Dual-channel prediction control method and electronic device

Abstract

The application provides a double-channel prediction control method and an electronic device. The double-channel prediction control method comprises the following steps: constructing a residual sequence of a target object based on an actual state sequence and a predicted state sequence of the target object; decomposing the residual sequence to obtain an operation intention residual sequence and an environmental disturbance residual sequence; correcting a reference trajectory of the target object based on the operation intention residual sequence to obtain a corrected trajectory; correcting model parameters used for generating the predicted state sequence based on the environmental disturbance residual sequence to obtain a corrected model; and generating a control instruction for driving the target object to operate based on the corrected trajectory and the corrected model, so that the control deviation caused by the mutual confusion of the operation intention and the environmental disturbance is reduced, and the accuracy of the operation of the target object is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology, and in particular to a dual-channel predictive control method and electronic device. Background Technology

[0002] In recent years, with the development of robotics and intelligent control technologies, predictive control methods have been widely used in fields such as robotic arms, mobile robots, and intelligent manufacturing equipment because they can predict the future operating state of a target object based on its current state.

[0003] However, existing predictive control methods typically adjust control based on the overall deviation between the actual and predicted states, making it difficult to accurately determine the cause of state deviations. When state deviations caused by different factors exhibit similar characteristics, it can easily lead to a mismatch between control adjustments and the actual needs of the target object, resulting in problems such as trajectory deviation, improper parameter adjustments, or operational oscillations, thus affecting the accuracy and stability of the target object's operation. Summary of the Invention

[0004] The dual-channel predictive control method and electronic device provided in this application aim to solve the defect in existing predictive control methods where the control adjustment does not match the actual operating requirements of the target object.

[0005] In a first aspect, embodiments of this application provide a dual-channel predictive control method. This dual-channel predictive control method includes: constructing a residual sequence of the target object based on its actual state sequence and predicted state sequence; decomposing the residual sequence to obtain an operational intention residual sequence and an environmental disturbance residual sequence; correcting a reference trajectory of the target object based on the operational intention residual sequence to obtain a corrected trajectory; correcting model parameters used to generate the predicted state sequence based on the environmental disturbance residual sequence to obtain a corrected model; and generating control commands to drive the target object to operate based on the corrected trajectory and the corrected model.

[0006] In some embodiments, constructing the residual sequence of the target object based on the actual state sequence and the predicted state sequence includes: aligning the actual state sequence and the predicted state sequence in time to determine the actual state and the predicted state corresponding to each control moment; calculating the difference between the actual state value and the corresponding predicted state value for each state dimension at each control moment to obtain multiple state residuals; combining the multiple state residuals corresponding to the same control moment to obtain a multidimensional residual vector for the corresponding control moment; and arranging the multidimensional residual vector according to the order of each control moment to obtain the residual sequence of the target object.

[0007] In some embodiments, decomposing the residual sequence to obtain an operational intention residual sequence and an environmental disturbance residual sequence includes: identifying operational intention residuals in the residual sequence based on the proportion of residuals with the same sign corresponding to multiple consecutive control times, the number of control times with the same residual sign, and the second derivative of the residual change; identifying environmental disturbance residuals in the residual sequence based on the number of residual direction changes within a preset time range, the amount of residual amplitude change corresponding to adjacent control times, and whether the residual corresponding to the current control time exceeds a preset fluctuation range; and arranging the operational intention residuals and the environmental disturbance residuals according to the order of each control time to obtain the operational intention residual sequence and the environmental disturbance residual sequence.

[0008] In some embodiments, the step of correcting the reference trajectory of the target object based on the operation intention residual sequence to obtain the corrected trajectory includes: determining the correction direction of the reference trajectory based on the sign of the operation intention residual in the operation intention residual sequence; determining the correction magnitude of the reference trajectory based on the magnitude of the operation intention residual; and correcting the reference trajectory of the target object based on the correction direction and the correction magnitude to obtain the corrected trajectory.

[0009] In some embodiments, the step of correcting the model parameters used to generate the predicted state sequence based on the environmental disturbance residual sequence to obtain a corrected model includes: determining the number of sign changes of the environmental disturbance residual within a preset time range; increasing the damping parameter in the model parameters by a first preset increment when the number of sign changes reaches a preset threshold; and correcting the model parameters used to generate the predicted state sequence based on the increased damping parameter to obtain the corrected model.

[0010] In some embodiments, the step of correcting the model parameters used to generate the predicted state sequence based on the environmental disturbance residual sequence to obtain a corrected model includes: determining the number of control moments with consecutively identical signs in the environmental disturbance residuals; increasing the stiffness parameter in the model parameters by a second preset increment when the number of control moments reaches a preset threshold; and correcting the model parameters used to generate the predicted state sequence based on the increased stiffness parameter to obtain the corrected model.

[0011] In some embodiments, generating control instructions for driving the target object based on the corrected trajectory and the corrected model includes: performing predictive control calculations based on the corrected trajectory and the corrected model to obtain control instructions for driving the target object; and applying the control instructions to the actuator of the target object to control the target object to run.

[0012] In some embodiments, the dual-channel predictive control method further includes: acquiring state data from multiple sensors; aligning the state data from the multiple sensors to a unified time reference using the control cycle; detecting outliers in the time-aligned state data, and replacing the outliers with normal state data corresponding to the previous control cycle if an outlier is detected; and arranging the state data after outlier detection and replacement processing according to the time sequence of each control cycle to obtain the actual state sequence.

[0013] In some embodiments, the dual-channel predictive control method further includes: acquiring the current state of the target object in the current control cycle and the current control command output by the controller in the current control cycle; inputting the current state and the current control command into a pre-established state transition model to generate a predicted state of the target object in the next control cycle; using each generated predicted state as the input state for the next state prediction, and recursively generating the predicted state of the target object in subsequent multiple control cycles based on the current control command through the state transition model; arranging each predicted state according to the order of the control cycles corresponding to each predicted state to obtain the predicted state sequence.

[0014] Secondly, embodiments of this application provide an electronic device. The electronic device includes a processor and a memory, wherein the memory stores computer program instructions, which, when invoked by the processor, cause the processor to execute the dual-channel predictive control method as described above.

[0015] The beneficial effects of the dual-channel predictive control method provided in this application are as follows: On the one hand, by comparing the actual state sequence of the target object with the predicted state sequence, the difference between the actual state and the predicted state of the target object is obtained, and the difference is decomposed into operation intention residual and environmental disturbance residual, which can distinguish the state deviation caused by the operator's active operation and the state deviation caused by external environmental disturbance; on the other hand, the reference trajectory and model are corrected according to the operation intention residual and the environmental disturbance residual respectively, and control commands are generated based on the corrected trajectory and model, which can reduce the control deviation caused by the confusion between operation intention and environmental disturbance and improve the accuracy of the target object's operation. Attached Figure Description

[0016] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0017] Figure 1 A schematic diagram of the dual-channel predictive control method provided in the embodiments of this application; Figure 2 A schematic diagram of a sub-process of step S110 of the dual-channel predictive control method provided in the embodiments of this application; Figure 3 A schematic diagram of a sub-process of step S120 of the dual-channel predictive control method provided in the embodiments of this application; Figure 4 A schematic diagram of a sub-process of step S130 of the dual-channel predictive control method provided in the embodiments of this application; Figure 5 A schematic diagram of the sub-process of step S140 of the dual-channel predictive control method provided in the embodiments of this application. Figure 1 ; Figure 6 A schematic diagram of the sub-process of step S140 of the dual-channel predictive control method provided in the embodiments of this application. Figure 2 ; Figure 7 A schematic diagram of a sub-process of step S150 of the dual-channel predictive control method provided in the embodiments of this application; Figure 8 This is a schematic diagram of the architecture of the dual-channel predictive control device provided in the embodiments of this application; Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0019] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0020] The technical solution of this application is described in detail below.

[0021] Figure 1This is a schematic diagram of the dual-channel predictive control method provided in an embodiment of this application. Figure 1 As shown, the dual-channel predictive control method includes: S110. Based on the actual state sequence and the predicted state sequence of the target object, construct the residual sequence of the target object; Here, "target object" refers to the controlled object that requires predictive control. Target objects can be robotic arms, mobile robots, unmanned vehicles, drones, etc.

[0022] "Actual state sequence" refers to a sequence of state data formed by arranging the actual operating states of a target object in chronological order across multiple consecutive control moments. The actual state sequence can include the target object's position, velocity, attitude, acceleration, joint angles, forces, torques, etc.

[0023] "Predicted state sequence" refers to the state data sequence obtained by predicting the state of the target object at multiple consecutive control moments based on a prediction model used to describe the motion law of the target object.

[0024] The "residual sequence" refers to the sequence formed by arranging the deviations between the actual state sequence and the predicted state sequence in chronological order.

[0025] In some embodiments, the actual state sequence of the target object over multiple consecutive control cycles is obtained, and a predicted state sequence corresponding to the actual state sequence in time is generated based on the prediction model. Then, the deviation between the actual state and the predicted state at the same control moment is calculated to obtain the residual sequence of the target object. This can transform the deviation of the target object in the actual operation process into analyzable time-series data, providing a data foundation for subsequent residual decomposition.

[0026] S120. Decompose the residual sequence to obtain the operation intention residual sequence and the environmental disturbance residual sequence; Among them, the "operational intention residual sequence" refers to the residual component sequence obtained from the residual sequence, which is used to characterize the operator's continuous intervention intention.

[0027] "Environmental disturbance residual sequence" refers to the residual component sequence obtained from the residual sequence and used to characterize external environmental disturbances.

[0028] In some embodiments, residuals from different sources in the residual sequence are distinguished based on the consistency, persistence, smoothness, frequency of direction change, and amplitude abrupt changes of the residual sequence over multiple consecutive control cycles. The residuals representing the operator's continuous intervention trend are classified as the operation intention residual sequence, and the residuals representing the effects of the external environment are classified as the environmental disturbance residual sequence. This enables structured analysis of mixed residual information and provides a basis for subsequent trajectory correction and model correction.

[0029] S130. Based on the residual sequence of the operation intention, the reference trajectory of the target object is corrected to obtain the corrected trajectory; Among them, "reference trajectory" refers to the target motion path that the control system expects the target object to track.

[0030] "The corrected trajectory" refers to the trajectory obtained by adjusting the reference trajectory according to the residual sequence of the operational intent.

[0031] In some embodiments, the reference trajectory points after the current control moment of the target object are offset and corrected according to the direction and magnitude of the residual in the residual sequence of the operation intention, so as to obtain the corrected trajectory. This allows the reference trajectory to be dynamically adjusted according to the operator's continuous intervention intention, thereby improving the target following and operation compliance in the human-machine collaborative control process.

[0032] S140. Based on the environmental disturbance residual sequence, the model parameters used to generate the predicted state sequence are corrected to obtain the corrected model; Here, "model parameters" refers to the parameters in the prediction model used to generate the predicted state sequence. Model parameters may include damping parameters, stiffness parameters, inertia parameters, etc.

[0033] "The modified model" refers to the prediction model obtained by adjusting the model parameters based on the residual sequence of environmental disturbances.

[0034] In some embodiments, the model parameters used to generate the predicted state sequence are adaptively adjusted according to the number of sign changes, the duration of continuous same direction, and the magnitude changes of the residuals in the environmental disturbance residual sequence, to obtain a corrected model. This allows the prediction model to adapt to state changes caused by external environmental disturbances and improves the accuracy of subsequent state predictions.

[0035] S150. Based on the corrected trajectory and the corrected model, generate control instructions to drive the target object to run.

[0036] In some embodiments, the corrected trajectory is used as the reference input of the target object, and the corrected model is used as the model for predictive control calculation. Based on the corrected trajectory and the corrected model, the control quantity of the target object in the current control cycle is solved, and control instructions for driving the target object are generated. This allows the control instructions to simultaneously reflect the operator's target adjustment intention and the model correction results under external environmental disturbances, thereby improving the trajectory tracking accuracy and anti-interference stability of the target object during operation.

[0037] In this embodiment, on the one hand, by comparing the actual state sequence of the target object with the predicted state sequence, the difference between the actual state and the predicted state of the target object is obtained, and the difference is decomposed into operation intention residual and environmental disturbance residual, which can distinguish the state deviation caused by the operator's active operation and the state deviation caused by external environmental disturbance; on the other hand, the reference trajectory and model are corrected according to the operation intention residual and the environmental disturbance residual respectively, and control commands are generated based on the corrected trajectory and model, which can reduce the control deviation caused by the confusion between operation intention and environmental disturbance and improve the accuracy of the target object's operation.

[0038] In some embodiments, such as Figure 2 As shown, S110 specifically includes: S111. Time-align the actual state sequence and the predicted state sequence to determine the actual state and predicted state corresponding to each control moment. In some embodiments, the control time corresponding to the control cycle is used as a unified time reference. The control time corresponding to each state data in the actual state sequence and the predicted state sequence is determined respectively. Actual states with the same control time are paired with predicted states to establish a one-to-one correspondence between each control time and the corresponding actual and predicted states. By aligning the actual state sequence and the predicted state sequence in time, it can be ensured that the actual state and the predicted state participating in the residual calculation are consistent in the time dimension, avoiding the impact of different control times corresponding to the state data on the accuracy of the residual calculation results.

[0039] S112. Calculate the difference between the actual state value and the corresponding predicted state value for each state dimension at each control time to obtain multiple state residuals; Here, "state dimension" refers to the different data dimensions used to describe the operational state of the target object. State dimensions can include position dimension, velocity dimension, attitude dimension, acceleration dimension, force dimension, torque dimension, joint angle dimension, etc.

[0040] "State residual" refers to the difference between the actual state value and the corresponding predicted state value of a certain state dimension at the same control time. It is used to characterize the deviation of the actual state of the target object from the predicted state in that state dimension.

[0041] In some embodiments, for each control moment, the difference between the actual position value and the corresponding predicted position value is calculated to obtain the position residual; the difference between the actual velocity value and the corresponding predicted velocity value is calculated to obtain the velocity residual; the difference between the actual attitude value and the corresponding predicted attitude value is calculated to obtain the attitude residual; and the difference between the actual torque value and the corresponding predicted torque value is calculated to obtain the torque residual. Thus, the state residuals corresponding to multiple state dimensions at that control moment can be obtained to characterize the degree of deviation of the actual state of the target object from the predicted state from multiple dimensions such as position, velocity, attitude, and torque.

[0042] S113. Combine multiple state residuals corresponding to the same control time to obtain a multidimensional residual vector corresponding to the control time. Among them, "multidimensional residual vector" refers to a vector formed by combining the state residuals corresponding to multiple state dimensions at the same control time according to a preset dimension arrangement order.

[0043] In some embodiments, the position residual, velocity residual, attitude residual and torque residual corresponding to the same control moment are combined according to a preset dimensional arrangement order to construct a multidimensional residual vector corresponding to that control moment.

[0044] S114. Arrange the multidimensional residual vectors according to the order of each control moment to obtain the residual sequence of the target object.

[0045] Among them, "residual sequence" refers to time series data formed by arranging the multidimensional residual vectors corresponding to multiple control times in chronological order of the control times.

[0046] In some embodiments, the multidimensional residual vectors corresponding to each control time are sequentially arranged according to the order of control times to construct a residual sequence of the target object. Each sequence element in the residual sequence corresponds to a control time and includes the state residuals corresponding to multiple state dimensions at that control time.

[0047] In this embodiment, by calculating the difference between the actual state values ​​and the corresponding predicted state values ​​of multiple state dimensions at each control moment, it can be ensured that the state data involved in the comparison remain consistent in the time dimension. By combining multiple state residuals corresponding to the same control moment into a multidimensional residual vector and arranging each multidimensional residual vector according to the order of the control moments, the deviation of the target object in different state dimensions and the variation law of the residuals over time can be characterized simultaneously, thereby providing an accurate and complete data foundation for subsequent identification of operational intention residuals and environmental disturbance residuals.

[0048] In some embodiments, such as Figure 3 As shown, S120 specifically includes: S121. Based on the proportion of residuals with the same sign corresponding to multiple consecutive control times, the number of control times with the same residual sign, and the second derivative of the residual change, identify the operation intention residual in the residual sequence; Among them, "operational intention residual" refers to the residual generated by the operator's active intervention in the operation of the target object, and can characterize the operator's expectation that the target object will change its direction of movement, speed of movement, posture of movement, or torque.

[0049] In some embodiments, residuals corresponding to multiple consecutive control times are selected from the residual sequence, and the proportion of residuals with the same sign is counted. When the proportion of residuals with the same sign is greater than or equal to a preset proportion threshold, the corresponding residual is determined to satisfy the direction consistency condition; when the number of consecutive control times with the same residual sign is greater than or equal to a preset duration threshold, the corresponding residual is determined to satisfy the persistence condition; when the absolute value of the second derivative of the residual change is less than or equal to a preset smoothing threshold, the corresponding residual is determined to satisfy the smoothness condition.

[0050] Furthermore, when the residuals corresponding to multiple consecutive control moments simultaneously satisfy the directional consistency condition, the persistence condition, and the smoothness condition, the residuals corresponding to the corresponding control moments are identified as containing operational intent signals.

[0051] S122. Based on the number of residual direction changes within a preset time range, the residual amplitude change corresponding to adjacent control times, and whether the residual corresponding to the current control time exceeds a preset fluctuation range, identify the environmental disturbance residual in the residual sequence. Among them, "environmental disturbance residual" refers to the residual caused by collisions, load changes, obstacles, external impacts, or other non-operational intention factors in the environment in which the target object is located.

[0052] In some embodiments, the number of times the residual direction changes within a preset time range is counted. When the number of residual direction changes is greater than or equal to a preset threshold for the number of direction changes, the corresponding residual is determined to meet the condition of frequent direction changes; when the residual amplitude change is greater than or equal to a preset amplitude change threshold, the corresponding residual is determined to meet the condition of abrupt amplitude change; when the residual corresponding to the current control moment is greater than the upper limit of a preset fluctuation range or less than the lower limit of a preset fluctuation range, the residual corresponding to the current control moment is determined to exceed the preset fluctuation range and meet the historical deviation condition.

[0053] Furthermore, when the residual corresponding to the control time satisfies at least one of the following conditions: frequent change in direction, abrupt change in amplitude, and historical deviation, the corresponding residual is identified as a residual containing environmental disturbance.

[0054] S123. Arrange the operation intention residual and the environmental disturbance residual according to the order of each control moment to obtain the operation intention residual sequence and the environmental disturbance residual sequence.

[0055] In some embodiments, for each control time point, the operational intent residuals corresponding to each state dimension at the same control time point are combined according to a preset dimensional arrangement order to obtain the operational intent residual vector corresponding to that control time point; the environmental disturbance residuals corresponding to each state dimension at the same control time point are combined according to the same dimensional arrangement order to obtain the environmental disturbance residual vector corresponding to that control time point. The operational intent residual vectors and environmental disturbance residual vectors corresponding to each control time point are sequentially arranged according to the order of the control time points to construct the operational intent residual sequence and the environmental disturbance residual sequence.

[0056] In this embodiment, by analyzing the consistency of direction, duration and smoothness of change of residuals over multiple consecutive control moments, operational intention residuals with continuous and stable change characteristics can be identified; by analyzing the number of directional changes of residuals within a preset time range, the amplitude changes between adjacent control moments and the deviation from the preset fluctuation range, environmental disturbance residuals with frequent changes, abrupt amplitude changes or abnormal deviations can be identified.

[0057] In some embodiments, such as Figure 4 As shown, S130 specifically includes: S131. Based on the sign of the operation intention residual in the operation intention residual sequence, determine the correction direction of the reference trajectory; Among them, "correction direction of reference trajectory" refers to the direction used when offsetting trajectory points in the reference trajectory, which is used to adjust the reference trajectory toward the direction of movement desired by the operator.

[0058] In some embodiments, one or more operation intention residuals at the current control time and prior to the current control time are obtained from the operation intention residual sequence, and the correction direction of the reference trajectory is determined according to the sign of each operation intention residual. When the operation intention residual is positive, the positive direction of the corresponding state dimension is determined as the correction direction of the reference trajectory; when the operation intention residual is negative, the negative direction of the corresponding state dimension is determined as the correction direction of the reference trajectory; when the operation intention residual is zero, it is determined that there is no valid trajectory correction direction for the corresponding state dimension.

[0059] Specifically, when the target object is a robotic arm, and the intentional position residual is positive in the horizontal axis direction of the target coordinate system, it indicates that the operator expects the end effector of the robotic arm to move in the positive direction of that horizontal axis. Therefore, this direction is determined as the position correction direction of the reference trajectory. When the intentional position residual is negative in the vertical axis direction of the target coordinate system, it indicates that the operator expects the end effector of the robotic arm to move in the negative direction of that vertical axis. Therefore, this direction is determined as the position correction direction of the reference trajectory.

[0060] S132. Based on the magnitude of the residual of the operational intent, determine the correction magnitude of the reference trajectory; Among them, "correction range of reference trajectory" refers to the magnitude of the offset of trajectory points in the reference trajectory along the correction direction, which is used to characterize the degree to which the reference trajectory is adjusted according to the operator's operational intention.

[0061] In some embodiments, corresponding trajectory correction gains can be set according to the type of the operational intent residual or the state dimension to which it belongs. For example, position correction gains, velocity correction gains, and attitude correction gains can be set for position residuals, velocity residuals, and attitude residuals, respectively, and the corresponding operational intent residuals can be converted into corresponding trajectory correction magnitudes. This can reduce the impact of differences in numerical scales between different state dimensions on the trajectory correction results, thereby improving the accuracy of trajectory correction.

[0062] S133. Based on the correction direction and the correction magnitude, the reference trajectory of the target object is corrected to obtain the corrected trajectory.

[0063] In some embodiments, the trajectory correction amount is determined based on the correction direction and the correction magnitude. The correction direction characterizes the direction in which the reference trajectory deviates, and the correction magnitude characterizes the amount by which the reference trajectory deviates along the correction direction. The trajectory correction amount is applied to one or more corresponding trajectory points in the reference trajectory, causing the corresponding trajectory points to deviate along the correction direction by a distance corresponding to the correction magnitude, thereby obtaining the corrected trajectory.

[0064] For example, when the correction direction of the reference trajectory is the positive direction of the horizontal coordinate axis in the target coordinate system, and the correction magnitude is a preset distance, a value corresponding to the preset distance can be added to the horizontal position coordinate of one or more trajectory points in the reference trajectory, so that the corresponding trajectory points are offset by the corresponding distance along the positive direction of the horizontal coordinate axis, thereby obtaining the trajectory corrected along that direction.

[0065] In this embodiment, by determining the correction direction and correction magnitude of the reference trajectory based on the sign and magnitude of the residual of the operational intent, and correcting multiple trajectory points after the current control moment, the corrected trajectory can be matched with the direction and degree of movement expected by the operator, thereby improving the accuracy of the target object's response to the operational intent and the continuity of trajectory adjustment.

[0066] In some embodiments, such as Figure 5 As shown, S140 specifically includes: S211. Determine the number of sign changes of the environmental disturbance residual within a preset time range; The "preset time range" refers to the time interval formed by selecting a preset number of consecutive control cycles backward from the current control time as the endpoint, which is used to limit the analysis range of environmental disturbance residuals.

[0067] "Number of sign changes" refers to the cumulative number of times the sign of the environmental disturbance residual changes from positive to negative or from negative to positive when comparing adjacent environmental disturbance residuals in the order of control times within a preset time range. It is used to characterize the frequency of changes in the direction of the environmental disturbance residual.

[0068] In some embodiments, the environmental disturbance residuals corresponding to each control moment within a preset time range are extracted from the environmental disturbance residual sequence, and the signs of adjacent environmental disturbance residuals are compared sequentially according to the order of the control moments. When two adjacent environmental disturbance residuals are positive and negative respectively (i.e., the residual sign changes from positive to negative or vice versa), a sign change is recorded; when two adjacent environmental disturbance residuals have the same sign, no sign change is recorded. After completing the sign comparison of all adjacent environmental disturbance residuals within the preset time range, the recorded sign changes are accumulated to obtain the number of sign changes of the environmental disturbance residuals.

[0069] S212. When the number of symbol changes reaches a preset threshold, the damping parameter in the model parameters is increased by a first preset increment. Among them, the "preset number threshold" refers to the critical number of times used to determine whether the residual of environmental disturbance has undergone frequent directional changes.

[0070] "Damping parameters" refer to model parameters in the prediction model used to characterize the motion resistance or energy dissipation characteristics of the target object. When the target object is a robotic arm, damping parameters can include joint space damping parameters, Cartesian space damping parameters, end effector damping parameters, or damping parameters in the admittance model.

[0071] "First preset increment" refers to the amount of parameter that is added to the current damping parameter when the number of sign changes of the environmental disturbance residual reaches a preset threshold, in order to adjust the damping level of the prediction model.

[0072] In some embodiments, the number of sign changes of the environmental disturbance residual is compared with a preset threshold. When the number of sign changes is less than the preset threshold, the current damping parameter remains unchanged; when the number of sign changes is greater than or equal to the preset threshold, a first preset increment is added to the current damping parameter to obtain an increased damping parameter. This can enhance the damping characteristics of the target object when the environmental disturbance residual frequently changes direction, thereby suppressing oscillations caused by environmental disturbances.

[0073] S213. Based on the increased damping parameters, the model parameters used to generate the predicted state sequence are modified to obtain the modified model.

[0074] In some embodiments, the increased damping parameter is written into the corresponding parameter position in the model used to generate the predicted state sequence to replace the original damping parameter, and the model parameters related to the damping characteristics in the model are updated according to the increased damping parameter, thereby completing the correction of the model parameters and obtaining the corrected model.

[0075] In this embodiment, by statistically analyzing the number of sign changes of the residual environmental disturbance within a preset time range, it can be determined whether the target object is subject to environmental disturbances with frequent changes in direction. When the number of sign changes reaches a preset threshold, the damping parameter is increased by a first preset increment, and the model used to generate the predicted state sequence is corrected based on the increased damping parameter. This can enhance the model's ability to suppress the cyclical changes or oscillation trends caused by environmental disturbances, reduce the oscillations generated during the operation of the target object, and improve the accuracy of the predicted state and the stability of the target object's operation.

[0076] In some embodiments, such as Figure 6 As shown, S140 specifically includes: S221. Determine the number of control moments in which the signs of the environmental disturbance residuals are consecutively identical. In some embodiments, the environmental disturbance residuals corresponding to the current control time and multiple control times preceding it are extracted from the environmental disturbance residual sequence. Then, according to the chronological order of the control times, the signs of the environmental disturbance residuals corresponding to adjacent control times are sequentially determined. When the environmental disturbance residuals are all positive or all negative for multiple consecutive control times, the number of control times with consecutively identical signs is accumulated to obtain the duration of the same direction of the environmental disturbance residuals.

[0077] For example, if the signs of the environmental disturbance residuals corresponding to multiple consecutive control times are positive, positive, positive, and negative in sequence, then the number of control times with the same sign for the positive environmental disturbance residuals is three; if the signs of the environmental disturbance residuals corresponding to multiple consecutive control times are negative, negative, negative, and negative in sequence, then the number of control times with the same sign for the negative environmental disturbance residuals is four.

[0078] S222. When the number of control moments reaches a preset threshold, the stiffness parameter in the model parameters is increased by a second preset increment. Among them, the "preset quantity threshold" refers to the quantity critical value used to determine whether the residual of environmental disturbance continues to exist in the same direction.

[0079] "Stiffness parameter" refers to the model parameter in the prediction model used to characterize the target object's ability to resist displacement or attitude deviation.

[0080] "Second preset increment" refers to the amount of parameter added to the current stiffness parameter when the number of consecutive control moments with the same symbol reaches a preset threshold, in order to improve the prediction model's ability to resist deviations from continuous environmental disturbances.

[0081] In some embodiments, the number of control moments with consecutive identical signs is compared with a preset threshold. When the number of control moments is less than the preset threshold, the current stiffness parameter remains unchanged; when the number of control moments is greater than or equal to the preset threshold, a second preset increment is added to the current stiffness parameter to obtain an increased stiffness parameter.

[0082] By using the above methods, when the environmental disturbance residuals persist in the same direction, the stiffness characteristics of the corresponding state dimension of the target object can be enhanced, enabling the prediction model to more accurately characterize the target object's resistance to deviation under continuous environmental influence, thereby reducing the possibility of the target object undergoing continuous deviation under continuous external disturbances.

[0083] S223. Based on the increased stiffness parameter, the model parameters used to generate the predicted state sequence are modified to obtain the modified model.

[0084] In some embodiments, the increased stiffness parameter is written into the corresponding parameter position in the model used to generate the predicted state sequence to replace the original stiffness parameter, and the model parameters related to stiffness characteristics in the model are updated according to the increased stiffness parameter, thereby completing the correction of the model parameters and obtaining the corrected model.

[0085] In this embodiment, by maintaining the same number of control moments with the same sign for the statistical environmental disturbance residuals, it can be determined whether the target object is subjected to a continuous environmental disturbance acting in the same direction. When the number of control moments reaches a preset threshold, the stiffness parameter is increased according to a second preset increment, and the model used to generate the predicted state sequence is corrected based on the increased stiffness parameter. This can enhance the model's ability to constrain state deviations caused by continuous environmental disturbances, reduce the degree of continuous offset of the target object in the same direction, and improve the accuracy of the predicted state and the stability of the target object's operation.

[0086] In some embodiments, such as Figure 7 As shown, S150 specifically includes: S151. Based on the corrected trajectory and the corrected model, predictive control calculations are performed to obtain control commands for driving the target object to run; In some embodiments, the controller uses the current state of the target object as the prediction starting point, uses the corrected trajectory as the reference trajectory in the predictive control process, and predicts the state of the target object at multiple future control moments based on the corrected model. Then, based on the deviation between each predicted state and the corresponding trajectory point in the corrected trajectory, and under the condition of satisfying preset control constraints, the controller optimizes and solves the control input in the preset prediction time domain to obtain a control input sequence, and determines the control input in the control input sequence corresponding to the current control moment as the control command used to drive the target object to run.

[0087] S152. Apply the control command to the actuator of the target object to control the operation of the target object.

[0088] In some embodiments, control commands are sent to the actuator of the target object, and the actuator drives the target object to run according to the control commands. For example, when the control command is a joint torque command, the actuator outputs the corresponding joint torque according to the joint torque command; when the control command is a speed command, the actuator adjusts the movement speed of the corresponding component of the target object according to the speed command; when the control command is a position command, the actuator drives the target object to move to the corresponding target position according to the position command.

[0089] In this embodiment, by performing predictive control calculations based on the corrected trajectory and the corrected model, the controller can generate control commands based on the dynamic characteristic changes caused by the operator's desired motion target and environmental disturbances. By applying the control commands to the actuators of the target object, the target object can be controlled to run according to the corrected trajectory, and the impact of environmental disturbances on the operation process can be reduced, thereby improving the matching degree between the control commands and the current control requirements, as well as the accuracy and stability of the target object's operation.

[0090] In some embodiments, the dual-channel predictive control method further includes: first, acquiring state data from multiple sensors; second, using the control cycle as a unified time reference, aligning the state data from the multiple sensors in time; then, performing outlier detection on the time-aligned state data, and replacing the outlier with normal state data corresponding to the previous control cycle if an outlier is detected; finally, arranging the state data after outlier detection and replacement processing according to the time sequence of each control cycle to obtain the actual state sequence.

[0091] "State data" refers to data acquired by sensors that characterizes the operating state of a target object at the corresponding acquisition time. State data can include position data, velocity data, acceleration data, attitude data, force data, torque data, etc.

[0092] In some embodiments, state data collected by position sensors, velocity sensors, attitude sensors, and torque sensors are received, and the acquisition time corresponding to each state data is recorded. Using the control time corresponding to each control cycle as a unified time reference, state data with the same or closest acquisition time to the control time are matched to the corresponding control time, so that the position data, velocity data, attitude data, and torque data corresponding to the same control time are combined to form a set of multi-dimensional state data.

[0093] After time alignment is completed, outlier detection is performed on each state data based on the preset normal value range and the amount of data change between adjacent control cycles. When the state data is within the preset normal value range and the amount of change relative to the corresponding state data in the previous control cycle is less than or equal to the preset change threshold, the state data is determined to be normal state data; when the state data exceeds the preset normal value range, or the amount of change relative to the corresponding state data in the previous control cycle is greater than the preset change threshold, the state data is determined to be an outlier.

[0094] Upon detecting an outlier, the system retrieves normal state data from the previous control cycle that belongs to the same state dimension as the outlier and replaces the outlier with the retrieved normal state data. If the state data corresponding to the previous control cycle is also an outlier, the system continues to search for the most recent normal state data in ascending order of control time and replaces it with the retrieved normal state data.

[0095] Subsequently, the state data after time alignment, outlier detection, and outlier replacement are arranged according to the order of control moments, so that each control moment corresponds to a set of actual state data, thereby constructing an actual state sequence.

[0096] In this embodiment, by performing outlier detection on the time-aligned state data and replacing the outliers with the normal state data corresponding to the previous control cycle, the impact of sensor failures, communication anomalies, or transient noise on the state data can be reduced. By arranging the processed state data according to the time sequence of the control cycle, a time-consistent and reliable actual state sequence can be obtained, thereby improving the accuracy of subsequent state comparisons and residual calculations.

[0097] In some embodiments, the dual-channel predictive control method further includes: first, acquiring the current state of the target object in the current control cycle and the current control command output by the controller in the current control cycle; second, inputting the current state and the current control command into a pre-established state transition model to generate a predicted state of the target object in the next control cycle; then, using each generated predicted state as the input state for the next state prediction, and recursively generating predicted states of the target object in subsequent multiple control cycles based on the current control command through the state transition model; finally, arranging each predicted state according to the order of the control cycles corresponding to each predicted state to obtain the predicted state sequence.

[0098] Here, "current state" refers to the operating state of the target object in the current control cycle. The current state can include current position, current velocity, current attitude, current torque, etc.

[0099] "Current control command" refers to the control input output by the controller in the current control cycle and used to drive the target object. Current control commands can include position commands, speed commands, force commands, torque commands, etc.

[0100] A "state transition model" is a model used to describe how a target object transitions from the state corresponding to one control cycle to the state corresponding to the next control cycle under the action of control commands.

[0101] In some embodiments, the current state data of the target object, such as position, velocity, attitude, and torque, in the current control cycle, as well as the current control command output by the controller in the current control cycle, are acquired. The current state and the current control command are input into a pre-established state transition model, and the state transition result of the target object under the action of the current control command is calculated through the state transition model to obtain the predicted state for the next control cycle.

[0102] For example, when the target object is a robotic arm, the current state data of each joint of the robotic arm, such as its current position, current velocity, and current posture, as well as the current joint torque command output by the controller, can be obtained. This current state data and current joint torque command are then input into a state transition model to predict the position and velocity of each joint in the next control cycle. Furthermore, based on the predicted state of each joint, the predicted position and predicted posture of the robotic arm's end effector in the next control cycle can be further determined.

[0103] After obtaining the predicted state corresponding to the next control cycle, this predicted state is used as the input state for the next state prediction, and the current control command is used as the control input for the next state prediction. Both are input into the state transition model to generate the predicted state for the next control cycle after that. Following this method, each generated predicted state is used as the input state for subsequent state predictions, and the current control command is continuously used for recursive deduction until a preset number of predicted states are obtained. The recursively generated predicted states are then arranged in the order of the control cycles corresponding to each predicted state to obtain a predicted state sequence.

[0104] In this embodiment, by inputting the current state of the target object and the current control command into the state transition model, the predicted state of the target object in the next control cycle can be obtained. By using each generated predicted state as the input state for the next state prediction and continuously performing recursive deduction based on the current control command, the predicted state of the target object in multiple subsequent control cycles can be obtained continuously, thereby forming a predicted state sequence that can reflect the future operating trend of the target object, providing a data foundation for subsequent residual calculation and predictive control.

[0105] Figure 8 This is a schematic diagram of the architecture of the dual-channel predictive control device provided in an embodiment of this application. Figure 8 As shown, the dual-channel prediction control device 200 includes: a construction module 210, a decomposition module 220, a first correction module 230, a second correction module 240, and a first generation module 250, wherein: Construction module 210 is used to construct the residual sequence of the target object based on the actual state sequence and the predicted state sequence of the target object; The decomposition module 220 is used to decompose the residual sequence to obtain the operation intention residual sequence and the environmental disturbance residual sequence; The first correction module 230 is used to correct the reference trajectory of the target object based on the residual sequence of the operation intention, so as to obtain the corrected trajectory; The second correction module 240 is used to correct the model parameters used to generate the predicted state sequence based on the environmental disturbance residual sequence, so as to obtain the corrected model. The first generation module 250 is used to generate control instructions for driving the target object to run based on the corrected trajectory and the corrected model.

[0106] In some embodiments, the construction module 210 is specifically used to: time-align the actual state sequence and the predicted state sequence to determine the actual state and predicted state corresponding to each control moment; calculate the difference between the actual state value and the corresponding predicted state value for each state dimension at each control moment to obtain multiple state residuals; combine the multiple state residuals corresponding to the same control moment to obtain a multidimensional residual vector for the corresponding control moment; and arrange the multidimensional residual vector according to the order of each control moment to obtain the residual sequence of the target object.

[0107] In some embodiments, the decomposition module 220 is specifically used to: identify the operational intention residual in the residual sequence based on the proportion of residuals with the same sign corresponding to multiple consecutive control moments, the number of control moments with the same residual sign, and the second derivative of the residual change; identify the environmental disturbance residual in the residual sequence based on the number of residual direction changes within a preset time range, the amount of residual amplitude change corresponding to adjacent control moments, and whether the residual corresponding to the current control moment exceeds a preset fluctuation range; and arrange the operational intention residual and the environmental disturbance residual according to the order of each control moment to obtain the operational intention residual sequence and the environmental disturbance residual sequence.

[0108] In some embodiments, the first correction module 230 is specifically configured to: determine the correction direction of the reference trajectory based on the sign of the operation intention residual in the operation intention residual sequence; determine the correction magnitude of the reference trajectory based on the magnitude of the operation intention residual; and correct the reference trajectory of the target object based on the correction direction and the correction magnitude to obtain the corrected trajectory.

[0109] In some embodiments, the second correction module 240 is specifically used to: determine the number of sign changes of the environmental disturbance residual within a preset time range; when the number of sign changes reaches a preset threshold, increase the damping parameter in the model parameters by a first preset increment; and based on the increased damping parameter, correct the model parameters used to generate the predicted state sequence to obtain the corrected model.

[0110] In some embodiments, the second correction module 240 is specifically used to: determine the number of control times in which the signs of the environmental disturbance residuals are consecutively the same; when the number of control times reaches a preset threshold, increase the stiffness parameter in the model parameters according to a second preset increment; and based on the increased stiffness parameter, correct the model parameters used to generate the predicted state sequence to obtain the corrected model.

[0111] In some embodiments, the first generation module 250 is specifically used to: perform predictive control calculations based on the corrected trajectory and the corrected model to obtain control instructions for driving the target object to run; and apply the control instructions to the actuator of the target object to control the target object to run.

[0112] In some embodiments, the dual-channel predictive control device 200 further includes: an acquisition module, an alignment module, a detection module, and a first arrangement module, wherein: the acquisition module is used to acquire state data from multiple sensors; the alignment module is used to time-align the state data of the multiple sensors using the control cycle as a unified time reference; the detection module is used to detect outliers in the time-aligned state data, and if an outlier is detected, to replace the outlier with normal state data corresponding to the previous control cycle; and the first arrangement module is used to arrange the state data after outlier detection and replacement processing according to the time sequence of each control cycle to obtain the actual state sequence.

[0113] In some embodiments, the dual-channel predictive control device 200 further includes: an acquisition module, a second generation module, a third generation module, and a second arrangement module, wherein: the acquisition module is used to acquire the current state of the target object in the current control cycle and the current control command output by the controller in the current control cycle; the second generation module is used to input the current state and the current control command into a pre-established state transition model to generate a predicted state of the target object in the next control cycle; the third generation module is used to use each generated predicted state as the input state for the next state prediction, and recursively generate the predicted state of the target object in subsequent multiple control cycles based on the current control command through the state transition model; the second arrangement module is used to arrange each predicted state according to the order of the control cycles corresponding to each predicted state to obtain the predicted state sequence.

[0114] Each functional module in the dual-channel predictive control device described in the foregoing one or more embodiments can be logically regarded as a virtual functional unit undertaking a specific processing task. Its function can be implemented by software program instructions, processor execution logic or dedicated hardware circuit. The specific implementation method can be flexibly selected according to the actual deployment requirements, and is not specifically limited here.

[0115] Figure 9 The diagram shows the structure of an electronic device according to an embodiment of this application. This embodiment does not limit the specific implementation of the electronic device.

[0116] like Figure 9As shown, the electronic device 400 may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440.

[0117] The processor 410, communication interface 420, and memory 430 communicate with each other via communication bus 440. Communication interface 420 is used for communication connections with other external devices. The processor 410 executes program 450 to implement the methods described in one or more of the above embodiments.

[0118] Specifically, program 450 may include program code that includes computer operation instructions. When program 450 is invoked, processor 410 executes the computer operation instructions to implement the steps in one or more embodiments.

[0119] Depending on the actual application scenario, the processor 410 can be of the appropriate type, including but not limited to mainstream embedded processors such as microcontrollers (MCUs) and ARM architecture processors, as well as other types of processors such as digital residual processors (DSPs), application-specific integrated circuits (ASICs), and field-programmable gate arrays (FPGAs), as long as they can provide the computing and control capabilities required by the actual application scenario.

[0120] The memory 430 is used to store the program 450. It includes a program storage area and a data storage area. The program storage area is used to store firmware programs, embedded applications, and various functional modules; the data storage area is used to store data and calculation results during program execution. For example, the memory 430 may include: on-chip RAM (for temporary data storage during program execution); on-chip Flash memory (for storing program code and configuration data); and EEPROM or other types of non-volatile memory (for storing parameters that need to be retained when power is off).

[0121] This application also provides a computer-readable storage medium. This computer-readable storage medium can be a non-volatile computer-readable storage medium. This computer-readable storage medium stores a computer program.

[0122] When executed by a processor, the computer program implements one or more steps of the method disclosed in the embodiments of this application. A complete computer program product is embodied on one or more computer-readable storage media containing the computer program disclosed in the embodiments of this application.

[0123] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Under the concept of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of this application as described above. For the sake of brevity, they are not provided in detail. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A dual-channel predictive control method, characterized in that, The method includes: Based on the actual state sequence and the predicted state sequence of the target object, construct the residual sequence of the target object; The residual sequence is decomposed to obtain the operation intention residual sequence and the environmental disturbance residual sequence; Based on the residual sequence of the operational intent, the reference trajectory of the target object is corrected to obtain the corrected trajectory. Based on the environmental disturbance residual sequence, the model parameters used to generate the predicted state sequence are corrected to obtain the corrected model; Based on the corrected trajectory and the corrected model, control instructions are generated to drive the target object to run.

2. The method according to claim 1, characterized in that, The construction of the residual sequence of the target object based on the actual state sequence and the predicted state sequence of the target object includes: The actual state sequence and the predicted state sequence are time-aligned to determine the actual state and predicted state corresponding to each control moment. The difference between the actual state value and the corresponding predicted state value for each state dimension at each control time is calculated to obtain multiple state residuals; By combining multiple state residuals corresponding to the same control time, a multidimensional residual vector corresponding to the control time is obtained; The multidimensional residual vectors are arranged in chronological order according to each control moment to obtain the residual sequence of the target object.

3. The method according to claim 1, characterized in that, The decomposition of the residual sequence to obtain the operational intent residual sequence and the environmental disturbance residual sequence includes: Based on the proportion of residuals with the same sign corresponding to multiple consecutive control moments, the number of control moments with the same residual sign, and the second derivative of the residual change, the operational intention residuals in the residual sequence are identified. Based on the number of residual direction changes within a preset time range, the amount of residual amplitude change corresponding to adjacent control times, and whether the residual corresponding to the current control time exceeds a preset fluctuation range, the environmental disturbance residual in the residual sequence is identified. The operation intention residual and the environmental disturbance residual are arranged in chronological order according to each control moment to obtain the operation intention residual sequence and the environmental disturbance residual sequence.

4. The method according to claim 1, characterized in that, The step of correcting the reference trajectory of the target object based on the residual sequence of the operational intent to obtain the corrected trajectory includes: Based on the sign of the operation intention residual in the operation intention residual sequence, the correction direction of the reference trajectory is determined; Based on the magnitude of the residual of the operational intent, the correction magnitude of the reference trajectory is determined; Based on the correction direction and the correction magnitude, the reference trajectory of the target object is corrected to obtain the corrected trajectory.

5. The method according to claim 1, characterized in that, The step of correcting the model parameters used to generate the predicted state sequence based on the environmental disturbance residual sequence to obtain the corrected model includes: Determine the number of sign changes of the environmental disturbance residual within a preset time range; When the number of symbol changes reaches a preset threshold, the damping parameter in the model parameters is increased by a first preset increment. Based on the increased damping parameters, the model parameters used to generate the predicted state sequence are modified to obtain the modified model.

6. The method according to claim 1, characterized in that, The step of correcting the model parameters used to generate the predicted state sequence based on the environmental disturbance residual sequence to obtain the corrected model includes: Determine the number of control moments in which the sign of the environmental disturbance residual is consecutively identical; When the number of control moments reaches a preset threshold, the stiffness parameter in the model parameters is increased by a second preset increment. Based on the increased stiffness parameter, the model parameters used to generate the predicted state sequence are modified to obtain the modified model.

7. The method according to claim 1, characterized in that, The step of generating control instructions for driving the target object based on the corrected trajectory and the corrected model includes: Based on the corrected trajectory and the corrected model, predictive control calculations are performed to obtain control commands for driving the target object to run; The control instructions are applied to the actuator of the target object to control the operation of the target object.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: Collect status data from multiple sensors; Using the control cycle as a unified time reference, the state data of the multiple sensors are time-aligned. Anomaly detection is performed on the time-aligned state data, and if an anomaly is detected, the normal state data corresponding to the previous control cycle is used to replace the anomaly. The actual state sequence is obtained by arranging the state data after outlier detection and replacement processing according to the time sequence of each control cycle.

9. The method according to any one of claims 1 to 7, characterized in that, The method further includes: Obtain the current state of the target object in the current control cycle and the current control command output by the controller in the current control cycle; The current state and the current control command are input into a pre-established state transition model to generate the predicted state of the target object in the next control cycle; Each generated predicted state is used as the input state for the next state prediction, and based on the current control command, the predicted state of the target object in the subsequent multiple control cycles is recursively generated through the state transition model. The predicted states are arranged in the order of the control cycles corresponding to each predicted state to obtain the predicted state sequence.

10. An electronic device, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when invoked by the processor, cause the processor to execute the dual-channel predictive control method as described in any one of claims 1-9.

Images