driving its movement. Specifically, the motion control signal can come from... Figure 1 The operator handle shown is sent to the VECU controller; in some embodiments, the aerial work platform can also achieve automated movement, and the motion control signal can also be generated by the built-in program of the aerial work platform.

[0049] See Figure 4 In some embodiments of this application, the vibration suppression method for aerial work machinery may further include: S402. Determine the total activation energy of the impact event atoms activated by the sparse activation vector in the physical dictionary of the haptic event; S404. When the total activation energy exceeds a preset energy threshold, reduce the sensitivity of the work platform to motion control signals.

[0050] During an emergency stop of an aerial work platform, the Z-axis acceleration curve of the platform under existing technology indicates that the platform experienced a severe impact and subsequent oscillation of approximately 0.4g. However, when the vibration suppression method for aerial work platforms provided in this application is applied, the maximum impact is limited to less than 0.25g due to the reduction of the platform's response sensitivity to motion control signals, i.e., the intervention of the impact suppression strategy. Furthermore, the subsequent oscillation is rapidly absorbed, resulting in a significant improvement in the perceived consistency of the user experience.

[0051] In some embodiments of this application, determining the total activation energy of the impact event atoms activated by the sparse activation vector in the physical dictionary of haptic events includes: determining the activation coefficients corresponding to the activated impact event atoms in the sparse activation vector; and determining the sum of squares of each activation coefficient to obtain the total activation energy.

[0052] In some embodiments of this application, when the total activation energy is greater than a preset energy threshold, reducing the response sensitivity of the work platform to the motion control signal includes: acquiring the motion control signal; when the total activation energy is greater than the preset energy threshold, reducing the slope of the motion control signal in the time domain according to a preset inhibition factor to obtain a weakened motion control signal; and sending the weakened motion control signal to the motion control unit of the work platform so that the motion control unit controls the movement of the work platform according to the weakened motion control signal.

[0053] In some embodiments of this application, when the total activation energy is greater than a preset energy threshold, reducing the slope of the motion control signal in the time domain according to a preset inhibition factor to obtain a weakened motion control signal includes: determining one or more sub-motion control signals of the motion control signal, wherein the sub-motion control signal represents a control signal for controlling the movement of the work platform in a specific degree of freedom; determining the degree of freedom of the impact event corresponding to the impact event; determining a target sub-motion control signal among the multiple sub-motion control signals according to the degree of freedom of the impact event, and reducing the slope of the target sub-motion control signal in the time domain according to the preset inhibition factor to obtain a weakened target sub-motion control signal; and sending the weakened motion control signal to the motion control unit of the work platform so that the motion control unit controls the movement of the work platform according to the weakened motion control signal includes: sending the weakened target sub-motion control signal to the motion control unit of the work platform so that the motion control unit controls the movement of the work platform according to the weakened target sub-motion control signal.

[0054] In some embodiments of this application, the inertial measurement unit determines the low-dimensional motion signal based on the following steps: monitoring the motion behavior of the working platform to obtain the original motion signal; processing the original motion signal by removing the mean and normalizing the variance to obtain the motion signal; compressing the motion signal according to the adaptive observation matrix to obtain the low-dimensional motion signal.

[0055] The following example illustrates the execution process of the vibration suppression method for aerial work machinery provided in the embodiments of this application: The method provided in this embodiment of the invention is executed within a continuous loop control cycle (set to 15 milliseconds in this embodiment), and its core algorithm flow is as follows: Figure 2 As shown, it includes three main steps: perception and analysis, attribution and focusing, and prediction and intervention. The perception and analysis step will be explained in detail below.

[0056] The overall goal of the Perception & Parsing step is to transform complex, high-dimensional raw dynamic signals on the working platform into structured, low-dimensional "sensory event" descriptions with clear physical meaning in real time, with minimal computational and communication overhead. It fundamentally resolves the core contradiction between high-fidelity perception and the limited computing power of embedded controllers, and is a technological prerequisite for achieving real-time active control. This step is further broken down into the following four closely linked sub-steps: Sub-step 201.1: Raw dynamic signal acquisition and preprocessing; Execution subject: Inertial sensing unit (IMU) 101. Technical motivation: To ensure the accuracy and stability of subsequent analysis, it is necessary to obtain a standardized data sample that can fully reflect the dynamic characteristics of the platform within a short time window.

[0057] Specific operations: High-frequency sampling: The sensing module 101 continuously collects raw data of its six degrees of freedom (three-axis acceleration, three-axis angular velocity) at a frequency of not less than 1kHz. Windowing processing: The continuous data stream is divided into segments with a 50% overlap and a fixed length. (For example A time window of (number of sampling points) is used. Overlapping settings ensure that dynamic events occurring at the boundaries are not missed due to window segmentation. Sample generation: At the beginning of each control cycle, a high-dimensional original dynamic signal vector is formed. In this embodiment, since it is six-axis data, the actual vector dimension processed is [dimension missing]. For the sake of simplicity, we will use a single axis as an example here. In actual applications, multi-axis parallel processing is used.

[0058] Local normalization: On the local microprocessor of the sensing module 101, the generated... Mean and variance normalization are performed. This step is crucial because it aims to eliminate signal DC bias and amplitude differences caused by different operating conditions (such as different boom extension lengths). This allows subsequent sparse parsing algorithms to focus more on the dynamic shape of the signal rather than its absolute size, thereby greatly improving the generalization ability and robustness of the dictionary.

[0059] Sub-step 201.2: Front-end adaptive compression. Execution entity: Built-in microprocessor / DSP of the sensing module (IMU). Technical motivation: To solve the data transmission bandwidth bottleneck and reduce the computational burden on the main controller. Direct transmission of high-dimensional raw signals. It will severely consume CAN bus bandwidth and impose a huge computational burden on the main controller.

[0060] Specific operation: Observation matrix acquisition: Sensing module 101 uses the adaptive observation matrix stored at the current time. (in ,For example It needs to be emphasized that here... It is updated in step S202 of the previous control cycle and (if necessary) sent here. This reflects the "adaptive" closed-loop characteristic of the entire system.

[0061] Compression transformation: Performs one matrix-vector multiplication operation, i.e. This operation transforms the high-dimensional, normalized original signal... (dimension) Linear projection onto a low-dimensional vector of measurements (dimension) The innovation of this step lies in the fact that it is not a simple random projection, but an intelligent and targeted observation. Because... Having already incorporated the previous focus strategy on "bad modes," this compression will naturally enhance the sampling of the dynamic features that the system is currently most concerned with (such as specific swing frequencies), thereby improving the signal-to-noise ratio of key information at the source.

[0062] Sub-step 201.3: Low-dimensional measurement value transmission. Execution entity: Communication link between the sensing module and the main controller module 103. Technical motivation: To achieve efficient, low-latency, and low-bandwidth information exchange between the sensing end and the control end.

[0063] Specific steps: Transform the measurement value vector, whose dimensionality has been significantly reduced... The data is packaged into one or more CAN messages and sent to the main controller module (VECU) via the vehicle's standard CAN bus.

[0064] Quantitative analysis: Taking 16-bit data as an example, transmitting the original signal need bytes, while transmitting compressed data Only The data transmission volume is reduced by 75%. This significant advantage allows the invention to be easily integrated into existing vehicle bus architectures without causing network congestion, ensuring the real-time performance and stability of the vehicle control system.

[0065] Sub-step 201.4: High-speed sparse reconstruction and parsing. Execution subject: Main controller module (VECU).

[0066] Technical motivation: To obtain received, linearly "encrypted" low-dimensional measurements It can accurately and quickly "decode" information into structured information with clear physical meaning.

[0067] Specific steps: Core problem definition: The mathematical essence of this step is solving a classic sparse recovery problem: Given... , and physics dictionary Find the sparsest activation vector. , making Solution Engine: This embodiment employs a sparse reconstruction algorithm based on deep neural network expansion, specifically PC-OMPNet v2. This network expands the solution process of the traditional, iterative Orthogonal Matching Pursuit (OMP) algorithm into a feedforward neural network of fixed depth. This approach transforms the complex iterative optimization problem into a series of efficient matrix multiplications and nonlinear activations, significantly improving computational speed.

[0068] Input preparation: To further improve efficiency, the main controller 103 receives... At the same time, utilize known and The dynamic response matrix is ​​calculated in advance. .

[0069] Execution parsing: and As input, the data is fed into the PC-OMPNet v2 network for a forward propagation. On the target VECU hardware, this process can be completed within 12 milliseconds.

[0070] Generate dual-core output: A) Sparse activation vector This is the most important output of this step. It is a highly sparse vector, where the positions of its non-zero elements precisely indicate which fundamental physical events from the dictionary constitute the current platform dynamics, while the values ​​of the non-zero elements quantify the intensity of these events. This vector serves as the direct input and decision-making basis for subsequent steps (new anomaly identification) and (predictive intervention). B) Pure Reconstructed Signal By parsing the With physics dictionary Multiplying these results in a clean time-domain signal that has been freed from random noise and contains only physically interpretable events. This signal has an extremely high signal-to-noise ratio, making it a perfect data source for accurate spectral analysis in step S202 to identify known swing patterns.

[0071] Through the precise collaboration of the above four sub-steps, the perception and analysis steps successfully provide high-quality, structured, and easy-to-understand decision inputs for the entire intelligent control system at the initial stage of each control cycle, which is a solid foundation and prerequisite for realizing all the advanced functions of this invention.

[0072] The initial "sensory event report" obtained from the perception and analysis steps undergoes in-depth processing and intelligent judgment. It not only attributes the current platform dynamics to their primary source, determining whether they are known risks or newly emerging situations, but also focuses on the system's perception resources to generate a forward-looking observation strategy to guide the perception process in the next control cycle. This step is a crucial bridge connecting "passive perception" and "active intervention," and is a concentrated manifestation of the system's adaptive capabilities. This step is implemented through a parallel dual-closed-loop mechanism: Closed Loop 1: Anticipatory Inconsistency Loop. Execution Entity: Main Controller Module (VECU). Technical Motivation: Among all dynamic responses of the boom crane, the low-frequency, high-amplitude swaying caused by the boom's flexible structure poses the greatest perceived hazard to operators and is the vibration behavior most in need of precise suppression. Input Data: Pure reconstructed signal. The reason for choosing this signal is that it has filtered out random noise, has an extremely high signal-to-noise ratio, and retains complete time-domain phase information, making it an ideal data source for accurate spectral analysis to capture periodic signals.

[0073] Specific implementation steps: A) Fast Fourier Transform (FFT): For Performing an FFT operation once yields its representation in the frequency domain. B) Key frequency matching: A pre-built library of known inconsistency patterns is used within the system. This library stores a set of frequencies characteristic of the most typical "bad modes" of this high-performance processor (e.g., first-order swing frequency 0.8Hz, second-order swing frequency 2.1Hz, etc.). The system iterates through this library... Find the matching frequency point with the strongest energy. ,Right now: C) Confidence Assessment: To quantify the significance of the current "bad pattern," its inconsistency confidence level is calculated. In this embodiment, Defined as the ratio of the energy at the matched frequency point to the average energy of the spectrum: The larger this value, the more prominent the swing pattern, and the higher its priority for attention. D) Target enhancement vector generation: If Exceeding a preset activation threshold If (e.g., 1.5), the system determines that this frequency needs to be observed in detail. At this point, the system will generate a target enhancement vector. This vector is a frequency domain representation of... A window function centered at a specific bandwidth (e.g., the Hanning window), and its corresponding waveform in the time domain. The physical meaning of this vector is to construct a "frequency domain searchlight" to enhance the observation sensitivity to this specific swing frequency at the next time step. If the threshold is not exceeded, then... It is a zero vector.

[0074] Loop Two: Novel Anomaly Loop. Execution Entity: Main Controller Module (VECU). Technical Motivation: The actual operating conditions of high-performance machines are complex and variable, potentially leading to unknown vibration or shock modes generated by multi-factor nonlinear coupling that were not covered in the simulation phase. This loop is designed to act like an alert "general practitioner in emergency medicine," keenly capturing these new and unconventional combinations of dynamic events and initiating exploratory observation mechanisms. Input Data: Sparse activation vectors of high-dimensional "sensory events". The reason for choosing this signal is that it directly describes the current dynamics as being composed of "which underlying events and at what intensity". The essence of a "new anomaly" is a rare "event combination pattern".

[0075] Specific implementation steps: A) Dynamic pattern fingerprint generation: To efficiently represent and query complex event combination patterns, the system employs the Locality Sensitive Hashing (SimHash) algorithm. For sparse vectors... Perform a SimHash operation to generate a low-dimensional (e.g., 64-bit) binary hash fingerprint. This fingerprint can map similar event combination patterns to similar hash codes with a high probability. B) Historical Dynamic Pattern Memory (HAPM) Query: HAPM is a database that stores all recently occurring dynamic pattern fingerprints. The hash table. The system will use the currently generated hash table. C) Novelty assessment: If the current fingerprint is compared with the records in HAPM using Hamming distance. If the Hamming distance to all records in HAPM is greater than a preset "novelty" threshold (e.g., 3 bits), the system determines that the current occurrence is a novel non-rhythmic dynamic (is_novel_dynamic=True). Simultaneously, this new fingerprint is... Stored in HAPM. HAPM also includes a time decay mechanism, which periodically reduces the weight of old fingerprints or removes them to ensure that only "recent" dynamic patterns are remembered, adapting to changes in operating conditions. D) Exploration Vector Generation: Once a new anomaly is confirmed, due to its unknown characteristics, the system adopts the most conservative comprehensive observation strategy. At this time, an exploration vector is generated. Its value is equal to the best time-domain estimate of the current abnormal dynamics, that is Using this vector to guide the update of the observation matrix has the effect of adjusting the overall observation method in the next time step in a direction most conducive to reproducing the current new anomaly. If it is not a new anomaly, then... It is a zero vector.

[0076] Collaborative enhancement and final decision-making. Execution entity: Main controller module (VECU). Technical motivation: To unify the outputs of the two closed loops, forming a single, collaborative, and optimal observation matrix update strategy.

[0077] Specific implementation steps: A) Weighted arbitration: The system needs to weigh the recommendations of "specialist diagnosis" and "general practitioner screening". This is done through a preset weighting function, based on... The value, novelty judgment result and current The overall energy is given by two enhancement vectors. and Dynamic weight allocation and B) Final update direction generation: The two weighted vectors are linearly combined to obtain the final collaborative enhancement vector. This vector combines a deep focus on known risks with a broad exploration of unknown anomalies. C) Observation matrix update: utilizing To calculate the update step size of the observation matrix In this embodiment, a gradient-ascent-like strategy is employed to ensure that the updated observation matrix... The projected energy in the direction is maximized. Finally, the updated matrix is ​​updated using the Gram-Schmidt orthogonalization process. The data is processed to obtain the final observation matrix for the next time step. The orthogonalization step ensures that the rows of the updated observation matrix maintain low correlation, avoiding information redundancy and the collapse of observation capabilities.

[0078] Through the above rigorous dual-closed-loop collaborative mechanism, we not only gained a deep understanding of the inherent attributes of current somatosensory events, but also dynamically reshaped the system's "attention" with forward-looking intelligent decision-making, paving the way for the next round of accurate perception and ultimately effective intervention.

[0079] The structured "hyper-sensory inconsistency" information identified in the preceding steps is transformed into real-time, physical, predictive compensation and proactive intervention commands for the main control system of the aerial work platform. This allows the system to proactively suppress or eliminate inconsistencies before they are fully perceived by the operator, rather than passively responding to worsening sensations. This step is the fundamental difference between the vibration suppression method for aerial work platforms provided in this application and all existing technologies, and it is the core execution link for achieving "software-defined sensory perception." This step mainly consists of the following two parallel intervention strategies: Strategy 1: Active Damping Strategy. Executor: Main Controller Module (VECU). Technical Motivation: Specifically designed to address the periodic low-frequency swaying caused by boom flexibility, which has the greatest impact on user comfort and operational accuracy. Traditional passive damping (structural damping) has limited effectiveness, while error-based feedback control suffers from time lag. This strategy employs a feedforward compensation approach, actively controlling the main hydraulic cylinder to generate a small additional force opposite to the swaying motion, thereby achieving efficient dissipation of swaying energy.

[0080] Triggering logic: When the previously known inconsistency prediction loop confirms the existence of significant periodic fluctuations, i.e., the inconsistency confidence level... Exceeding the preset damping start-up threshold (e.g., 2.0), this policy is activated.

[0081] Input data: Pure reconstructed signal It provides all the information needed for precise compensation—the real-time amplitude, frequency, and phase of the swing. The matched swing frequency... Used for parameterized compensation algorithms.

[0082] Intervention command generation algorithm (Calculate_Damping_Signal): A) State observer: Using a Kalman filter or similar state observer, the input signal is... The process is performed to estimate the displacement of the sway in the boom coordinate system in real time. and speed B) Compensation force calculation: The core principle of compensation is to apply a damping force in the opposite direction to the swaying velocity, i.e. ,in It is a calibrable damping gain coefficient. C) Inverse command solution: Through the inverse Jacobian matrix of the high-performance machine, the required compensation force is obtained. The reverse analysis traces back to the main hydraulic actuator (such as a luffing cylinder or telescopic cylinder) that requires the force, calculating the additional acceleration or speed command that the cylinder needs to generate. D) Valve control signal conversion: Finally, through the transfer function model of the hydraulic system, this additional acceleration / speed command is converted into a small, high-frequency modulated additional current command for the corresponding electro-hydraulic proportional valve. .

[0083] Integration with lower-level control: The motion control unit will send this additional current command. This is linearly superimposed on the base valve-controlled current from the operator's main command. The final effect: while performing the macroscopic main motion, the main hydraulic cylinder also performs minute but precise high-frequency reciprocating motions. The force generated by this additional motion precisely counteracts the swaying energy, making the boom appear from the outside as if it's being "held down" by an invisible hand, with swaying quickly suppressed. Because the compensation is feedforward and phase-precise, the operator will only feel an unusually stable platform and will hardly perceive the compensation action itself.

[0084] Strategy Two: Jerk Suppression Strategy. Executor: Main Controller Module (VECU). Technical Motivation: Specifically designed to address mechanical and hydraulic shocks caused by discontinuous or abrupt changes in operating commands (such as a handle being pushed from zero to full or back to zero instantaneously). The essence of this shock is excessive acceleration (jerk). This strategy employs a command-shaping approach, smoothing out the "brutal" original commands at the source, making them "gentle," thereby preventing or greatly reducing the impact.

[0085] Triggering logic: When the parsed sparse activation vector In the process, the sum of the activation energies of the physical atoms that were labeled as "impact events" during the offline training phase (i.e., Exceeding a preset impact warning threshold This strategy is activated at that time. Predictability is reflected in the fact that this judgment is instantaneous. Because the dictionary atoms capture the "morphological characteristics" of the impact, the system can predict an impending impact event based solely on the initial trend of signal changes, before the acceleration curve reaches its peak.

[0086] Input data: sparse activation vectors The original command signal from the operator handle.

[0087] Intervention command generation algorithm (command training): A) Command channel identification: The system first identifies which degree of freedom operation command (such as amplitude change, slewing, etc.) is currently triggering the impact warning.

[0088] B) Dynamic Slope Limit Adjustment: The main controller typically has a slope limiter (SlewRateLimiter) for each command channel to limit the maximum rate of change of the command value. Normally, this value is relatively large to ensure response sensitivity. Once an impact warning is triggered, this strategy dynamically and instantaneously lowers the slope limit parameter command_slew_rate for the corresponding channel (e.g., temporarily multiplying it by a suppression factor of 0.1 to 0.5). C) "S-shaped" Curve Generation: After processing by the slope limiter, the original command, which was originally a step or steep change, will have its rising and falling edges forcibly smoothed into a continuous "S-shaped" acceleration curve. D) Automatic Recovery Mechanism: When subsequent... When no further impact events are detected, the slope limit parameter automatically and smoothly returns to its normal value to ensure that the sensitivity of operation is not unnecessarily affected. Integration with lower-level control: This strategy intervenes at the highest-level command processing stage. The motion control unit receives a smoothed command curve that has already been "trained".

[0089] Final Result: Because the drive commands themselves have become very smooth, the opening and closing process of the electro-hydraulic proportional valve also becomes correspondingly gentle, avoiding sudden pressure changes within the hydraulic system and rigid impacts on the mechanical structure. The operator's subjective experience is that even with a very "aggressive" operating style, the equipment's response always maintains a "flexible" rather than "rigid" feel, with shocks and jerks greatly reduced. Through the parallel implementation and synergistic effect of these two predictive intervention strategies, the system's intelligent analysis capabilities are successfully transformed into tangible, beneficial, and proactive changes to the physical world, ultimately achieving the core objective of this invention: improving the consistency of machine feel.

[0090] See Figure 5 In some embodiments of this application, the vibration suppression method for aerial work machinery further includes the step of obtaining a physical dictionary of somatosensory events, which is obtained based on the following steps: S502. Acquire multiple training data segments, which represent the motion signals of the working platform under specific motion behaviors; S504. Based on multiple training data segments, the final dictionary and the sparsest activation vector of each training data segment are obtained iteratively through the K-SVD algorithm. S506. For each atom in the final dictionary, determine the training data segment with the maximum activation sparsity for the atom based on the sparsest activation vector of each training data segment, and obtain the activation data segment of the atom. S508. If the proportion of data segments representing the same motion behavior in the active data segments of an atom is greater than a preset proportion threshold, then the atom is determined to be a valid atom representing a specific motion behavior. S510. Retain all valid atoms in the final dictionary to obtain the physical dictionary of haptic events.

[0091] The motion-sensing event physics dictionary is an incomplete base that autonomously learns and extracts data from massive, high-fidelity, high-performance dynamic data using machine learning methods. It is capable of representing various physical dynamic events. The training process for the motion-sensing event physics dictionary is completed offline during the product development phase. It assumes any complex and chaotic platform dynamic signal... (That is, the operator's actual physical sensations) can all be precisely approximated as a dictionary. A few "fundamental physical events" in the atom sparse linear combinations, i.e. ,in It is a small set of indexes. Therefore, by training the dictionary offline, this optimal and most representative set of "physical atoms" can be found, allowing them to be combined in the sparsest and most robust way to form various possible platform dynamics. The resulting physical dictionary of haptic events is as follows: Figure 6 As shown.

[0092] In some embodiments of this application, step S504 may include: for each training data segment, determining the sparsest activation vector of the current dictionary using a sparse coding algorithm, wherein the sparsest activation vector minimizes the reconstruction error of the training data segment and the number of non-zero terms in the sparsest activation vector is less than a preset sparsity; updating each atom in the current dictionary according to the sparsest activation vector of each training data segment to obtain an updated dictionary, and re-determining the sparsest activation vector of each training data segment based on the updated dictionary; if the updated dictionary meets a preset termination condition, determining the updated dictionary as the final dictionary, and obtaining the sparsest activation vector of each training data segment determined based on the final dictionary.

[0093] In some embodiments of this application, obtaining multiple training data segments includes: obtaining a training database, wherein the training database includes multiple training motion signals, the training motion signals representing the motion signals of the working platform under specific motion behaviors; for each training motion signal, dividing the training motion signal into multiple original training data segments with overlapping signal segments in time according to a preset signal length; for each original training data segment, processing the original training data segment by removing the mean and normalizing the variance to obtain the training data segment.

[0094] In some embodiments of this application, obtaining a training database may include: constructing a dynamic model and an electro-hydraulic control model of the aerial work platform; performing joint simulation based on a joint simulation model of the dynamic model and the electro-hydraulic control model, wherein the joint simulation model includes a virtual working platform of the aerial work platform and a virtual inertial measurement unit on the virtual working platform; during the joint simulation, recording motion signals obtained by the virtual inertial measurement unit monitoring the motion behavior of the working platform under different working conditions as training motion signals; and saving multiple training motion signals as a training database.

[0095] This embodiment uses high-fidelity multibody dynamics as the dynamic model of the aerial work platform and co-simulates it with the hydraulic system to generate training data; to ensure the completeness of the dictionary (covering all working conditions) and physical interpretability.

[0096] As an example, the simulation platform is built as follows: Using Adams or similar multibody dynamics software, an accurate 3D rigid-flexible coupled dynamic model of the high-performance machine is established. This model needs to include the flexible characteristics of the boom (imported through finite element analysis), joint clearances and friction, tire nonlinear characteristics, etc. Using Amesim or Simulink / Simscape, a detailed electro-hydraulic control system model matching the actual vehicle is established, including the dynamic response of the hydraulic pump and proportional valve, the physical properties of the hydraulic oil, etc. The two are then co-simulated to construct a digital twin that can accurately reproduce the dynamic response of the real vehicle.

[0097] 2. Training Data Generation (Scene Coverage): Using this digital twin, systematically run and collect operational data covering the entire lifecycle of the high-performance machine. The collected signals are six-axis data from the platform's central IMU. Scenario design must include: A) Basic Independent Motion: For each degree of freedom (start, amplitude change, extension, rotation, etc.), execute the entire process from rest to maximum speed and then to stop. The aim is to learn the most basic acceleration, constant speed, and deceleration atoms. B) Composite Linkage Operations: Execute commonly used operator combinations, such as "extending the arm while rotating" or "walking while raising the arm." The aim is to learn the complex dynamic atoms generated when multiple degrees of freedom motions are coupled. C) Limit and Boundary Conditions: Execute operations such as emergency braking (E-Stop), high-speed impact on obstacles (virtual), and rapid crossing of uneven ground under extreme conditions such as maximum amplitude, maximum arm length, and maximum load. The aim is to learn key atoms representing "bad physical sensations" such as impact, violent swaying, and instability. D) Active Injection of Specific Events: To make the physical meaning of the dictionary purer, actively inject known, single excitation signals into the simulation. For example, when the system is stationary, a precise half-sine wave impact force is applied only to the Z-axis to collect pure "Z-axis impact" data samples; or a lateral excitation of a specific frequency is applied to the arm tip to collect pure "first-order / second-order swing" data samples.

[0098] Data Output: After completing the above simulation, a large raw time series database with scene labels is obtained. Each data file is clearly labeled with its source (e.g., "E-stop from full slew at max reach").

[0099] Data preprocessing. Segmentation and slicing: The long time-series data collected is segmented into fixed lengths (e.g., ...). Overlapping signal segments (sampling points) Each signal segment inherits the scene label from its source. Normalization: for each signal segment... The signal undergoes mean and variance normalization. This eliminates the dimensional influence of different operating amplitudes, allowing the learning algorithm to focus more on the signal's shape rather than its absolute amplitude.

[0100] Core dictionary learning algorithm. This embodiment uses the K-SVD algorithm to perform dictionary learning. K-SVD is a highly efficient dictionary learning algorithm that iteratively alternates between sparse encoding and dictionary updates, eventually converging to obtain an optimal dictionary. Initialization: Randomly select from the preprocessed data segment. One (e.g.) (to form an overcomplete dictionary) as the initial dictionary Iterative process: A) Sparse coding stage: Fix the current dictionary For each training data segment Using sparse coding algorithms such as Orthogonal Matching Pursuit (OMP), a minimum sparse activation vector is found. , making Minimum, and The number of non-zero terms is less than the preset sparsity. B) Dictionary update phase: Fix all sparse activation vectors Update the dictionary column by column. For the first column of the dictionary... Column (atoms) Find the set of all data segments that used that atom. Then, the error matrix is ​​calculated. and to In the index set Singular Value Decomposition (SVD) is performed on the constraints above, and the results are updated using its principal left singular vector. 3. Termination condition: Training terminates when the change in the dictionary is less than a very small threshold, or when the preset maximum number of iterations is reached.

[0101] Physical interpretability verification and annotation of the dictionary. The K-SVD algorithm itself is purely mathematical, and it does not guarantee that the learned atoms have clear physical meanings. Therefore, after training, physical meaning must be labeled and verified. This is a key step that distinguishes this invention from traditional signal processing. Reverse mapping and clustering: For the trained final dictionary... Each atom in We iterate through the entire training dataset to find those elements in its sparse representation that are... It has the maximum activation coefficient ( The largest data segment. Tag statistical analysis: Statistically analyze the scene tags carried by these strongly activated data segments. Clear physical meaning: If atoms are found... Since an atom is almost exclusively activated by data segments originating from the "E-Stop" and "impact injection" scenarios, we can confidently label it as an "impact event atom." Similarly, we can label it as a "swaying atom," "acceleration atom," etc. Meaningless mixed atoms: If an atom is activated by various unrelated scenario data segments, it indicates that it lacks a clear physical meaning and is a "mixed noise" atom. Dictionary pruning and optimization: All "mixed noise" atoms that cannot be assigned a clear physical meaning are removed from the dictionary. The remaining atoms collectively constitute the final, pure, and physically interpretable "body-sensing event physics dictionary." Through the rigorous process described above, we obtain the final dictionary. It is no longer a black box mathematical matrix, but a structured knowledge base containing advanced computer science expert knowledge and dynamic characteristics, providing a solid foundation for subsequent real-time analysis and precise intervention.

[0102] The vibration suppression method for aerial work machinery provided in this application aims to solve the technical problem in the prior art that the consistency of the aerial work platform's body sensation depends on passive design and cannot be evaluated and actively controlled in real time. Its core idea is to regard the dynamic response signal of the aerial work platform as sparse under a certain "physical dictionary of body sensation events", to conduct efficient observation through adaptive compressed sensing technology, to perform real-time analysis through a high-speed sparse reconstruction algorithm, and finally to transform the analysis result into an active intervention command for the system. The vibration suppression method for aerial work platforms provided in this application represents a leap from "passive evaluation" to "active control": For the first time, it treats motion consistency as a target for real-time closed-loop control, significantly improving the smoothness and stability of the aerial work platform's operation through active damping and shock suppression; it greatly enhances sensing efficiency and real-time performance: by compressing sensing, it significantly reduces the transmission bandwidth of sensor data and the computational load on the controller, making it possible to implement complex algorithms on cost- and computing-sensitive embedded controllers. The entire "sensing-analysis-intervention" closed loop can be completed within 15 milliseconds; it enhances the system's adaptability and intelligence: through adaptively adjusted observation matrices and the ability to identify emerging anomalies, the system can adapt to dynamic changes under different working conditions, maintaining optimal control performance; and it achieves "software-defined motion sensing": providing a core technical framework for future optimization and customization of the aerial work platform's motion sensing performance through OTA (Over-The-Air) software upgrades, greatly increasing the product's added value and lifecycle.

[0103] It should be understood that although the steps in the flowcharts provided in the embodiments of this application are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts provided in the embodiments of this application may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0104] This application also provides a vibration suppression device for aerial work machinery. The vibration suppression device for aerial work machinery includes a processor configured to execute the vibration suppression method for aerial work machinery provided according to the embodiments of this application.

[0105] This application also provides an aerial work platform, which includes: a work platform equipped with an inertial measurement unit; a vibration suppression device for the aerial work platform provided in the above embodiments of this application; and a motion control unit for controlling the movement of the work platform and for receiving suppression signals from the vibration suppression device to suppress the swaying of the work platform according to the suppression signals.

[0106] Specifically, the inertial measurement unit in this embodiment is, for example, an industrial-grade, high-precision, six-axis inertial measurement unit, comprising: a three-axis accelerometer: used to accurately capture the linear motion of the platform in a spatial rectangular coordinate system, especially the impact and jerking sensation in the vertical (Z-axis) direction, and the smoothness of acceleration and deceleration in the horizontal direction. A three-axis gyroscope: used to accurately capture the angular velocity of the platform around three axes, which is crucial for identifying and quantifying low-frequency swaying and torsional vibrations caused by the flexible deformation of the boom. To ensure the data fidelity required by the algorithm, the performance indicators of the inertial measurement unit must meet the following requirements: a sampling frequency of not less than 1kHz to capture high-frequency impact details; low gyroscope zero-bias instability to avoid long-term integration errors; and low accelerometer noise density to provide a clean raw signal. Internally, it typically integrates a microelectromechanical system (MEMS) sensor and a small microcontroller (MCU) or digital signal processor (DSP) to provide local computing capabilities.

[0107] In some embodiments of this application, the inertial measurement unit is securely mounted at the center of the work platform floor. The choice of this location is crucial because it is where the operator is located, and the dynamic response at this location most directly reflects the operator's "body sensation." Furthermore, the central location minimizes measurement noise introduced by local floor vibrations and better represents the overall rigid body motion and structural vibration modes of the entire platform.

[0108] In this embodiment, the inertial measurement unit serves as an entry point into the physical world, continuously acquiring the platform's high-frequency dynamic response and generating a high-dimensional original signal vector. It utilizes its built-in MCU / DSP to perform adaptive compressed sensing. It receives the updated adaptive observation matrix from the VECU. and the collected Perform real-time matrix multiplication to obtain low-dimensional measurements. This front-end processing reduces data dimensionality by more than 75%. The inertial measurement unit can transmit the compressed, high-input-entropy, low-dimensional measurement values ​​via the vehicle's Controller Area Network (CAN) bus. It periodically sends data to the VECU with extremely low bandwidth utilization and can selectively receive instructions from the main controller module to update the observation matrix stored internally. .

[0109] The VECU in this embodiment can be, for example, Danfoss's PLUS+1 series or IFM's eComatController series. This VECU needs to possess: a powerful computing core, such as a multi-core processor based on the ARM Cortex-A architecture, to ensure that all complex calculations can be completed within each control cycle (<15ms); and sufficient memory, including RAM for program execution and memory for storing a "physical dictionary of haptic events". and non-volatile memory (such as Flash) of the History Dynamic Pattern Memory (HAPM); The VECU internally contains complete software logic responsible for executing the entire process of perception and analysis, attribution and focusing, prediction and intervention: it can perform high-speed analysis and receive... Running PC-OMPNet v2 to reconstruct the network instantly resolves the core sparse activation vectors. and pure signal Intelligent attribution: Executes a dual closed-loop mechanism to determine the root cause of the current sensory event (whether it is a known sway or a new anomaly); Strategy focus: Calculates and updates the observation matrix for the next time step based on the attribution results. This enables dynamic focusing of perceived resources; command fusion and generation: receiving and processing raw intent commands from the operator's handle. Simultaneously, based on... and The system performs parallel calculations of the compensation commands required for active damping and the slope adjustments required for shock suppression. Finally, all information is fused to generate the final high-level commands for the motion control unit; interaction with other modules includes receiving compression measurements from the sensing module 101. ; Receive raw control commands from the operator handle; Output fused and compensated high-level logic commands (e.g., target velocity, target acceleration, slope limit parameters, etc.) to the motion control unit.

[0110] The core function of the motion control unit in this application embodiment is instruction translation and synthesis. It can accurately convert the abstract logic instructions of the VECU into physical electrical signals (such as PWM current with a specific voltage or duty cycle) that can directly drive the electro-hydraulic proportional valves in the main hydraulic system. Its key functions are: Main instruction execution: converting the operator's main instructions, "trained" by the shock suppression strategy, into basic valve control current. Compensation instruction superposition: accurately superimposing the high-frequency, minute compensation instruction current generated by the active damping strategy with the basic valve control current. Final drive: outputting the synthesized final drive current to the solenoid coil of the hydraulic valve. Interaction with other modules: receiving all high-level control instructions from the VECU. Sending precise, physical drive electrical signals to each proportional valve in the main hydraulic system. In some embodiments of this application, the VECU includes a main controller module and a motion control unit. The main controller module is used to execute the vibration suppression method for aerial work machinery provided in this application embodiment, and the motion control unit is the high-power output port on the VECU and its drive circuit, such as a pulse width modulation (PWM) driver.

[0111] The aerial work platform machinery may also include: a main hydraulic system and an operator's handle: acting as the "muscle system" of the machine, this is the ultimate physical manifestation of the control effect of this invention. The output of this invention ultimately acts on its internal electro-hydraulic proportional valve, changing the flow rate of the hydraulic cylinder by finely adjusting the valve opening, thereby achieving precise regulation of the boom movement; as the input source of the operator's "intention," it provides the desired direction and speed of the system's movement. This invention is not intended to replace the operator, but rather to act as an "intelligent optimizer" and "protective layer" for their intentions, faithfully executing their intentions while actively eliminating components that may cause physical discomfort.

[0112] This application provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform a vibration suppression method for aerial work machinery according to an embodiment of this application.

[0113] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05, and a memory (not shown) connected via a system bus. The processor A01 provides computing and control capabilities. The memory includes internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 stored in the non-volatile storage medium A06. The network interface A02 is used for communication with external terminals via a network connection. When the computer program is executed by the processor A01, it implements a vibration suppression method for aerial work machinery. The display screen A04 can be a liquid crystal display (LCD) or an e-ink display. The input device A05 can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0114] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0115] In one embodiment, the vibration suppression device for aerial work machinery provided in this application can be implemented as a computer program, which can be implemented in the form of, for example... Figure 6 The device operates on the computer shown. The computer's memory can store various program modules that make up the vibration suppression device for aerial work machinery. The computer program, composed of these program modules, causes the processor to execute the steps in the vibration suppression methods for aerial work machinery described in the various embodiments of this application.

[0116] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0118] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0119] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0120] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0121] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0122] Computer-readable media include both permanent and non-permanent, removable and non-removable media, which can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0123] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0124] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A vibration suppression method for aerial work machinery, characterized in that, The aerial work platform includes a working platform equipped with an inertial measurement unit (IMU) for acquiring the platform's raw dynamic signals; the vibration suppression method includes: The low-dimensional motion signal of the working platform is acquired, and the low-dimensional motion signal is obtained by the inertial measurement unit compressing the original dynamic signal using the current adaptive observation matrix; Based on the low-dimensional motion signal, the adaptive observation matrix, and the pre-trained somatosensory event physical dictionary, a sparse activation vector is obtained through a sparse reconstruction algorithm, and a reconstruction signal is determined based on the sparse activation vector and the somatosensory event physical dictionary. The somatosensory event physical dictionary is trained by a dictionary learning algorithm based on high-altitude operation machinery simulation data, and its atomic representation is the basic dynamic event pattern. Based on a preset swing frequency library, the target swing frequency in the spectrum corresponding to the reconstructed signal is determined to generate an enhancement vector that reflects the target swing frequency. An inhibition signal is generated based on the motion parameters of the working platform determined by the reconstructed signal, and an exploration vector is generated based on the physical dictionary of the somatosensory event and the sparse activation vector. The adaptive observation matrix is ​​updated based on the enhancement vector and the exploration vector, and the suppression signal is sent to the motion control unit of the working platform so that the motion control unit suppresses the swaying of the working platform based on the suppression signal.

2. The vibration suppression method according to claim 1, characterized in that, The steps of determining the target swing frequency in the spectrum corresponding to the reconstructed signal according to a preset swing frequency library to generate an enhancement vector reflecting the target swing frequency, generating a suppression signal based on the working platform motion parameters determined by the reconstructed signal, and generating an exploration vector based on the somatosensory event physical dictionary and the sparse activation vector include: Based on a preset swing frequency library, the target swing frequency in the spectrum corresponding to the reconstructed signal is determined, and the confidence level of the target swing frequency is determined. If the confidence level is greater than a preset confidence threshold, an enhancement vector reflecting the target swaying frequency is generated, and a suppression signal is generated based on the work platform motion parameters determined by the reconstructed signal. If the sparse activation vector does not exist in the historical dynamic pattern memory, an exploration vector is generated based on the somatosensory event physics dictionary and the sparse activation vector.

3. The vibration suppression method according to claim 2, characterized in that, The step of determining the target swing frequency in the spectrum corresponding to the reconstructed signal according to a preset swing frequency library includes: The reconstructed signal is processed by Fast Fourier Transform to obtain the frequency domain signal of the reconstructed signal; The target swing frequency is determined from the preset swing frequency library as the swing frequency that best matches the frequency domain peak value of the frequency domain signal. The confidence level of the target swing frequency is determined by the ratio of the energy value of the target swing frequency in the frequency domain signal to the average energy of the frequency domain signal.

4. The vibration suppression method according to claim 2, characterized in that, The historical dynamic pattern memory includes hash fingerprints of sparse activation vectors that have appeared historically; the step of generating an exploration vector based on the somatosensory event physical dictionary and the sparse activation vector when the sparse activation vector does not exist in the historical dynamic pattern memory includes: Query the sparse activation vector in the historical dynamic pattern memory; Generate the first hash fingerprint of the sparse activation vector; For each hash fingerprint in the historical dynamic pattern memory, determine the Hamming distance between the hash fingerprint and the first hash fingerprint; If all the Hamming distances are greater than a preset distance threshold, it is determined that the sparse activation vector does not exist in the historical dynamic pattern memory, and an exploration vector is generated based on the somatosensory event physical dictionary and the sparse activation vector.

5. The vibration suppression method according to claim 1, characterized in that, The generation of the enhancement vector reflecting the target's swaying frequency includes: A Hanning window centered on the target swing frequency is generated based on a preset bandwidth as the enhancement vector.

6. The vibration suppression method according to claim 1, characterized in that, The step of updating the adaptive observation matrix based on the enhancement vector and the exploration vector includes: The adaptive observation matrix is ​​updated based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector.

7. The vibration suppression method according to claim 6, characterized in that, The step of updating the adaptive observation matrix based on the update step size obtained by weighted fusion of the enhancement vector and the exploration vector includes: The updated vector is obtained by weighted fusion of the enhancement vector and the exploration vector; The update step size that maximizes the projection energy of the adaptive observation matrix in the direction of the update vector is obtained based on the gradient ascent strategy. The adaptive observation matrix is ​​obtained by processing the sum of the adaptive observation matrix and the update step size using the Gram-Schmidt orthogonalization algorithm.

8. The vibration suppression method according to claim 1, characterized in that, The step of generating a suppression signal based on the motion parameters of the working platform determined according to the reconstructed signal includes: The reconstructed signal is calculated to obtain the motion speed of the working platform; The damping force, which is opposite to the direction of the motion speed, is obtained by multiplying the magnitude of the motion speed and the preset damping gain coefficient. The damping force is used to generate the suppression signal to suppress the swaying of the working platform; Sending the suppression signal to the motion control unit of the work platform so that the motion control unit suppresses the swaying of the work platform based on the suppression signal includes: The suppression signal is sent to the motion control unit of the working platform so that the motion control unit controls the movement of the working platform and suppresses the swaying of the working platform according to the linear superposition signal of the suppression signal and the motion control signal.

9. The vibration suppression method according to claim 1, characterized in that, The vibration suppression method further includes: Determine the sum of activation energies of the impact event atoms activated by the sparse activation vector in the physical dictionary of the somatosensory events; If the total activation energy exceeds a preset energy threshold, the sensitivity of the working platform to motion control signals is reduced.

10. The vibration suppression method according to claim 9, characterized in that, The determination of the total activation energy of the impact event atoms activated by the sparse activation vector in the physical dictionary of the somatosensory events includes: Determine the activation coefficients corresponding to the activated impact event atoms in the sparse activation vector; The sum of squares of each activation coefficient is determined to obtain the total activation energy.

11. The vibration suppression method according to claim 9, characterized in that, When the total activation energy exceeds a preset energy threshold, reducing the sensitivity of the working platform to motion control signals includes: Acquire the motion control signal; When the total activation energy is greater than a preset energy threshold, the slope of the motion control signal in the time domain is reduced according to a preset inhibition factor to obtain a weakened motion control signal. The weakened motion control signal is sent to the motion control unit of the work platform so that the motion control unit controls the movement of the work platform according to the weakened motion control signal.

12. The vibration suppression method according to claim 1, characterized in that, The inertial measurement unit determines the low-dimensional motion signal based on the following steps: The original motion signal is obtained by monitoring the motion behavior of the working platform; The motion signal is obtained by removing the mean and normalizing the variance from the original motion signal. The motion signal is compressed according to the adaptive observation matrix to obtain the low-dimensional motion signal.

13. The vibration suppression method according to claim 1, characterized in that, The physical dictionary of somatosensory events is obtained based on the following steps: Acquire multiple training data segments, wherein the training data segments represent the motion signals of the working platform under specific motion behaviors; Based on the multiple training data segments, the final dictionary and the sparsest activation vector of each training data segment are obtained iteratively through the K-SVD algorithm. For each atom in the final dictionary, the training data segment with the maximum activation sparsity for the atom is determined based on the sparsest activation vector of each training data segment, thus obtaining the activation data segment of the atom. If the proportion of data segments representing the same motion behavior in the active data segments of the atom is greater than a preset proportion threshold, the atom is determined to be a valid atom representing a specific motion behavior. By retaining each valid atom in the final dictionary, the physical dictionary of the haptic event is obtained.

14. The vibration suppression method according to claim 13, characterized in that, The step of iteratively obtaining the final dictionary and the sparsest activation vector for each training data segment using the K-SVD algorithm based on the multiple training data segments includes: For each training data segment, the sparsest activation vector of the current dictionary is determined by a sparse coding algorithm. The sparsest activation vector minimizes the reconstruction error of the training data segment, and the number of non-zero terms in the sparsest activation vector is less than a preset sparsity. Update each atom in the current dictionary according to the sparsest activation vector of each training data segment to obtain the updated dictionary, and re-determine the sparsest activation vector of each training data segment based on the updated dictionary; If the updated dictionary satisfies the preset termination condition, the updated dictionary is determined as the final dictionary, and the sparsest activation vectors of each training data segment determined based on the final dictionary are obtained.

15. The vibration suppression method according to claim 13, characterized in that, The acquisition of multiple training data segments includes: Acquire a training database, wherein the training database includes multiple training motion signals, and the training motion signals represent the motion signals of the working platform under specific motion behaviors; For each of the training motion signals, the training motion signal is divided into multiple original training data segments with overlapping signal segments in time according to a preset signal length; For each of the original training data segments, the original training data segments are processed by removing the mean and normalizing the variance to obtain the training data segments.

16. The vibration suppression method according to claim 15, characterized in that, The acquisition of the training database includes: Construct the dynamic model and electro-hydraulic control model of the aerial work machinery; A joint simulation is performed based on the dynamic model and the electro-hydraulic control model, wherein the joint simulation model includes a virtual working platform of the aerial work machinery and a virtual inertial measurement unit on the virtual working platform; During the joint simulation, the motion signals obtained by the virtual inertial measurement unit monitoring the motion behavior of the working platform under different working conditions are recorded as the training motion signals; Multiple training motion signals are saved as the training database.

17. A vibration suppression device for aerial work machinery, characterized in that, The vibration suppression device for aerial work machinery includes a processor configured to perform the vibration suppression method for aerial work machinery according to any one of claims 1-16.

18. An aerial work platform, characterized in that, The aerial work machinery includes: The working platform is equipped with an inertial measurement unit; Vibration suppression device for aerial work machinery as described in claim 17; A motion control unit is used to control the movement of the work platform and to receive the suppression signal from the vibration suppression device to suppress the swaying of the work platform according to the suppression signal.

19. A machine-readable storage medium storing instructions thereon, characterized in that, When executed by a processor, the instruction causes the processor to be configured to perform the vibration suppression method for aerial work machinery according to any one of claims 1 to 16.

Images