A multi-layer algorithm fusion method, system and storage medium for controlling an actuator

Abstract

The application relates to the technical field of machine control, and discloses a multi-layer algorithm fusion method and system for controlling actuators and a storage medium. The method comprises the following steps: acquiring multiple sensor data, distributing the multiple sensor data to built-in control algorithms, calculating single-line control data, and obtaining a single-line control data matrix according to the multiple single-line control data; forming a sensor data matrix from the multiple sensor data, and calculating a fusion control data matrix by using a fusion algorithm; calculating a control matrix distance according to the single-line control data matrix and the fusion control data matrix; if the control matrix distance is smaller than a reference matrix distance, calculating a final control data matrix according to the single-line control data matrix and the fusion control data matrix, and controlling multiple actuators according to the final control data matrix; otherwise, controlling the multiple actuators by using the single-line control data matrix; and through the synergistic effect of an upgrading algorithm and a solidification algorithm, the control precision and stability after vehicle motion control are improved.

Description

Technical Field

[0001] This application relates to the technical field of machine control, and in particular to a multi-layer algorithm fusion method, system and storage medium for controlling actuators. Background Technology

[0002] In the control process of vehicles or robots, the corresponding sensors first accurately acquire various relevant data during operation according to the preset rules and system-defined acquisition logic. The directly acquired data is then input into the calculation module. The calculation module processes the raw data to obtain control data for the actuators. Finally, this control data is transmitted to the corresponding actuators to drive them to perform corresponding control actions, thus achieving initial regulation of the vehicle's operating status.

[0003] The computational algorithms used to process the acquired data and generate control data are pre-embedded within the vehicle's chip. Through predetermined computational logic, calculations are performed to obtain control data corresponding to different control requirements. This control data is then precisely allocated to various actuators within the vehicle, such as the power actuator responsible for regulating vehicle power output, the brake actuator ensuring driving safety, the steering actuator controlling the vehicle's direction of travel, and the suspension actuator maintaining vehicle stability. Each actuator then performs its corresponding control action based on the received control data.

[0004] However, the algorithms embedded in vehicle chips have significant limitations, as they cannot be directly upgraded or updated. With the increasing demands for vehicle control and the growing number of complex driving scenarios, the control data generated by the existing embedded algorithms can no longer meet the requirements for higher-precision motion control, resulting in unsatisfactory control performance. Summary of the Invention

[0005] In order to upgrade the algorithm by performing fusion calculations on the fixed algorithm, and to improve the control accuracy and stability after vehicle motion control through the synergistic effect of the upgraded algorithm and the fixed algorithm, this application provides a multi-layer algorithm fusion method, system and storage medium for control actuators.

[0006] Firstly, this application provides a multi-layer algorithm fusion method for controlling actuators, employing the following technical solution:

[0007] A multi-layer algorithm fusion method for controlling actuators includes the following steps:

[0008] The system acquires various sensor data based on preset sensors, assigns the various sensor data to corresponding built-in control algorithms, calculates single-line control data using the built-in control algorithms, and obtains a single-line control data matrix based on the various single-line control data.

[0009] Multiple sensor data are combined to form a sensor data matrix, and a fusion control data matrix is ​​obtained by inputting the sensor data matrix into a preset fusion algorithm.

[0010] The control matrix distance is calculated using a matrix distance algorithm based on the single-line control data matrix and the fused control data matrix.

[0011] If the distance between the control matrices is less than the preset distance between the reference matrices, the final control data matrix is ​​calculated based on the single-line control data matrix and the fused control data matrix, and multiple actuators are controlled based on the final control data matrix; otherwise, the single-line control data matrix is ​​used to control multiple actuators.

[0012] By adopting the above technical solution, data is allocated to a one-to-one corresponding built-in control algorithm to generate single-line control data and form a single-line control data matrix, thus preserving the stable output of the original algorithm. Multiple sensor data are combined to form a sensor data matrix, which is then used by a fusion algorithm to obtain a fused control data matrix. The control matrix distance is calculated using a matrix distance algorithm and compared with a reference matrix distance. When the control matrix distance is less than the reference matrix distance, the single-line control data matrix and the fused control data matrix are combined to obtain the final control data matrix to control multiple actuators; otherwise, the single-line control data matrix is ​​used. This achieves an organic integration of the old and new algorithms, balancing control effectiveness and reliability.

[0013] Optionally, the fusion algorithm includes:

[0014] Obtain actuator type data;

[0015] Based on the type of data, match the associated control parameters corresponding to the sensing data of different sensors;

[0016] Based on the sensing data from different sensors, the execution control data is calculated using a weighted algorithm with associated control parameters, where each execution control data corresponds one-to-one with an actuator.

[0017] Multiple execution control data are combined to form a fused control data matrix.

[0018] By adopting the above technical solutions, the fusion algorithm can accurately associate sensor data and control logic based on the characteristics of the actuator, making the associated control parameters more adaptable. The weighted algorithm can specifically integrate multi-source sensor data and improve the control effect of the execution control data.

[0019] Optionally, the matrix distance algorithm includes:

[0020] The difference matrix is ​​obtained by calculating the difference between corresponding elements of the single-line control data matrix and the fused control data matrix;

[0021] The mean absolute error of the difference matrix is ​​calculated as the distance to the control matrix; where the mean absolute error is... .

[0022] By adopting the above technical solution, the average absolute error is calculated using the difference matrix as the control matrix distance, which can accurately quantify the difference between the two control data matrices.

[0023] Optionally, the step of calculating the final control data matrix based on the single-line control data matrix and the fused control data matrix further includes the following sub-steps:

[0024] Associate the elements at corresponding positions in the single-line control data matrix with the elements at the corresponding positions in the fused control data matrix one by one;

[0025] The final element is calculated based on the associated elements using a pre-defined weighted average algorithm.

[0026] The final elements form the final control data matrix.

[0027] By adopting the above technical solution, the elements of the single-line control data matrix and the fusion control data matrix are accurately correlated. The final elements obtained by the weighted average algorithm are more in line with the actual control requirements, so that the final control data matrix can take into account the advantages of both matrices and provide better data for actuator control.

[0028] Optionally, within a preset time period, the percentage of times the control matrix distance is less than the preset reference matrix distance is recorded, and the weighting coefficient of the corresponding element in the fused control data matrix is ​​adjusted according to the positive correlation of the percentage of times. The larger the percentage of times, the larger the weighting coefficient, and the smaller the percentage of times, the smaller the weighting coefficient.

[0029] By adopting the above technical solution, the weighting coefficient is dynamically adjusted according to the proportion of occurrences within a time period. When the proportion of occurrences is high, increasing the weight can enhance the fusion effect, while decreasing the weight when the proportion of occurrences is low can reduce the impact of deviation.

[0030] Optionally, within a preset time period, the percentage of times the control matrix distance is less than the preset reference matrix distance is recorded, and the average distance is calculated based on the corresponding number of control matrix distances to update the reference matrix distance.

[0031] By adopting the above technical solution, the reference matrix distance is dynamically updated, and the average value of the corresponding control matrix distance is calculated by selecting the proportion of times, thereby improving the adaptability of the reference matrix distance and making subsequent matrix distance comparisons more accurate.

[0032] Optionally, within a preset time period, the percentage of times the control matrix distance is less than the preset reference matrix distance is recorded, and the reference matrix distance is adjusted inversely based on the percentage of times. The larger the percentage of times, the smaller the reference matrix distance, and vice versa.

[0033] By adopting the above technical solution, the control matrix distance can be reduced when the proportion of occurrences is high, thereby increasing the control threshold of the fused control data matrix; when the proportion is low, the control matrix distance can be expanded to ensure the continuity of system control and make the control matrix distance judgment more in line with reality.

[0034] Optionally, the step of controlling multiple actuators according to the final control data matrix further includes the following sub-steps:

[0035] The correlation control parameters are adjusted based on the inverse correlation of the control matrix distance. The larger the control matrix distance, the smaller the correlation control parameters, and vice versa.

[0036] By adopting the above technical solution, the associated control parameters can be dynamically adjusted according to the distance of the control matrix. When the distance of the control matrix is ​​large, reducing the associated control parameters can reduce the influence of deviation, while increasing the associated control parameters when the distance of the control matrix is ​​small can enhance the fusion effect.

[0037] Secondly, this application provides a multi-layer algorithm fusion system for controlling actuators, which adopts the following technical solution:

[0038] A multi-layer algorithm fusion system for controlling an actuator includes a processor, wherein the processor executes the steps of the multi-layer algorithm fusion method for controlling an actuator as described in any of the preceding claims.

[0039] Thirdly, this application provides a storage medium, which adopts the following technical solution:

[0040] A storage medium storing a program, wherein the program, when executed by a processor, implements the steps of the multi-layer algorithm fusion method for controlling an actuator as described above.

[0041] In summary, this application includes at least one of the following beneficial technical effects: Data is allocated to a one-to-one corresponding built-in control algorithm to generate single-line control data and form a single-line control data matrix, preserving the stable output of the original algorithm. Multiple sensor data are combined to form a sensor data matrix, which is then used by a fusion algorithm to obtain a fused control data matrix. The control matrix distance is calculated using a matrix distance algorithm and compared with a reference matrix distance. When the control matrix distance is less than the reference matrix distance, the single-line control data matrix and the fused control data matrix are combined to obtain the final control data matrix to control multiple actuators; otherwise, the single-line control data matrix is ​​used. This achieves an organic integration of the old and new algorithms, balancing control effectiveness and reliability. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the steps of a multi-layer algorithm fusion method for controlling actuators.

[0043] Figure 2 This is a diagram of the sub-steps of the fusion algorithm.

[0044] Figure 3 This is a diagram of the sub-steps of the matrix distance algorithm. Detailed Implementation

[0045] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0046] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0047] This application discloses a multi-layer algorithm fusion method for controlling actuators, referring to... Figure 1 It includes the following steps:

[0048] The system initiates a pre-defined sensor acquisition process, controlling various functional sensors mounted on the vehicle or robot to synchronously acquire sensor data in their respective dimensions at a preset sampling frequency (e.g., 10Hz). These sensors include vehicle speed sensors for power control, brake pressure sensors for braking control, steering angle sensors for steering control, and suspension displacement sensors for suspension control. During acquisition, filtering algorithms, such as Kalman filtering, are used to remove data noise, ensuring that the accuracy error of the sensor data does not exceed a preset threshold, such as ±0.1%. The acquired sensor data is then categorized and mapped: vehicle speed data is assigned to the power control algorithm embedded in the chip, brake pressure data to the braking control algorithm, steering angle data to the steering control algorithm, and suspension displacement data to the suspension control algorithm, achieving a one-to-one correspondence between sensor data and built-in control algorithms. Each built-in control algorithm operates according to its pre-defined computational logic. For example, the power control algorithm calculates the input single-dimensional sensor data based on the PID formula of vehicle speed and target power demand, outputting single-line control data corresponding only to a single actuator. For instance, the power control algorithm outputs the torque adjustment value of the power motor, and the braking control algorithm outputs the pressure control value of the brake caliper. Finally, all single-line control data are arranged in the order of actuators for power, braking, steering, and suspension, constructing a single-line control data matrix with a dimension of "1 × number of actuators". Each element in the matrix uniquely maps to the basic control command of an actuator, ensuring the stable output characteristics of the original fixed algorithm.

[0049] The collected multi-sensor data undergoes preprocessing: data format standardization and missing data completion are performed. Standardizing the data format involves converting all data to decimal floating-point numbers. Missing data is filled using linear interpolation to fill in occasionally missing sampling points. A sensor data matrix is ​​then constructed according to the rule of rows for sensor type and columns for sampling time sequence. For example, the row dimension could be "vehicle speed-braking pressure-steering angle-suspension displacement," and the column dimension could be the number of samples, ensuring the integrity and dimensional consistency of the matrix data. This sensor data matrix is ​​then input into a preset fusion algorithm to generate a fused control data matrix through multi-dimensional data collaborative calculation. Figure 2 The specific implementation process is as follows:

[0050] By reading the hardware configuration file of the vehicle or robot's ECU (Electronic Control Unit), the type, model and quantity of actuators that need to be controlled can be obtained, such as 1 power actuator, 1 brake actuator, 1 steering actuator and 4 suspension actuators. At the same time, the control response characteristics of each actuator can be extracted, such as the maximum pressure response speed of the brake actuator and the angle adjustment range of the steering actuator.

[0051] The system calls a preset parameter matching database, which stores the mapping relationship between "actuator type - sensor type - associated control parameters". For example, the association weight of the steering angle sensor corresponding to the steering actuator is 0.6, and the association weight of the vehicle speed sensor is 0.4. Based on the acquired actuator type data, the system matches the associated control parameters corresponding to each actuator and the sensor data of different sensors to ensure that the parameters are adapted to the response characteristics of the actuator. For example, the associated parameters of the displacement sensor data of the suspension actuator need to match the stiffness adjustment requirements of the suspension.

[0052] For each row of sensor data in the sensor data matrix, combined with the matched associated control parameters, a preset weighted summation algorithm is used for calculation. The weighting algorithm, such as the weight ratio, is dynamically allocated based on the actuator priority, with safety actuators having a higher weight than comfort actuators. For example, for a steering actuator, its execution control data, i.e., the steering angle adjustment, is calculated using the formula: steering angle data × 0.6 + vehicle speed data × 0.4. After calculating for each actuator, the execution control data is arranged into a fusion control data matrix of "1 × number of actuators" according to the actuator order consistent with the single-line control data matrix, realizing the collaborative integration of multi-source sensor data and improving the scenario adaptability of the control data.

[0053] The control matrix distance is calculated using a matrix distance algorithm based on the single-line control data matrix and the fused control data matrix; a preset matrix distance algorithm is then invoked, employing a matrix distance algorithm based on mean absolute error (MAE), with reference to... Figure 3 To achieve accurate quantification of the differences between the single-line control data matrix and the fused control data matrix, the specific steps are as follows:

[0054] Let A be the single-line control data matrix, with dimensions m×n, where m is the number of rows and n is the number of columns, and its elements be denoted as aij. Let B be the fused control data matrix, with the same dimensions as A, and its elements be denoted as bij. Associate corresponding elements of A and B, and calculate the difference element by element:

[0055] For the element in the i-th row and j-th column of the matrix, calculate the difference dij = aij - bij; or dij = bij - aij, and the results are consistent after taking the absolute value. Traverse all elements of the matrix, i from 1 to m, j from 1 to n, and arrange all differences dij according to the original matrix dimensions to construct a difference matrix D (dimension m×n, elements are dij).

[0056] Based on the difference matrix D, the distance of the control matrix is ​​calculated using the mean absolute error formula, which is defined as follows: Where: |dij| represents the absolute value of each element in the difference matrix, used to eliminate the offsetting effect of positive and negative deviations and highlight the actual magnitude of the deviation; m×n represents the total number of elements in the matrix, which is averaged to make the result suitable for distance comparisons of matrices with different dimensions; during calculation, all elements of the difference matrix are first traversed and summed. Then divide by the total number of elements m×n to obtain the mean absolute error (MAE), and use this value as the final control matrix distance.

[0057] The preset reference matrix distance is retrieved from the parameter storage module. This distance value is obtained through previous real vehicle / robot testing and represents a reasonable deviation threshold between single-line control data and fused control data. For example, the preset value is 0.15. The calculated control matrix distance is compared with the reference matrix distance, and control is executed according to the following logic:

[0058] If the fusion control condition is met, i.e., the control matrix distance < the reference matrix distance: the fusion control data is deemed reliable. The generated final control data matrix is ​​then invoked, and control commands are sent to each actuator, such as the final torque adjustment value to the power actuator and the final pressure control value to the brake actuator, driving the actuators to coordinate their actions. Simultaneously, the inverse correlation adjustment mechanism of the associated control parameters is activated: the difference between the control matrix distance and the reference matrix distance is calculated. For example, if the control matrix distance is 0.1 and the reference matrix distance is 0.15, the difference is 0.05. The larger the difference, i.e., the closer the control matrix distance is to the reference matrix distance, the more the associated control parameters are decreased by a preset adjustment step size, such as 0.05 times per adjustment. For example, the steering angle association weight of the steering actuator is reduced from 0.6 to 0.55 to reduce the impact of fusion data deviation. Conversely, the smaller the difference, i.e., the smaller the control matrix distance is to the reference matrix distance, the more the associated control parameters are increased by a preset adjustment step size. For example, the steering angle association weight is increased from 0.6 to 0.65 to enhance the effect of the fusion algorithm and achieve dynamic optimization of the associated control parameters.

[0059] If the fusion control condition is not met, i.e., the distance between the control matrix and the distance between the reference matrix are greater than or equal to the distance between the reference matrix, it is determined that there is a large deviation in the fusion control data. In order to ensure control safety and continuity, the final control data matrix is ​​abandoned, and the generated single-line control data matrix is ​​directly called to send single-line control commands to each actuator, ensuring that the actuator actions conform to the original stable control logic and avoiding control errors caused by fusion data deviation.

[0060] Through the above steps, this method achieves the organic synergy between the built-in control algorithm (old algorithm) and the fusion algorithm (new algorithm): it retains the stable output advantage of the old algorithm, improves control accuracy through multi-source data fusion of the new algorithm, and balances control effect and system reliability through matrix distance judgment and dynamic parameter adjustment, making it suitable for the high-precision actuator control needs of vehicles or robots in complex scenarios.

[0061] Within a preset time period, i.e., a pre-defined statistical cycle, this time period needs to be determined based on the actuator's control response frequency and the actual application scenario, such as setting it to 5 minutes or 100 control cycles, to ensure that the statistical sample size is sufficient to reflect data patterns and avoid single-event deviations affecting the results. Within this time period, the control matrix distance calculated using the matrix distance algorithm is recorded in real time each time, and compared one by one with a preset reference matrix distance. Valid events with control matrix distances less than the reference matrix distance are filtered out. Simultaneously, the number of occurrences of these valid events (denoted as N) and the total number of control matrix distance calculations within this time period (denoted as M) are accumulated. The percentage of occurrences within this time period is calculated using the formula: Percentage of Counts = Number of Valid Events N / Total Number of Calculations M. Preset precision is maintained during the calculation, such as retaining two decimal places, to ensure the accuracy of the percentage data.

[0062] After obtaining the percentage of occurrences within a time period, the weighting coefficients of the corresponding elements in the fused control data matrix are adjusted based on the positive correlation adjustment rule. The specific implementation is as follows:

[0063] Clarify the correspondence between each element in the fusion control data matrix; each element in the matrix corresponds to the fusion control data of a specific actuator, and its weighting coefficient is the weight value of the element when it is weighted in the matrix distance algorithm with the corresponding element of the single-line control data matrix. For example, the weighting coefficient of the element corresponding to the power actuator in the fusion control data matrix is ​​initially set to 0.6.

[0064] If the calculated percentage of occurrences is higher, such as 90%, it indicates that the deviation between the fused control data and the single-line control data is small and the reliability is high. In this case, the weighting coefficient of the corresponding element is increased by a preset adjustment range. For example, if the percentage increases by 10% each time, the weighting coefficient increases by 0.1, with an upper limit of 0.9, to strengthen the influence of the fused control data in the final control data matrix and further improve control accuracy. If the percentage of occurrences is lower, such as 30%, it indicates that the risk of deviation in the fused control data is high. In this case, the weighting coefficient of the corresponding element is decreased by a preset adjustment range. For example, if the percentage decreases by 10% each time, the weighting coefficient decreases by 0.1, with a lower limit of 0.3, to reduce the impact of fused data deviation on the final control result and ensure control stability.

[0065] The adjusted weighting coefficients are updated in the parameter library of the matrix distance algorithm. In the control loop of the next preset time period, the elements of the single-line control data matrix and the fused control data matrix are calculated based on the updated weighting coefficients to achieve dynamic iterative optimization of the weighting coefficients.

[0066] A preset statistical time period is established, which must match the actuator control cycle and the frequency of scene changes, such as 3 minutes or 50 control cycles, to ensure the sample is both timely and representative. Within this time period, after each control matrix distance calculation, the distance value is recorded and compared with the preset initial reference matrix distance. Valid distance data where the control matrix distance is less than the initial reference matrix distance are marked. The number of valid distance data (denoted as A) is accumulated, along with the total number of control matrix distance calculations within this time period (denoted as B). The percentage of calculations within this time period is calculated using the formula: Percentage of Calculations = Valid Counts A / Total Counts B. The calculation result retains a preset precision, such as one decimal place, to provide a basis for subsequent distance filtering.

[0067] Based on the frequency percentages obtained above, the number of control matrix distances to be selected is determined. If the frequency percentage is C, such as C=80%, then from all control matrix distances recorded within that time period, distance data consistent with the number of effective frequencies A are selected. That is, A effective distance data points with control matrix distances less than the initial reference matrix distance are selected to ensure the reliability of the selected data. The selected A effective distance data points are summed and then divided by the number of data points A to obtain the average distance. During the calculation, the same precision as the control matrix distances is maintained, such as retaining three decimal places, to avoid calculation errors affecting subsequent updates.

[0068] The calculated average distance is used as the new reference matrix distance, replacing the original initial reference matrix distance, and stored in the algorithm parameter storage module. The updated reference matrix distance is directly applied to the control matrix distance comparison stage in the next preset time period, realizing dynamic iteration of the reference matrix distance. If the proportion of occurrences changes in subsequent time periods, such as increasing to 90% or decreasing to 60%, the above steps are repeated. Based on the new proportion of occurrences, the corresponding number of valid distance data are selected to calculate the average value, continuously updating the reference matrix distance to ensure that it always adapts to the deviation pattern of the current control scenario and improves the accuracy of matrix distance comparison.

[0069] A preset statistical time period is established, which needs to be set in conjunction with the actuator's control response cycle and the dynamic characteristics of the actual scenario. For example, it could be set to 4 minutes or 60 control cycles to ensure that the statistical sample reflects the control deviation patterns under the current scenario. Within this time period, after each calculation of the control matrix distance using the matrix distance algorithm, the distance value is immediately recorded and compared with the currently effective preset reference matrix distance. Valid events where the control matrix distance is less than the reference matrix distance are marked. Simultaneously, the number of occurrences of valid events (denoted as X) and the total number of control matrix distance calculations within this time period (denoted as Y) are accumulated. The percentage of occurrences within this time period is calculated using the formula: Percentage of Counts = Valid Counts X / Total Counts Y. The calculation result retains a preset precision, such as two decimal places, to provide a quantitative basis for subsequent adjustment of the reference matrix distance.

[0070] Based on the frequency proportions calculated above, the distance to the reference matrix is ​​dynamically adjusted according to the inverse correlation adjustment rule, as follows:

[0071] Preset and correlate the adjustment range: Preset the basic adjustment range of the reference matrix distance, such as 5% to 10% of the initial reference matrix distance for each adjustment, and set the upper and lower limits of the adjusted distance to avoid exceeding the safe control range of the actuator. This range needs to be determined in combination with the control accuracy requirements of the actuator and historical deviation data.

[0072] If the percentage of occurrences is higher, such as 90%, it indicates that the deviation between the current fused control data and the single-line control data is small and the reliability is high. In this case, the reference matrix distance is reduced by a preset adjustment range. For example, if the initial reference matrix distance is 0.2 and the adjustment range is 10%, then the new reference matrix distance = 0.2 × (1 - 10%) = 0.18. By reducing the control threshold, the accuracy requirements of the fused control data matrix are further improved, and the control effect is enhanced. If the percentage of occurrences is lower, such as 40%, it indicates that the risk of deviation in the fused control data is higher. In this case, the reference matrix distance is increased by a preset adjustment range. For example, if the initial reference matrix distance is 0.2 and the adjustment range is 10%, then the new reference matrix distance = 0.2 × (1 + 10%) = 0.22. By increasing the control threshold, system control interruption caused by frequent triggering of single-line control logic is avoided, ensuring control continuity.

[0073] The adjusted new reference matrix distance replaces the original reference matrix distance and is stored in the algorithm's parameter storage unit. It is then directly called in the control matrix distance comparison stage of the next preset time period to achieve dynamic iterative optimization of the reference matrix distance, ensuring that each judgment fits the actual deviation of the current control scenario.

[0074] This application also discloses a multi-layer algorithm fusion system for controlling an actuator, including a processor, wherein the processor executes the steps of the multi-layer algorithm fusion method for controlling an actuator as described in any of the above embodiments.

[0075] This application also discloses a storage medium storing a program, which, when executed by a processor, implements the steps of the multi-layer algorithm fusion method for controlling the actuator described in any of the above-mentioned embodiments.

[0076] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A multi-layer algorithm fusion method for controlling actuators, characterized in that, Includes the following steps: The system acquires various sensor data based on preset sensors, assigns the various sensor data to corresponding built-in control algorithms, calculates single-line control data using the built-in control algorithms, and obtains a single-line control data matrix based on the various single-line control data. Multiple sensor data are combined to form a sensor data matrix, and a fusion control data matrix is ​​obtained by inputting the sensor data matrix into a preset fusion algorithm. The control matrix distance is calculated using a matrix distance algorithm based on the single-line control data matrix and the fused control data matrix. If the control matrix distance is less than the preset reference matrix distance, the final control data matrix is ​​calculated based on the single-line control data matrix and the fused control data matrix, and multiple actuators are controlled based on the final control data matrix. Otherwise, a single-line control data matrix is ​​used to control multiple actuators.

2. The multi-layer algorithm fusion method for controlling actuators according to claim 1, characterized in that, The fusion algorithms include: Obtain actuator type data; Based on the type of data, match the associated control parameters corresponding to the sensing data of different sensors; Based on the sensing data from different sensors, the execution control data is calculated using a weighted algorithm with associated control parameters, where each execution control data corresponds one-to-one with an actuator. Multiple execution control data are combined to form a fused control data matrix.

3. The multi-layer algorithm fusion method for controlling actuators according to claim 1, characterized in that, Matrix distance algorithms include: The difference matrix is ​​obtained by calculating the difference between corresponding elements of the single-line control data matrix and the fused control data matrix; The mean absolute error of the difference matrix is ​​calculated as the distance to the control matrix; where the mean absolute error is... ; where |dij| is the absolute value of the difference matrix element, used to eliminate the offsetting effect of positive and negative deviations and highlight the actual magnitude of the deviation; m×n is the total number of elements in the matrix, which is averaged to make the result suitable for distance comparison of matrices of different dimensions.

4. The multi-layer algorithm fusion method for controlling actuators according to claim 1, characterized in that, The step of calculating the final control data matrix based on the single-line control data matrix and the fused control data matrix also includes the following sub-steps: Associate the elements at corresponding positions in the single-line control data matrix with the elements at the corresponding positions in the fused control data matrix one by one; The final element is calculated based on the associated elements using a pre-defined weighted average algorithm. The final elements form the final control data matrix.

5. The multi-layer algorithm fusion method for controlling actuators according to claim 1, characterized in that, Within a preset time period, the percentage of times the control matrix distance is less than the preset reference matrix distance is recorded. The weighting coefficient of the corresponding element in the fused control data matrix is ​​adjusted according to the positive correlation of the percentage of times. The larger the percentage of times, the larger the weighting coefficient, and the smaller the percentage of times, the smaller the weighting coefficient.

6. The multi-layer algorithm fusion method for controlling actuators according to claim 1, characterized in that, Within a preset time period, the percentage of times the control matrix distance is less than the preset reference matrix distance is recorded. Based on the percentage of times, the average distance is calculated using the corresponding number of control matrix distances to update the reference matrix distance.

7. The multi-layer algorithm fusion method for controlling actuators according to claim 1, characterized in that, Within a preset time period, the percentage of times the control matrix distance is less than the preset reference matrix distance is recorded. The reference matrix distance is adjusted inversely based on the percentage of times: the larger the percentage of times, the smaller the reference matrix distance, and vice versa.

8. The multi-layer algorithm fusion method for controlling actuators according to claim 2, characterized in that, The steps of controlling multiple actuators based on the final control data matrix also include the following sub-steps: The correlation control parameters are adjusted based on the inverse correlation of the control matrix distance. The larger the control matrix distance, the smaller the correlation control parameters, and vice versa.

9. A multi-layer algorithm fusion system for controlling actuators, characterized in that, Includes a processor, wherein the processor performs the steps of the multi-layer algorithm fusion method for controlling the actuator as described in any one of claims 1-8.

10. A storage medium, characterized in that, The storage medium stores a program, which, when executed by a processor, implements the steps of the multi-layer algorithm fusion method for controlling the actuator as described in any one of claims 1-8.

Images