Multi-sensor based robot balancing auto-adjustment method and system

Abstract

This invention relates to the field of robot control technology and discloses a method and system for automatic robot balance adjustment based on multiple sensors. The method includes: acquiring multi-source synchronous vibration signals and ground condition feedback data, and preprocessing them to obtain a preliminary vibration signal sequence; performing frequency domain analysis on the preliminary vibration signal sequence to obtain resonant frequency distribution characteristics; determining whether mode shift exists based on the resonant frequency distribution characteristics and obtaining a mode shift index; fusing the mode shift index with the ground condition feedback data to update the vibration model and obtain optimized vibration model parameters. If the optimized vibration model parameters exceed the stable range, the control gain is calculated based on the deviation from the target stable state, and a compensation signal and a corrected balance strategy are generated. This invention, through multi-source data fusion and dynamic model optimization, can significantly improve the balance stability and operational reliability of robots in complex dynamic environments.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method and system for automatic robot balance adjustment based on multiple sensors. Background Technology

[0002] Currently, robotics plays a crucial role in industry, healthcare, and service sectors, with its stability and precision directly impacting the efficiency and safety of task execution. As robot applications become increasingly complex, such as moving across uneven terrain or performing high-precision operations, maintaining dynamic balance has become a core requirement. Precise balance control is not only fundamental for robots to complete complex tasks but also key to ensuring their long-term operational reliability and a vital link in achieving precise process control.

[0003] In existing technologies, robot balance control methods primarily rely on data from a single type of sensor to capture the robot's posture information. The system adjusts actuators based on static, pre-calibrated compensation parameters using a built-in control program to cope with posture changes. This method can provide some balance maintenance in relatively simple environments. However, these methods often reveal significant shortcomings when dealing with dynamic balance problems in complex environments. Due to reliance on a single sensor, the system struggles to comprehensively capture the robot's complete vibration characteristics under complex working conditions, resulting in weak adaptability to environmental changes. More importantly, these methods are typically based on fixed calibration parameters, making it difficult to adapt in real-time to changes in the robot's dynamic vibration patterns caused by load variations, changes in ground conditions, or mechanical wear. For example, when a service robot is carrying heavy objects, if the ground suddenly becomes uneven, its vibration pattern will change instantaneously. A system based on fixed parameters cannot identify and adjust its control strategy in time, making it highly susceptible to tilting or even falling due to imbalance.

[0004] In summary, existing technologies suffer from insufficient dynamic adaptability, making it difficult to ensure the long-term stable operation of robots in complex environments. Summary of the Invention

[0005] This invention provides a method and system for automatic robot balance adjustment based on multiple sensors, in order to solve the problem that existing technologies have insufficient dynamic adaptability and are difficult to ensure long-term stable operation of robots in complex environments.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for automatic robot balance adjustment based on multiple sensors, comprising: Acquire multi-source synchronous vibration signals and ground condition feedback data, perform data preprocessing, and obtain a preliminary vibration signal sequence; Based on the preliminary vibration signal sequence, frequency domain analysis is performed to obtain the resonant frequency distribution characteristics; Based on the resonant frequency distribution characteristics, mode shift is determined to obtain the mode shift index; Based on the mode offset index, the ground condition feedback data is integrated to update the preset vibration model and obtain optimized vibration model parameters. If the optimized vibration model parameters exceed the preset stability range, the control gain value is calculated to obtain the adjusted control gain. Based on the adjusted control gain, compensation signals are generated and balance control commands are updated to form closed-loop control until the system stability reaches the preset threshold.

[0007] Preferably, the step of acquiring multi-source synchronous vibration signals and ground condition feedback data, performing data preprocessing, and obtaining a preliminary vibration signal sequence includes: Ground condition feedback data is obtained from ground stress sensors; Raw vibration data is obtained from accelerometers and gyroscopes, and timestamps are aligned to obtain a time-synchronized data sequence. Based on the time-synchronized data sequence, multi-level decomposition and reconstruction are performed to obtain a denoised data sequence; Based on the denoised data sequence, a smoothing process is performed to obtain a preliminary vibration signal sequence.

[0008] Preferably, the step of performing frequency domain analysis based on the preliminary vibration signal sequence to obtain the resonant frequency distribution characteristics includes: Based on the preliminary vibration signal sequence, windowing and frequency domain transformation are performed to obtain the frequency domain spectrum; Based on the frequency domain spectrum, peak search and feature identification are performed to obtain the resonant frequency distribution characteristics.

[0009] Preferably, the step of determining the mode shift based on the resonant frequency distribution characteristics to obtain a mode shift index includes: The main harmonic components are extracted from the resonant frequency distribution characteristics. The harmonic components are compared with a preset reference harmonic template and the deviation is calculated to obtain the deviation value; If the deviation value exceeds the preset deviation range, it is determined that there is a mode shift, and the spectral shift is calculated to obtain the mode shift index.

[0010] Preferably, the step of updating the preset vibration model based on the mode offset index and fusing the ground condition feedback data to obtain optimized vibration model parameters includes: Based on the mode offset index, the data is fused with the ground condition feedback data to obtain the fused state estimation sequence. Based on the state estimation sequence, the power spectral density is calculated to obtain the frequency domain energy distribution; Based on the frequency domain energy distribution, the preset vibration model is optimized by parameter fitting to obtain the optimized vibration model parameters.

[0011] Preferably, if the optimized vibration model parameters exceed a preset stability range, then a control gain value is calculated to obtain the adjusted control gain, including: If the optimized vibration model parameters exceed the preset stability range, then the deviation between the optimized vibration model parameters and the preset target stability state is calculated. Based on the deviation between the optimized vibration model parameters and the preset target stable state, dynamic compensation calculations are performed to obtain multiple control components; The adjusted control gain is obtained by performing a weighted summation based on the multiple control components.

[0012] Preferably, the step of generating compensation signals and updating balance control commands based on the adjusted control gain to form closed-loop control until the system stability reaches a preset threshold includes: Based on the adjusted control gain, a signal is generated to obtain the initial compensation signal; Based on the initial compensation signal, output limiting processing is performed to obtain the final compensation signal; Based on the final compensation signal, the actuator is driven and monitored until the system stability reaches the preset threshold.

[0013] Secondly, the present invention provides a robot balance automatic adjustment system based on multiple sensors, comprising: The acquisition and preprocessing module is used to acquire multi-source synchronous vibration signals and ground condition feedback data, perform data preprocessing, and obtain a preliminary vibration signal sequence. The frequency domain analysis module is used to perform frequency domain analysis based on the preliminary vibration signal sequence to obtain the resonant frequency distribution characteristics; The mode determination module is used to determine the mode shift based on the resonant frequency distribution characteristics and obtain the mode shift index. The model update module is used to update the preset vibration model based on the mode offset index and the ground condition feedback data to obtain optimized vibration model parameters. The gain calculation module is used to calculate the control gain value and obtain the adjusted control gain if the optimized vibration model parameters exceed the preset stability range. The closed-loop control module is used to generate compensation signals and update balance control commands based on the adjusted control gain to form closed-loop control until the system stability reaches the preset threshold.

[0014] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the multi-sensor-based automatic robot balance adjustment method described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the multi-sensor-based automatic robot balance adjustment method described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains synchronous vibration signals by fusing multiple source sensors such as accelerometers and gyroscopes, and performs preprocessing such as noise reduction and smoothing. This overcomes the shortcomings of incomplete information from a single sensor, forming a more complete and accurate description of the robot's complex vibration modes, thereby improving the accuracy and comprehensiveness of the robot's dynamic state perception and providing a high-quality data foundation for subsequent balance control.

[0017] (2) This invention analyzes the resonant frequency characteristics to determine the mode shift caused by load changes and integrates ground condition feedback data to dynamically update the vibration model, thus eliminating the dependence of existing technologies on static, preset control parameters. At the same time, by monitoring the vibration mode changes caused by factors such as load in real time and dynamically adjusting its own model in combination with external environmental information, the robot can recognize and adapt to the current working conditions in real time, thereby enhancing the robot's dynamic adaptability and robustness in environments with load changes and complex and variable ground conditions.

[0018] (3) By constructing a closed-loop control circuit, this invention dynamically generates a precise compensation signal based on the deviation of the model parameters from the target stable state when the model parameters exceed the stable range. This signal is then used to correct the balance strategy until the system returns to stability, ensuring that any trend deviating from the stable state can be responded to in a timely and accurate manner. At the same time, by dynamically calculating the compensation signal and correcting it in real time, the system can actively suppress unstable vibrations rather than responding passively, ensuring the real-time performance and accuracy of the robot's balance control, thereby significantly improving its operational stability and safety under highly dynamic tasks. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the automatic robot balance adjustment method based on multiple sensors provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a multi-sensor-based automatic robot balance adjustment system provided in the second embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 The first embodiment of the present invention provides a method for automatic robot balance adjustment based on multiple sensors, including the following steps: S11: Acquire multi-source synchronous vibration signals and ground condition feedback data, perform data preprocessing, and obtain a preliminary vibration signal sequence; S12, Based on the preliminary vibration signal sequence, perform frequency domain analysis to obtain the resonant frequency distribution characteristics; S13, Based on the resonant frequency distribution characteristics, perform mode shift judgment to obtain the mode shift index; S14. Based on the mode offset index, the ground condition feedback data is fused to update the preset vibration model and obtain optimized vibration model parameters. S15, if the optimized vibration model parameters exceed the preset stability range, then the control gain value is calculated to obtain the adjusted control gain; S16, based on the adjusted control gain, generate compensation signals and update balance control commands to form closed-loop control until the system stability reaches the preset threshold.

[0022] In step S11, multi-source synchronous vibration signals and ground condition feedback data are acquired, and data preprocessing is performed to obtain a preliminary vibration signal sequence, including: Ground condition feedback data is obtained from ground stress sensors; Raw vibration data is obtained from accelerometers and gyroscopes, and timestamps are aligned to obtain a time-synchronized data sequence. Based on the time-synchronized data sequence, multi-level decomposition and reconstruction are performed to obtain a denoised data sequence; Based on the denoised data sequence, a smoothing process is performed to obtain a preliminary vibration signal sequence.

[0023] This invention acquires multi-source dynamic data during robot operation and generates a preliminary vibration signal sequence that accurately reflects the robot's vibration state through a series of signal processing operations. This step is the foundation for all subsequent analysis and control. Specifically, ground condition feedback data is first obtained from one or more piezoresistive ground stress sensors placed at the contact points between the robot and the operating plane (such as the soles of its feet). These sensors convert the sensed physical stress into electrical signals in real time and sample and quantize them at a preset sampling frequency to obtain a one-dimensional time series composed of timestamps and stress values. The sampling frequency is set based on experimental statistics. By analyzing the ground stress signal spectrum when the robot performs actions on planes of various typical materials, the sampling frequency is set to more than twice the upper limit of the main energy distribution frequency according to the Nyquist sampling theorem to ensure information integrity.

[0024] Next, raw vibration data containing independent timestamps, generated by the three-axis accelerometers and three-axis gyroscopes, is acquired from the inertial measurement unit (IMU) installed at the robot's center of mass. To obtain complete motion data at the same moment, timestamp alignment processing is required. This processing uses the accelerometer timestamp sequence as a reference. For each reference timestamp, the corresponding angular velocity value of the gyroscope at that moment is calculated using a linear interpolation algorithm. For example, the interpolation formula for the X-axis angular velocity is: in, The desired accelerometer timestamp X-axis angular velocity interpolation at the location; and In the raw data of the gyroscope The two closest timestamps; and These are the raw X-axis angular velocities measured by the gyroscope at these two timestamps. After processing all axes using this method, a time-synchronized data sequence containing a unified timestamp and six motion components can be obtained.

[0025] Subsequently, based on the time-synchronized data sequence, a multi-level decomposition and reconstruction denoising process was performed using Discrete Wavelet Transform (DWT). This process employed the "Daubechies4" wavelet basis, known for its performance in processing transient signals, to decompose the signal into multiple levels (e.g., decomposition level L=3). This decomposition process separated the signal into low-frequency approximation coefficients representing the main signal profile and high-frequency detail coefficients representing signal details and noise. The number of decomposition levels L was determined based on a trade-off analysis between signal-to-noise ratio and signal distortion. By performing different levels of decomposition and denoising reconstruction on sample signals containing typical vibration characteristics, a decomposition level was selected that maximizes noise removal without significantly distorting the original signal waveform profile.

[0026] Next, soft thresholding is performed on the high-frequency detail coefficients representing noise at each layer. This process is based on a pre-calculated wavelet denoising threshold. The magnitude of the coefficient is determined by a threshold value. If the absolute value of the coefficient is less than the threshold, it is set to zero; if it is greater than the threshold, it is shrunk towards zero by a threshold value, thus filtering out noise while preserving the effective components of the signal. The wavelet denoising threshold is... The calculation is based on the general threshold rule, and the formula is as follows: in, It is the length of the detail coefficient of that layer. This represents the standard deviation of the noise in that layer, which can usually be estimated by the absolute value of the median of the wavelet coefficients in that layer. After thresholding, the signal is reconstructed using inverse discrete wavelet transform (IDWT). This reconstruction process uses the unprocessed low-frequency approximation coefficients and the thresholded high-frequency detail coefficients to restore the signal from the frequency domain to the time domain, resulting in a denoised data sequence.

[0027] Finally, based on the denoised data sequence, a moving average filter is used for smoothing to remove residual minor spikes. This filter operates on a size of... When sliding within the window, the value at the window's center point is replaced with the arithmetic mean of all data points within the window, calculated using the following formula: It should be noted that the window size... The window size is set to a positive odd number to ensure symmetry around the center point. The setting is based on experimental statistics. By evaluating the smoothness and signal delay of the filtered signal under different window sizes, a window size that achieves the best balance between effectively filtering out glitches and introducing acceptable signal delay is selected. Finally, a preliminary vibration signal sequence is obtained for subsequent frequency domain analysis.

[0028] In step S12, frequency domain analysis is performed based on the preliminary vibration signal sequence to obtain the resonant frequency distribution characteristics, including: Based on the preliminary vibration signal sequence, windowing and frequency domain transformation are performed to obtain the frequency domain spectrum; Based on the frequency domain spectrum, peak search and feature identification are performed to obtain the resonant frequency distribution characteristics.

[0029] Specifically, this step first performs windowing and frequency domain transformation on the initial vibration signal sequence as input. For example, to reduce spectral leakage caused by signal truncation, a Hanning window function is used to preprocess the signal sequence; that is, the start and end of signal segments are smoothed by multiplying the signal sequence with the window function point by point. In one implementation, a Fast Fourier Transform (FFT) algorithm is used to transform the windowed signal sequence from the time domain to the frequency domain. This transformation is performed independently on the six motion components in the sequence, thereby obtaining the frequency domain spectrum of each component. The frequency domain spectrum is a dataset containing frequency points and their corresponding amplitudes, and its mathematical expression can be defined as… ={( , ) |k=1,2,...,K} where Let k be the frequency value at the k-th frequency point. This represents the vibration amplitude corresponding to that frequency point.

[0030] Subsequently, peak search and feature identification are performed based on the frequency domain spectrum. Preferably, the peak search process involves traversing the spectrum data to identify frequency points with amplitude values ​​greater than the amplitudes of their two immediately preceding and following data points as local peaks. To distinguish significant resonance peaks from random noise, the amplitude of each local peak is... With a preset peak amplitude threshold Comparisons are made. It should be noted that if the peak amplitude... Greater than If the frequency is greater than or equal to the resonant frequency, then that frequency point is identified as a significant resonant frequency; conversely, if it is less than or equal to the resonant frequency... If the peak value is not specified, it is considered noise and ignored.

[0031] Peak amplitude threshold The threshold is set based on experimental statistics. Vibration signal samples of the robot under various typical working conditions during normal and stable operation are collected, and their spectra are calculated. The statistical distribution characteristics of vibration peaks under normal conditions are analyzed, and the threshold is set as a specific percentile (e.g., 99th or 99.5th percentile) of the amplitude distribution of all sample spectra. Therefore, setting the threshold here means that any peak exceeding this threshold is an extremely low probability event and can be considered a statistical outlier, thus providing a robust criterion for distinguishing significant resonant signals from normal background vibrations. It is understood that all the resonant frequencies identified as significant and their corresponding amplitudes together constitute the resonant frequency distribution characteristics, which provide direct data basis for subsequent mode shift judgment.

[0032] In step S13, based on the resonant frequency distribution characteristics, mode shift is determined to obtain a mode shift index, including: The main harmonic components are extracted from the resonant frequency distribution characteristics. The harmonic components are compared with a preset reference harmonic template and the deviation is calculated to obtain the deviation value; If the deviation value exceeds the preset deviation range, it is determined that there is a mode shift, and the spectral shift is calculated to obtain the mode shift index.

[0033] Specifically, this step first extracts the main harmonic components from the resonant frequency distribution characteristics used as input. In one implementation, the system first determines the frequency point with the largest vibration amplitude in the resonant frequency distribution characteristics as the fundamental frequency. Subsequently, the system uses this fundamental frequency as a reference to search for frequencies that are integer multiples (n, n) of the fundamental frequency within the resonant frequency distribution characteristics. The frequency points (n=2,3,...) are determined, and a small frequency tolerance is allowed based on experimental statistics of the maximum frequency drift under different operating conditions of the motor characteristics and mechanical structure, thus obtaining a set of main harmonic components consisting of harmonic order, harmonic frequency and corresponding amplitude.

[0034] Next, the extracted harmonic components are compared and the deviation is calculated with a preset reference harmonic template. It should be noted that the reference harmonic template is a digital description of the robot's vibration characteristics under standard, no-load conditions. Its construction method is based on experimental statistics, establishing a standard vibration amplitude range for each order harmonic by collecting and analyzing a large amount of vibration data under no-load conditions. , The comparison process will compare the current real-time harmonic amplitude. The amplitude range of the corresponding order in the template is compared, and a comprehensive deviation value is calculated. (This deviation value is a normalized dimensionless index that comprehensively reflects the deviation of the amplitude of each harmonic from its reference range.) For example, this deviation value... The normalized out-of-range values ​​of each harmonic exceeding the template range can be obtained by weighted summation, and the calculation formula is as follows: in The preset weighting coefficients are typically assigned higher weights to lower-order harmonics that better reflect load changes, based on domain experience. These weights are set through experimental statistics and sensitivity analysis. The robot is subjected to a series of known, increasing loads (e.g., no load, 2 kg, 4 kg, etc.) under controlled conditions, and vibration data is collected for each load. Harmonic analysis is performed on the data, and the correlation coefficient between the amplitude of each harmonic order and the load magnitude is calculated. Harmonic orders with higher correlation coefficients, i.e., more sensitive to load changes, are assigned higher weight values. This ensures that the calculation of the deviation value can more accurately reflect the mode shift caused by load changes.

[0035] Finally, the calculated deviation value is judged. Preferably, the deviation value is... Compare with a preset deviation threshold. If If the deviation exceeds a threshold, the robot is considered to have experienced a mode shift due to load changes; conversely, if the deviation is less than or equal to the threshold, the system is considered to be operating normally. The deviation threshold is also based on experimental statistics, determined by testing the robot's maximum deviation under acceptable light loads, ensuring the reliability of the judgment. If a mode shift is detected, the system will further calculate the spectral shift to obtain a mode shift index. In one possible implementation, this index can be quantified as the current base frequency. The change in the fundamental frequency relative to the unloaded state in the reference harmonic template, i.e., the mode offset index. This index is ultimately output and used to precisely adjust the vibration model in the next step.

[0036] In step S14, based on the mode offset index and by fusing the ground condition feedback data, the preset vibration model is updated to obtain optimized vibration model parameters, including: Based on the mode offset index, the data is fused with the ground condition feedback data to obtain the fused state estimation sequence. Based on the state estimation sequence, the power spectral density is calculated to obtain the frequency domain energy distribution; Based on the frequency domain energy distribution, the preset vibration model is optimized by parameter fitting to obtain the optimized vibration model parameters.

[0037] In one implementation, the fusion process uses a Kalman filter. It should be noted that the system pre-defines a state vector to describe the robot's vibration characteristics, which is defined as follows:

[0038] ,in Let be the equivalent damping coefficient of the robot system at time k. This is the equivalent stiffness coefficient. The state vector is set based on the fact that the robot's vibration behavior can be effectively simplified and modeled as a second-order mass-spring-damped system, and damping and stiffness are two core physical parameters that determine the dynamic response characteristics of this system. The initial prediction process of the Kalman filter includes: first, based on the optimal state estimate of the previous time step k-1... The state at time k is predicted using the state transition matrix F, i.e. In this application, F can be set as the identity matrix, representing the system parameters changing steadily over a short period of time. Meanwhile, the prediction error covariance... Where Q is the process noise covariance matrix, the value of which is determined by experimentally analyzing the natural drift range of the robot's physical parameters under normal operating conditions. The filter correction process includes: firstly, calculating the Kalman gain. Then, the predicted state is corrected using the current measurement values ​​(i.e., model offset index and ground condition feedback data). The core update formula is: ,in Let H be the measurement vector, and let H be the observation matrix that maps the state vector to the measurement space. Specifically, the observation matrix H is set based on the physical model of the system. It establishes a linear approximation relationship between the state parameters (damping c, stiffness k) and the measured values ​​(mode offset index, ground feedback data), allowing the filter to infer changes in the state parameters from changes in the measured values. Ultimately, a parameter estimation sequence with lower noise and closer to the true physical state is obtained.

[0039] It should be noted that the specific form of the observation matrix H needs to be determined through system identification or mechanism derivation. In an exemplary embodiment, the measurement vector... It can be defined as [ ,in This is a mode offset indicator. These are the main frequency components of the ground stress sensor readings. Through offline experiments, state parameters of the robot under different loads and ground conditions were collected. , With the corresponding measured value The observation matrix H can be fitted using methods such as linear regression. For example, if we determine... Mainly related to stiffness Related, Mainly related to damping If there is a correlation, the H matrix may be approximated as a diagonal matrix H=diag(h1, h2), where h1 and h2 are the coefficients obtained from the regression.

[0040] Subsequently, the system calculates the power spectral density (PSD) of the preliminary vibration signal sequence obtained in step S11 to obtain its frequency domain energy distribution. Preferably, this calculation uses the Welch method, and its specific steps are as follows: First, the preliminary vibration signal sequence of length N is divided into M segments of length L, which overlap each other by 50%, wherein the segment length L is set based on the target frequency resolution. With sampling frequency Relationship ( Next, a Hanning window function is applied to each segment to reduce spectral leakage. Then, a fast Fourier transform is performed on each windowed segment and the square of its amplitude is calculated to obtain the power spectrum of that segment. Finally, the power spectra of all M segments are arithmetically averaged to obtain a smooth and low-noise overall power spectral density estimate, which accurately shows the distribution of vibration energy at different frequencies.

[0041] Finally, based on the frequency domain energy distribution, a preset vibration model is optimized by parameter fitting. For example, the preset vibration model is a second-order linear mass-spring-damped system, whose physical structure is defined by differential equations. Description, in which quality Given the robot's known parameters, the parameters to be optimized are the model's dynamic parameters c (representing the system's equivalent damping coefficient) and k (representing the system's equivalent stiffness coefficient). This model is a physical model and does not require training on a dataset. The parameter fitting process uses the least squares method, which iteratively adjusts the model parameters c and k (using the output of a Kalman filter as initial values ​​to accelerate convergence) to minimize the mean square error between the theoretical power spectrum calculated from these parameters and the frequency domain energy distribution actually calculated in the previous step. The parameter set obtained after iterative convergence is the final output optimized vibration model parameter set, which provides an accurate model basis for subsequent stability assessment and control. In step S15, if the optimized vibration model parameters exceed the preset stability range, the control gain value is calculated to obtain the adjusted control gain, including: If the optimized vibration model parameters exceed the preset stability range, then the deviation between the optimized vibration model parameters and the preset target stability state is calculated. Based on the deviation between the optimized vibration model parameters and the preset target stable state, dynamic compensation calculations are performed to obtain multiple control components; The adjusted control gain is obtained by performing a weighted summation based on the multiple control components.

[0042] Specifically, this step first inputs the optimized vibration model parameters, i.e., the optimized damping coefficients. and stiffness coefficient The value is compared with a preset stability range. It should be noted that the preset stability range is a multi-dimensional region defined in parameter space, its boundaries based on experimental statistical data of the system. This range is determined by collecting and analyzing a set of parameter samples demonstrating the robot's long-term stable operation under various typical loads and ground conditions; for example, it can be defined as a rectangular region. , ]and[ , ],in and These are the lower and upper limits of stability for the equivalent damping coefficient, respectively. and These are the lower and upper limits of the equivalent stiffness coefficient, respectively. If the current optimization parameter is determined to exceed this range, the system initiates deviation calculation, which compares the current parameter with a preset target stable state. , ] for comparison, where The target equivalent damping coefficient, The target equivalent stiffness coefficient is used. The preset target stable state is set based on control theory and experimental verification, representing the ideal operating point where the robot achieves the optimal balance between response speed, overshoot, and energy consumption. Through comparison, the deviation vector is finally obtained. The two components of this vector are the differences between the optimized damping and stiffness coefficients and their corresponding target values, respectively, and their calculation formulas are as follows: Next, based on the calculated deviation vector Dynamic compensation calculations are performed to obtain multiple control components. In one implementation, this calculation includes three parts: proportional, integral, and derivative calculations. First, the deviation from the current value is calculated. proportional components Its formula is First, it is used to quickly respond to the current deviation; second, it calculates the deviation. Integral components that are proportional to the cumulative value over time Its formula is This is used to eliminate potential steady-state errors in the system; finally, the deviation is calculated. Differential component proportional to the rate of change Its formula is It is used to predict future trends of deviations and suppress system oscillations. The proportional gain coefficient is used to... Integral gain coefficient Differential gain coefficient The setting is based on the Ziegler-Nichols tuning method, which is well-known in the field of control systems. This method involves gradually increasing the proportional gain in a real or simulated system and recording the critical gain when the system reaches critical oscillation. and oscillation period Then calculate according to the standard formula , , Recommended value.

[0043] Finally, the obtained control components are weighted and summed to obtain the adjusted control gain. Understandably, this step involves applying the proportional components calculated in the previous step. Integral components and differential components The formula for linear superposition is as follows: The final adjusted control gain It is a control command that integrates current deviation, historical deviation, and future deviation trends. This command is output to the next step to generate specific physical compensation signals.

[0044] In step S16, based on the adjusted control gain, a compensation signal is generated and a balance control command is updated to form closed-loop control until the system stability reaches a preset threshold, including: Based on the adjusted control gain, a signal is generated to obtain the initial compensation signal; Based on the initial compensation signal, output limiting processing is performed to obtain the final compensation signal; Based on the final compensation signal, the actuator is driven and monitored until the system stability reaches the preset threshold.

[0045] Specifically, this step first involves adjusting the control gain as input. Signal generation is performed to obtain an initial compensation signal. In one implementation, the adjusted control gain is a vector representing the desired adjustment of system dynamic parameters (such as damping and stiffness). The signal generation process linearly converts this high-dimensional control gain vector into physical signals (such as voltage or PWM duty cycle) recognizable by the actuators of each joint at the robot's bottom layer through a preset control allocation matrix M. The control allocation matrix M is set based on the robot's dynamic model and system identification experiments, establishing a mapping relationship between the top-level balance control objective (adjusting c and k) and the torque output of each joint at the bottom layer. For example, the calculation formula for this conversion process can be expressed as follows: This transforms the abstract control gain into a specific initial compensation signal vector. .

[0046] Next, based on the initial compensation signal, output limiting processing is performed to obtain the final compensation signal. It should be noted that this processing is to protect the robot's hardware actuators from physical damage or saturation effects caused by excessive control commands. Preferably, the limiting processing reduces the initial compensation signal... Each component is within a preset signal amplitude safety range. , The upper and lower limits of the safety range are compared. and The setting is based on the manufacturer's product specifications for the actuator (such as a motor and its driver), with particular attention to its "rated continuous operating voltage / current" range, and a certain safety margin is set on this basis. For example, for a DC servo motor with a rated continuous operating voltage of ±24V, its safety range can be set to ±22V. If the signal component exceeds this range, its value is constrained to the boundary, and the calculation formula is as follows: The result after processing is the final compensation signal. This signal can effectively execute control intentions while ensuring the absolute safety of the hardware.

[0047] Finally, actuator driving and deviation monitoring are performed based on the final compensation signal. The system sends the final compensation signal to the corresponding joint motor drive units of the robot. Specifically, these voltage signals drive the joint motors to generate precise torques, restoring balance by adjusting the robot's posture and movement. For example, if the robot is detected to be tilting forward, the compensation signal will drive the ankle joint motors to generate a torque that rotates the robot's upper body backward, and may simultaneously drive the hip joint motors to produce a coordinated movement. This fine adjustment based on ankle or hip joint strategies can effectively shift the robot's center of gravity back to a stable region within its support polygon.

[0048] While driving the actuator, the system enters a continuous monitoring loop, repeatedly executing steps S14 and S15 to calculate the latest deviation vector and its norm (deviation value) in real time, and compare it with a preset stability requirement threshold. The stability requirement threshold is a scalar limit used to determine whether the closed-loop control has converged. It is set based on task performance-oriented experimental statistics. Specifically, a set of performance evaluation indicators for stable robot operation is first defined, such as the torso's angular sway being less than 0.5 degrees when performing a specific task.

[0049] Subsequently, the robot was operated under various typical operating conditions, and deviation vectors were collected within the "acceptable stability" time period that met all performance evaluation indicators. Finally, the norm of these deviation vectors (i.e., deviation values) under these "acceptable stability" states was statistically analyzed, and the 99th percentile of this statistical distribution was set as the stability requirement threshold. This method aims to provide an objective and robust boundary for the "steady state." Choosing the 99th percentile means that the vast majority (99%) of normal deviation fluctuations observed within the acceptable performance range are considered stable states, thus avoiding overreaction by the control system to small, harmless disturbances due to excessive sensitivity. This method ensures that active control continues only when the system's control deviation enters a verified, statistically significant outlier range. This closed-loop control process continues until the monitored deviation value falls below the required threshold, at which point the system is considered to have reached a steady state.

[0050] In summary, this invention significantly improves the vibration stability and operational reliability of mechanical systems in complex environments by integrating multi-source data fusion, frequency domain analysis, and dynamic control, thereby achieving precise process control.

[0051] Reference Figure 2 The second embodiment of the present invention provides a robot balance automatic adjustment system based on multiple sensors, comprising: The acquisition and preprocessing module is used to acquire multi-source synchronous vibration signals and ground condition feedback data, perform data preprocessing, and obtain a preliminary vibration signal sequence. The frequency domain analysis module is used to perform frequency domain analysis based on the preliminary vibration signal sequence to obtain the resonant frequency distribution characteristics; The mode determination module is used to determine the mode shift based on the resonant frequency distribution characteristics and obtain the mode shift index. The model update module is used to update the preset vibration model based on the mode offset index and the ground condition feedback data to obtain optimized vibration model parameters. The gain calculation module is used to calculate the control gain value and obtain the adjusted control gain if the optimized vibration model parameters exceed the preset stability range. The closed-loop control module is used to generate compensation signals and update balance control commands based on the adjusted control gain to form closed-loop control until the system stability reaches the preset threshold.

[0052] It should be noted that the multi-sensor-based automatic robot balance adjustment system provided in this embodiment of the invention is used to execute all the process steps of the multi-sensor-based automatic robot balance adjustment method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0053] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a multi-sensor-based automatic robot balance adjustment program. When the processor executes the computer program, it implements the steps in the various embodiments of the multi-sensor-based automatic robot balance adjustment method described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the frequency domain analysis module.

[0054] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0055] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0056] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0057] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0058] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0059] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0060] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for automatic balance adjustment of a robot based on multiple sensors, characterized in that, include: Acquire multi-source synchronous vibration signals and ground condition feedback data, perform data preprocessing, and obtain a preliminary vibration signal sequence; Based on the preliminary vibration signal sequence, frequency domain analysis is performed to obtain the resonant frequency distribution characteristics; Based on the resonant frequency distribution characteristics, mode shift is determined to obtain the mode shift index; Based on the mode offset index, the ground condition feedback data is integrated to update the preset vibration model and obtain optimized vibration model parameters. If the optimized vibration model parameters exceed the preset stability range, the control gain value is calculated to obtain the adjusted control gain. Based on the adjusted control gain, compensation signals are generated and balance control commands are updated to form closed-loop control until the system stability reaches the preset threshold.

2. The automatic robot balance adjustment method based on multiple sensors according to claim 1, characterized in that, The process of acquiring multi-source synchronous vibration signals and ground condition feedback data, performing data preprocessing, and obtaining a preliminary vibration signal sequence includes: Ground condition feedback data is obtained from ground stress sensors; Raw vibration data is obtained from accelerometers and gyroscopes, and timestamps are aligned to obtain a time-synchronized data sequence. Based on the time-synchronized data sequence, multi-level decomposition and reconstruction are performed to obtain a denoised data sequence; Based on the denoised data sequence, a smoothing process is performed to obtain a preliminary vibration signal sequence.

3. The automatic robot balance adjustment method based on multiple sensors according to claim 1, characterized in that, The step of performing frequency domain analysis based on the preliminary vibration signal sequence to obtain the resonant frequency distribution characteristics includes: Based on the preliminary vibration signal sequence, windowing and frequency domain transformation are performed to obtain the frequency domain spectrum; Based on the frequency domain spectrum, peak search and feature identification are performed to obtain the resonant frequency distribution characteristics.

4. The automatic robot balance adjustment method based on multiple sensors according to claim 1, characterized in that, The step of determining the mode shift based on the resonant frequency distribution characteristics to obtain the mode shift index includes: The main harmonic components are extracted from the resonant frequency distribution characteristics. The harmonic components are compared with a preset reference harmonic template and the deviation is calculated to obtain the deviation value; If the deviation value exceeds the preset deviation range, it is determined that there is a mode shift, and the spectral shift is calculated to obtain the mode shift index.

5. The automatic robot balance adjustment method based on multiple sensors according to claim 1, characterized in that, The step of updating the preset vibration model based on the mode offset index and integrating the ground condition feedback data to obtain optimized vibration model parameters includes: Based on the mode offset index, the data is fused with the ground condition feedback data to obtain the fused state estimation sequence. Based on the state estimation sequence, the power spectral density is calculated to obtain the frequency domain energy distribution; Based on the frequency domain energy distribution, the preset vibration model is optimized by parameter fitting to obtain the optimized vibration model parameters.

6. The automatic robot balance adjustment method based on multiple sensors according to claim 1, characterized in that, If the optimized vibration model parameters exceed the preset stability range, then the control gain value is calculated to obtain the adjusted control gain, including: If the optimized vibration model parameters exceed the preset stability range, then the deviation between the optimized vibration model parameters and the preset target stability state is calculated. Based on the deviation between the optimized vibration model parameters and the preset target stable state, dynamic compensation calculations are performed to obtain multiple control components; The adjusted control gain is obtained by performing a weighted summation based on the multiple control components.

7. The automatic robot balance adjustment method based on multiple sensors according to claim 1, characterized in that, The step of generating compensation signals and updating balance control commands based on the adjusted control gain to form closed-loop control until the system stability reaches a preset threshold includes: Based on the adjusted control gain, a signal is generated to obtain the initial compensation signal; Based on the initial compensation signal, output limiting processing is performed to obtain the final compensation signal; Based on the final compensation signal, the actuator is driven and monitored until the system stability reaches the preset threshold.

8. A robot balance automatic adjustment system based on multiple sensors, characterized in that, include: The acquisition and preprocessing module is used to acquire multi-source synchronous vibration signals and ground condition feedback data, perform data preprocessing, and obtain a preliminary vibration signal sequence. The frequency domain analysis module is used to perform frequency domain analysis based on the preliminary vibration signal sequence to obtain the resonant frequency distribution characteristics; The mode determination module is used to determine the mode shift based on the resonant frequency distribution characteristics and obtain the mode shift index. The model update module is used to update the preset vibration model based on the mode offset index and the ground condition feedback data to obtain optimized vibration model parameters. The gain calculation module is used to calculate the control gain value and obtain the adjusted control gain if the optimized vibration model parameters exceed the preset stability range. The closed-loop control module is used to generate compensation signals and update balance control commands based on the adjusted control gain to form closed-loop control until the system stability reaches the preset threshold.

Images