Mechanical arm control method and system based on action prediction and delay compensation
By establishing a mathematical model and a signal delay monitoring mechanism, combined with closed-loop control using pseudo-feedforward and Smith predictors, the problems of transmission delay and disturbance signal lag in traditional robotic arm control systems were solved, achieving improved stability and positioning accuracy in high-precision and high-dynamic environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-11
- Publication Date
- 2026-04-14
AI Technical Summary
Traditional robotic arm control systems struggle to meet the requirements for control stability and positioning accuracy in high-precision and high-dynamic scenarios when faced with problems such as transmission delay superposition and disturbance signal lag. In particular, they exhibit system instability and error deviation in complex working environments.
By establishing a mathematical model of the controlled robotic arm and the disturbance channel, setting a signal transmission delay monitoring mechanism, constructing a total lag model and a prediction model, obtaining a pseudo-feedforward compensation transfer function, and combining it with a Smith predictor for error calculation and feedback control, the model parameters are corrected in real time to form a closed-loop control structure, and the signal transmission architecture is optimized to reduce the impact of delay and disturbance.
It significantly improves the control stability and positioning accuracy of the robotic arm in high-precision scenarios, enhances the system's adaptability to time-varying parameters and complex disturbances, reduces the risk of failure, and adapts to the control requirements of robotic arms in high-precision and high-dynamic scenarios.
Smart Images
Figure CN121848403A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robotic arm motion planning and control technology, specifically to a robotic arm control method and system based on motion prediction and delay compensation. Background Technology
[0002] Currently, robotic arms are increasingly used in precision industrial assembly, minimally invasive medical surgery, and collaborative services, placing extremely high demands on their control precision, response speed, and operational safety. Traditional robotic arm control systems rely on real-time sensing and current state to generate control commands. However, in actual control processes, delays are unavoidable in sensor perception, data transmission, and actuator response. This delay leads to "lag" in control commands, causing the robotic arm's trajectory to deviate from expectations, resulting in overshoot or oscillation. In high-precision operation scenarios (such as micron-level part docking) or human-robot collaboration scenarios, this can easily lead to insufficient precision, low efficiency, or even safety accidents.
[0003] Complex operating environments in modern manufacturing, medical fields, and other areas (such as dynamic production lines) require robotic arms to possess stronger environmental adaptability and delay resistance. Traditional control methods (such as PID control and position impedance control) exhibit significant limitations in handling these scenarios because they do not compensate for delays. Robotic arm control technology based on motion prediction and delay compensation can effectively offset the effects of delays by predicting the delayed target motion state in advance and compensating for and correcting control commands. This meets the control requirements of robotic arms in high-precision and high-dynamic scenarios, becoming a key technological direction for breaking through traditional control bottlenecks and promoting the intelligent application of robotic arms.
[0004] The movement of a robotic arm involves multiple physical states and has significant latency. Communication latency compensation is necessary to ensure the safety and efficiency of task completion. Motion prediction and latency compensation technology can predict the next state of the robotic arm in real time and perform communication compensation to improve its operational efficiency and safety. This technology can reduce the risk of failure and adapt to high-precision, high-risk tasks.
[0005] Delay compensation is the core of precise and stable control of robotic arms. It can correct delay deviations in the sensing, communication and execution links, and avoid inaccurate movements or system instability. It is the key to its adaptation to high-precision and high-dynamic scenarios.
[0006] Robotic arm control methods based on motion prediction and delay compensation have been widely applied in various fields due to their adaptability to dynamic scenarios and ability to eliminate control delays. In the industrial sector, it supports high-precision automated production. For example, in automotive welding production lines, the robotic arm predicts the dynamic position of the car body on the conveyor belt and compensates for the delays of visual sensors and actuators, enabling precise welding of the high-speed moving car body and avoiding welding misalignment problems caused by lag in traditional control. In the aerospace field, it assists in the execution of complex tasks. For instance, robotic arms for maintenance inside spacecraft can predict minute attitude changes of the spacecraft in orbit and compensate for long-distance transmission delays of ground control commands, enabling precise grasping and repair of satellite components and ensuring the smooth progress of space missions.
[0007] There are three main challenges in robotic arm trajectory tracking control:
[0008] 1) How to accurately establish the nonlinear coupled dynamics model of the robotic arm to provide a reliable basis for prediction and compensation.
[0009] 2) How to design a computationally efficient real-time prediction algorithm to ensure that state calculation is completed within milliseconds without introducing new delays.
[0010] 3) How to maintain the robust stability of the system under model errors and external disturbances, and ensure the reliability of the prediction compensation mechanism under various working conditions.
[0011] In existing technologies, Smith Predictor is used to compensate for results, such as Figure 1 As shown, the system delay includes control delay and sensor delay, and the closed-loop transfer function is: ; In the formula, This is the actual output signal of the system. The input signal is used as a system reference. For the controller's transfer function, The transfer function for the actual controlled object. To control the delay time, For sensor delay time, For the Laplace operator, It is a natural constant.
[0012] Taking a first-order system as an example, the transfer function of the natural constant transfer function is: ; Using the PID algorithm, the transfer function is: ; in, , , These are the proportional gain, derivative time constant, and integral time constant, respectively. The controller parameters are selected as follows: , , To reduce the impact of latency, a Smith estimator is introduced into the system. The control block diagram is as follows: Figure 2 As shown in the figure The total system delay, . The predicted compensation transfer function is introduced. To predict the transfer function of the compensation model, To estimate the delay time of the compensation stage, the closed-loop transfer function of this structure is: ; Observation reveals that, compared to the system without delay, the transfer function of this system is simply multiplied by an additional factor. This is equivalent to a closed-loop system without any lag terms, and lag elements do not affect the system's characteristic equation. In other words, compared to a system without delay, the system after Smith predictive compensation simply postpones the control process. .
[0013] Based on the Smith predictor architecture, this approach uses a nominal dynamics model and the current system state to predict future outputs, and uses feedforward compensation to offset tracking errors caused by time delays. This scheme effectively improves compensation for fixed delays, enhancing system stability and tracking accuracy. However, its performance is highly dependent on model accuracy, and its effectiveness decreases when faced with model mismatch, time-varying delays, and external disturbances. It needs to be combined with adaptive or robust control to enhance its adaptability to uncertainties. The main drawbacks of this technical solution are: 1) No adaptive correction mechanism for model mismatch: It is highly dependent on the dynamic model of the robotic arm (parameters such as inertia and damping). When the load changes or the equipment ages and causes the model parameters to drift, there is no online correction strategy, the compensation effect drops sharply, and it may even lead to system oscillation.
[0014] 2) Error calculation fails to offset the effects of lag: The traditional error logic of "target value - actual output" is used only. The controller cannot predict the deviation caused by lag and correct the action lag. The optimization of system overshoot and steady-state time is limited.
[0015] In current feedforward predictive control, the core principle is to calculate the required control input in advance based on the dynamic model of the controlled object and the future reference trajectory (or external disturbance prediction), and then combine this with appropriate feedback correction to improve tracking accuracy and response speed even in the presence of time delays or disturbances. A common feedforward predictive control law can be expressed as: ; in, This represents the feedforward predictive control law. This is the feedforward control quantity calculated in advance based on the prediction model and the delay amount; This represents the current state of the system. For the desired trajectory; This is the feedback gain matrix, used to compensate for model uncertainties and external disturbances. For delay amount, It is a time variable.
[0016] This method predicts future system behavior, aligning the control input to the desired trajectory at execution time, thereby effectively compensating for system latency and improving control performance. However, it has the following drawbacks: 1) Disconnect between disturbance signal transmission and compensation: Although the feedforward control is designed based on disturbance prediction, the transmission link of the disturbance signal is not optimized (such as not using a real-time bus). The signal acquisition and transmission delay is too long, which causes the predicted disturbance information to be out of sync with the actual disturbance. The feedforward compensation amount cannot accurately offset the impact of real-time disturbance.
[0017] 2) Insufficient robustness of model mismatch: The effect of feedforward compensation is highly dependent on the accuracy of the controlled object model. When load changes or equipment aging causes model mismatch, there is no adaptive correction mechanism. It is difficult to make up for the deviation by relying solely on feedback control, and the system stability decreases.
[0018] Specifically, in traditional robotic arm signal transmission schemes, a series of performance bottlenecks arise due to the mismatch between the transmission link design and the requirements of the improved Smith predictor algorithm: First, traditional transmission often uses conventional buses such as RS485, whose signal transmission delay can reach 10-20ms. Meanwhile, the robotic arm joint actuators (such as harmonic reducers) themselves have an elastic hysteresis of 5-8ms. The combined hysteresis results in a total hysteresis that prevents the Smith predictor's hysteresis compensation stage from accurately matching the actual control link, leading to excessive deviations in the robotic arm's end-effector trajectory tracking and making it difficult to meet the positioning accuracy requirements of precision assembly, parts grinding, and other scenarios. Second, common measurable disturbances in robotic arm operations (such as fluctuations in the weight of grasped parts, visual recognition...) (Workpiece position deviation, etc.) The signal needs to be collected by the sensor and forwarded by multiple nodes before it can be transmitted to the controller. In the traditional solution, the transmission delay often exceeds 15ms, which causes the pseudo feedforward module to be unable to obtain disturbance information in time and generate advance compensation. The disturbance suppression capability is greatly reduced, and the end effector is prone to motion deviation due to load changes or workpiece deviation. In addition, the signals of each joint of the multi-axis robot arm lack an effective synchronization mechanism in the traditional transmission architecture. The signal transmission time difference between different joints can reach ±3ms. This synchronization difference will amplify the error of the Smith predictor in multi-joint collaborative control, further reducing the overall motion accuracy of the robot arm and making it unsuitable for the high-precision operation requirements of ±0.03mm in scenarios such as 3C electronics manufacturing. Summary of the Invention
[0019] The technical problem to be solved by this invention is to solve the above-mentioned problems of transmission delay superposition and disturbance signal lag, and improve the control stability and positioning accuracy of the robotic arm in high-precision scenarios.
[0020] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A robotic arm control method based on motion prediction and delay compensation includes: S10, Establish a mathematical model of the controlled robotic arm and the disturbance channel; S20, based on the mathematical model established in S10, sets up a signal transmission delay monitoring mechanism; S30, based on the signal transmission delay monitoring mechanism, constructs a total lag model and a prediction model; S40: Based on the mathematical model established in S10 and the prediction model in S30, obtain the pseudo-feedforward compensation transfer function and set the calculation logic to be converted to the time domain in order to obtain the feedforward control quantity. S50, set up the Smith prediction error calculation module to obtain signal error and feedback control quantity; S60 superimposes the feedforward control quantity and the feedback control quantity to obtain the total control quantity, and outputs the total control quantity after limiting the amplitude. S70 compares the deviation between the actual output of the robotic arm control system and the output of the prediction model in real time. When the deviation exceeds the threshold, the parameters of the prediction model are corrected online. S80 feeds back the actual output of the actuator to the Smith prediction error calculation module in real time, forming a closed control structure and synchronously monitoring whether the effect meets the standard.
[0021] In this embodiment, a signal transmission delay monitoring mechanism is set up, including: Collect the following key timestamps from both the master and slave stations: the time the master station issues the command. The time when the main station receives instructions Feedback time from the station , Main site receives feedback time ; Calculate uplink one-way delay Downlink one-way delay : , ; By taking into account the bus symmetry and eliminating interference from slave processing time, the total transmission delay can be obtained. : ; Total transmission delay Combined with execution lag time Get the total latency : .
[0022] In this embodiment, the total lag model and the prediction model are constructed, including: The formula for obtaining the total delay time is used as the total lag model; The controlled robotic arm is used to describe the hysteresis-free dynamic transfer function. Using the prototype, the estimated lag time is taken. Total delay time Introducing a pure time delay element Obtain the prediction model : ; For the Laplace operator, It is a natural constant.
[0023] In this embodiment, the pseudo-feedforward compensation transfer function is expressed by the following formula: ; For pseudo-feedforward compensation transfer function, For external disturbance channel transfer function, For the transfer function of the PD controller, This is the total delay time. For the Laplace operator.
[0024] In this embodiment, obtaining the feedforward control quantity includes: The sensor collects disturbance electrical signals and converts them into physical quantities. ; The physical quantity that converts the disturbance electrical signal. The value obtained during sampling is the AD conversion value. To disturb the physical range, For AD conversion range, For continuous time variables; The disturbance is smoothed by first-order filtering: ; This is the current filtered disturbance electrical signal. The filtered value from the previous time step. These are the filter coefficients; Substituting into the difference equation, calculate the feedforward control quantity at the current time: ; This is the feedforward control quantity at the current moment. For the dynamic attenuation coefficient, This refers to the static gain coefficient. The actuator range is limited, and the final output is a feedforward control quantity; ; It serves as the feedforward control quantity for the actuator and as the feedforward compensation quantity. The minimum / maximum compensation amount allowed by the actuator.
[0025] In this embodiment, the Smith prediction error calculation module is represented by the following formula: ; The feedback control value is obtained using the following formula: ; In the formula, To estimate the signal error between the model's output and the hysteresis-free predicted output, The target value for the control command of the robotic arm. This is the actual output of the robotic arm control system. This is the amount of delayed compensation. To predict the model's output, This is a forecast output without lag. For continuous time variables, For feedback control, This is the proportionality coefficient. is the differential coefficient.
[0026] In this embodiment, the total control quantity is output after being limited, and expressed by the following formula: ; In the formula, This is the total control input signal after amplitude limiting. For feedback control, This is the feedforward control quantity. It is a saturation or limiting function.
[0027] In this embodiment, online correction of the parameters of the prediction model includes gain correction and time constant correction: The gain correction is as follows: ; The time constant is corrected to: ; In the formula, This is the corrected gain. The gain before correction. The deviation between the actual output of the robotic arm control system and the predicted model output. The corrected time constant. The time constant before correction is , , These are the correction step sizes used for gain correction and time constant correction, respectively.
[0028] In this embodiment, whether the monitoring effect meets the standard includes synchronously monitoring disturbance suppression performance indicators during closed-loop operation: suppression speed and steady-state error; when the suppression performance indicators are found to be below the predetermined standard, the delay monitoring parameters or compensation parameters are adjusted retrospectively.
[0029] This invention also provides a robotic arm control system based on motion prediction and delay compensation, which applies the above-described robotic arm control method based on motion prediction and delay compensation, including: Passive object and disturbance module, used to establish mathematical models of the controlled robotic arm and disturbance channels; The signal monitoring module is used to set up a signal transmission delay monitoring mechanism based on the mathematical model established in the passive object and disturbance module. The lag and prediction module is used to construct the total lag model and prediction model based on the signal transmission delay monitoring mechanism. The feedforward control module is used to obtain the pseudo-feedforward compensation transfer function and set the calculation logic to convert it to the time domain based on the mathematical model established in the passive object and disturbance module and the prediction model in the lag and prediction module, so as to obtain the feedforward control quantity. The feedback control module is used to set up the Smith prediction error calculation module and obtain signal error and feedback control quantity. The main control module is used to superimpose the feedforward control quantity and the feedback control quantity to obtain the total control quantity, and then output the total control quantity after limiting the amplitude. The prediction correction module is used to compare the deviation between the actual output of the robotic arm control system and the output of the prediction model in real time. When the deviation exceeds the threshold, the parameters of the prediction model are corrected online. The closed-loop control and monitoring module is used to feed back the actual output of the actuator to the Smith prediction error calculation module in real time, forming a closed control structure, and to monitor whether the effect meets the standard synchronously.
[0030] Compared with the prior art, the beneficial effects of the present invention are: This invention significantly improves the dynamic response speed and control accuracy of systems with large time delays: While traditional Smith predictors can compensate for lag, they are prone to large overshoot and slow response when facing load disturbances or sudden changes in setpoints, due to their reliance solely on the lag in feedback regulation. However, by introducing pseudo-feedforward, the solution can proactively acquire disturbance signals (such as changes in the controlled object's input load or external environmental interference) and apply them directly to the PD controller output via a pseudo-feedforward channel, initially offsetting the disturbance before feedback regulation begins. For example, in industrial temperature control scenarios, when the heating source power fluctuates, the pseudo-feedforward module can quickly detect this change and adjust the control quantity in advance, preventing significant temperature deviations due to lag. Simultaneously, the pseudo-feedforward compensation and the Smith predictor's feedback compensation form a "double insurance" mechanism, eliminating the impact of lag on system stability and reducing overshoot during dynamic processes, enabling the system to reach steady state faster. This improves control accuracy compared to traditional solutions, making it particularly suitable for large-lag industrial scenarios such as chemical reactors and long-distance transmission pipelines.
[0031] This invention enhances the system's adaptability to time-varying parameters and complex disturbances: In industrial settings, the parameters of the controlled object (such as inertia time constant and gain) often change over time due to variations in operating conditions (such as changes in material concentration and equipment aging), and there are complex disturbances from multiple sources (such as fluctuations in grid voltage and changes in ambient temperature). Traditional Smith predictors, due to their fixed models, are prone to compensation failure and system instability. The improved pseudo-feedforward scheme enhances adaptability through two optimizations: First, the pseudo-feedforward channel has an adaptive adjustment function, which can filter and limit the acquired signal to avoid overcompensation or undercompensation under complex disturbances by using fixed feedforward parameters. Second, the pseudo-feedforward signal is fused with the feedback signal, allowing the controller to simultaneously refer to disturbance information and system output deviation, more accurately correcting the control strategy and reducing the impact of time-varying parameters on the compensation effect. For example, in a liquid level control system, when the feed flow rate of the storage tank fluctuates due to changes in pump pressure, the pseudo-feedforward module can adapt to the changing pattern of flow disturbance in real time. Combined with the Smith predictor's compensation for liquid level lag, it ensures that the liquid level is always stable near the set value. Compared with traditional solutions, the system's stability in time-varying parameter scenarios is improved, and its anti-disturbance capability is significantly enhanced.
[0032] This invention addresses the issues of transmission delay superposition and disturbance signal lag by optimizing the signal transmission architecture and delay monitoring. This ensures that the algorithm performance of the pseudo-feedforward improved Smith predictor is fully utilized in the control of the robotic arm, ultimately improving the control stability and positioning accuracy of the robotic arm in high-precision scenarios.
[0033] This invention enhances disturbance suppression capabilities and adapts to time-varying parameters and complex disturbance scenarios: This invention addresses the problems of "latency in transmission of measurable disturbance signals and failure of pseudo-feedforward compensation" in traditional solutions by designing a full-process disturbance processing technology: First, disturbance signals such as gripping weight and workpiece position deviation are collected in real time by sensors. The electrical signals are converted into physical quantities through scaling conversion. Then, the disturbance is smoothed by first-order filtering. Finally, the compensation amount is calculated by substituting it into the difference equation and the actuator is limited.
[0034] This technology significantly reduces the total delay from the acquisition of disturbance signals to the generation of compensation quantities. Even if the weight of the parts grasped by the robotic arm fluctuates, the deviation of the end effector's movement can still be controlled within a high-precision range. At the same time, the deployed model mismatch adaptive correction mechanism can adjust the model parameters in real time, and can still maintain system stability when the load changes or equipment aging causes parameter drift. Compared with the oscillation risk of traditional Smith predictors when model mismatch occurs, this solution significantly improves the stability when the working conditions change.
[0035] This invention ensures robustness through closed-loop monitoring: The closed-loop monitoring and effect verification system established in this invention forms a control closed loop by providing real-time feedback of the actual output to the error calculation module. Simultaneously, it monitors disturbance suppression speed and steady-state error, and retrospectively adjusts delay monitoring parameters or compensation parameters when the indicators fail to meet standards. This system significantly reduces the risk of failure during long-term continuous operation of the robotic arm. In industrial welding scenarios, by monitoring welding deviation error and optimizing the delay compensation amount in real time, it significantly improves the welding qualification rate, while reducing production line downtime due to failures and increasing the average daily output of a single workstation. Attached Figure Description
[0036] Figure 1 This is a schematic diagram of a typical control system.
[0037] Figure 2 This is a schematic diagram of the Smith estimator.
[0038] Figure 3 This is a flowchart of a robotic arm control method based on motion prediction and delay compensation according to an embodiment of the present invention. Detailed Implementation
[0039] To facilitate understanding of the technical solution of the present invention by those skilled in the art, the technical solution of the present invention will now be further described in conjunction with the accompanying drawings.
[0040] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0041] This invention provides a robotic arm control method based on motion prediction and delay compensation, comprising: S10, Establish a mathematical model of the controlled robotic arm and the disturbance channel.
[0042] In one embodiment of the invention, the input of the robotic arm control system is acquired using the step response method. With output The data is then subjected to a Laplace transform to fit the dynamic transfer function of the controlled object. ; In the formula, Let be the dynamic transfer function of the controlled robotic arm, describing the hysteresis-free dynamics. The output signal of the controlled robotic arm. The input signal for the controlled robotic arm. The actual delay function of the controlled robotic arm. The actual delay time of the controlled robotic arm. For the Laplace operator, It is a natural constant.
[0043] And measure the execution lag time to establish the perturbation channel transfer function of the actuator: ; In the formula, For the actuator disturbance channel transfer function, For the total gain of the actuator, The equivalent time constant, The execution delay time.
[0044] Simultaneously collect measurable disturbance data (such as weight and flow rate). Establish the external disturbance channel transfer function: ; In the formula, For external disturbance channel transfer function, Let be the disturbance signal function. These are the output values of the external disturbance. These parameters will serve as the basis for subsequently constructing the prediction model and compensation function. Functions in the frequency domain It is a function in the time domain.
[0045] S20, based on the mathematical model established in S10, sets up a signal transmission delay monitoring mechanism.
[0046] In one embodiment of the present invention, an EtherCAT real-time bus supporting the IEEE 1588 PTP protocol is selected, and four key timestamps are collected at the master and slave stations respectively: the time when the master station issues the command. The time when the main station receives instructions Feedback time from the station , Main site receives feedback time .
[0047] Calculate uplink one-way delay Downlink one-way delay : , ; By taking into account the bus symmetry and eliminating interference from slave processing time, the total transmission delay can be obtained. : ; Total transmission delay Combined with execution lag time Get the total latency : .
[0048] The total delay time provides parameter support for the subsequent total lag model.
[0049] S30, based on the signal transmission delay monitoring mechanism, constructs a total lag model and a prediction model.
[0050] In one embodiment of the present invention, when constructing the total lag model, the actual lag of the object in step S10 is first considered. Transmission lag in step S20 Since the total lag is the sum of the two, the core formula is: This is a key parameter for the system's latency characteristics.
[0051] When establishing the prediction model, the hysteresis-free dynamic transfer function described in step S10 is used. Using the prototype, the estimated lag is taken. Introducing a pure time delay element (Total lag in reproduction), ultimately yielding For subsequent error calculation (such as...) (Provides a benchmark.)
[0052] S40: Based on the mathematical model established in S10 and the prediction model in S30, obtain the pseudo-feedforward compensation transfer function and set the calculation logic to be converted to the time domain in order to obtain the feedforward control quantity.
[0053] In one embodiment of the present invention, in conjunction with step S10 Executor transfer function and step S30 Derive the pseudo-feedforward transfer function The derivation process is as follows: Establish the relationship between the compensation quantity and the controlled quantity: The compensation quantity (feedforward control quantity) needs to be applied to the controlled quantity through the actuator and robotic arm, and the corresponding controlled quantity becomes: ; For changes in the controlled quantity, For the actuator disturbance channel transfer function, The dynamic transfer function of the controlled object. For compensation amount, For the Laplace operator.
[0054] Setting up a disturbance cancellation equation: The core of feedforward control is to allow the influence of the compensation amount to cancel out the influence of the disturbance, that is: ; in, This represents the output of external disturbances.
[0055] Substitute: ; have to: ; For external disturbance channel transfer function, The disturbance signal function; Based on the definition of the pseudo-transfer function, combined with the perturbation cancellation equation, and by transforming it, the initial pseudo-transfer function is solved: ; This is a pseudo-feedforward compensation transfer function; By combining the total delay time, the initial pseudo-transfer function is optimized to obtain the optimized pseudo-transfer function: ; This represents the total delay time.
[0056] Solving the pseudo-feedforward transfer function : The pseudo-feedforward transfer function is defined as: ; Transform the above equation: ; Combination optimization: Depend on Derivation: ; This is the transfer function for the PD controller.
[0057] Substituting into the above equation, we get: ; Write logic to measure the disturbance signal The compensation quantity (feedforward control quantity) is converted into the time domain. This prepares for the subsequent superposition of total control quantities, including measurable disturbance signals. With disturbance signal function The relationship between them is , This is the symbol for the Laplace transform. Pseudo-feedforward transfer function. With compensation amount The relationship between them is Specifically, the conversion process is as follows: The sensor collects disturbance electrical signals and converts them into physical quantities. ; The physical quantity that converts the disturbance electrical signal. This is the AD conversion value at the current sampling time. To disturb the physical range, For AD conversion range, For continuous time variables; The disturbance is smoothed by first-order filtering: ; This is the current filtered disturbance electrical signal. The filtered value from the previous time step. Let be the filter coefficients, where , The larger the value, the faster the response and the weaker the noise reduction.
[0058] Substituting into the difference equation, calculate the feedforward control quantity at the current time: ; This is the feedforward control quantity at the current moment. For the dynamic attenuation coefficient, This refers to the static gain coefficient, specifically... , , The sampling period is To compensate for the dynamic time constant, This is used to compensate for the static gain during disturbances.
[0059] The actuator range is limited, and the final output is a feedforward control quantity. Based on the actuator range (e.g., valve opening 0-100%, motor speed 0-3000 rpm), the following parameters are applied: To apply amplitude limiting, the formula is: ; It serves as the feedforward control quantity for the actuator and as the feedforward compensation quantity. The minimum / maximum compensation amount allowed by the actuator.
[0060] S50, set up the Smith prediction error calculation module to obtain signal error and feedback control quantity.
[0061] In one embodiment of the present invention, the core of the Smith prediction error calculation module is to optimize the error calculation logic by fusing prediction model information. The specific implementation process is as follows: First, the prediction model constructed in step S30 is called. Compared with the prototype model in step S10 The estimated output with lag was calculated separately. and lag-free forecast output .
[0062] For prediction models with lag , The target value for the control command of the corresponding robotic arm can be obtained through Laplace transform and inverse transform: ; In the formula, To predict the model's output, The target value for the control command of the robotic arm. For Laplace transform, This is the inverse Laplace transform.
[0063] The corresponding prototype model without hysteresis (dynamic transfer function of the controlled robotic arm) The input is Through Laplace transform and inverse transform, we can obtain: ; For prototype models without hysteresis Estimated output.
[0064] The signal error is calculated based on the following formula.
[0065] ; To estimate the signal error between the model's output and the hysteresis-free predicted output, This is the actual output of the robotic arm control system. This is the hysteresis compensation amount. It includes the signal error. Input the PD controller and calculate the feedback control quantity using the following formula: ; For feedback control, This is the proportionality coefficient. is the differential coefficient.
[0066] S60 superimposes the feedforward control quantity and the feedback control quantity to obtain the total control quantity, and outputs the total control quantity after limiting the amplitude.
[0067] In one embodiment of the present invention, the feedforward control quantity obtained in step S40 is... The feedback control quantity calculated in step S50 according to By superimposing the signals, the total control input signal of the robotic arm control system is obtained. In practical applications, the superimposed control signal also needs to be amplitude-limited. ; in, This represents a saturation or limiting function, used to ensure the control signal remains within the actuator's permissible operating range. The total control input signal after limiting. The output is sent to the actuator (such as a motor, valve, etc.) to drive the controlled robotic arm to move, thereby realizing the implementation of control commands.
[0068] S70 compares the actual output of the robotic arm control system with the output of the predicted model in real time. When the deviation exceeds the threshold, the parameters of the predicted model are corrected online.
[0069] In one embodiment of the present invention, during the control operation, the actual output of the system is compared in real time. The estimated output of the prediction model established in step S30 Deviation between: ; when deviation Exceeding the set threshold When, explain the prediction model Significant mismatch exists between the predicted model and the actual system, necessitating online correction of the predicted model parameters. Online correction of the predicted model parameters includes gain correction and time constant correction. The gain correction is as follows: ; The time constant is corrected to: ; In the formula, This is the corrected gain. The gain before correction. The deviation between the actual output of the robotic arm control system and the predicted model output. The corrected time constant. The time constant before correction is , , These are the correction step sizes used for gain correction and time constant correction, respectively. The corrected predicted model transfer function. It can be represented as: ; The updated version will be reused. Error calculation and feedforward / feedback compensation in step S50 are used to improve the robustness and adaptability of the control algorithm.
[0070] S80 feeds back the actual output of the actuator to the Smith prediction error calculation module in real time, forming a closed control structure and synchronously monitoring whether the effect meets the standard.
[0071] The actual output of the robotic arm control system The error is fed back in real time to the error calculation module in step S50, forming a closed-loop control structure. The error signal is... ; in, This serves as the reference input for the robotic arm control system. This error... It is used to adjust the PD controller output in real time to maintain system stability and tracking performance. During closed-loop operation, disturbance suppression performance indicators, such as suppression speed and steady-state error, are also monitored synchronously. When the suppression performance indicators are found to be below the predetermined standards, the delay monitoring parameters or compensation parameters are adjusted retrospectively to achieve online optimization and robustness improvement of the control strategy, ensuring that the entire control scheme meets the design requirements in terms of dynamic performance, steady-state accuracy, and disturbance suppression.
[0072] In this embodiment, the suppression speed (disturbance elimination time) is obtained using the following formula: ; In the formula, For disturbance elimination time, For the desired output, This is the deviation threshold.
[0073] The steady-state error is obtained using the following formula: ; In the formula, This represents the steady-state error.
[0074] In this embodiment, the delay monitoring parameters include the prediction model. hysteresis-free dynamic transfer function The compensation parameters include the lag compensation amount. Filter coefficients Disturbance compensation static gain .
[0075] This invention also provides a robotic arm control system based on motion prediction and delay compensation, which applies the above-described robotic arm control method based on motion prediction and delay compensation, including: Passive object and disturbance module, used to establish mathematical models of the controlled robotic arm and disturbance channels; The signal monitoring module is used to set up a signal transmission delay monitoring mechanism based on the mathematical model established in the passive object and disturbance module. The lag and prediction module is used to construct the total lag model and prediction model based on the signal transmission delay monitoring mechanism. The feedforward control module is used to obtain the pseudo-feedforward compensation transfer function and set the calculation logic to convert it to the time domain based on the mathematical model established in the passive object and disturbance module and the prediction model in the lag and prediction module, so as to obtain the feedforward control quantity. The feedback control module is used to set up the Smith prediction error calculation module and obtain signal error and feedback control quantity. The main control module is used to superimpose the feedforward control quantity and the feedback control quantity to obtain the total control quantity, and then output the total control quantity after limiting the amplitude. The prediction correction module is used to compare the deviation between the actual output of the robotic arm control system and the output of the prediction model in real time. When the deviation exceeds the threshold, the parameters of the prediction model are corrected online. The closed-loop control and monitoring module is used to feed back the actual output of the actuator to the Smith prediction error calculation module in real time, forming a closed control structure, and to monitor whether the effect meets the standard synchronously.
[0076] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.
[0077] The above embodiments are merely examples of implementation methods of the invention. The scope of protection of the present invention is not limited to the above embodiments. For those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention.
Claims
1. A robotic arm control method based on motion prediction and delay compensation, characterized in that, include: S10, Establish a mathematical model of the controlled robotic arm and the disturbance channel; S20, based on the mathematical model established in S10, sets up a signal transmission delay monitoring mechanism; S30, based on the signal transmission delay monitoring mechanism, constructs a total lag model and a prediction model; S40: Based on the mathematical model established in S10 and the prediction model in S30, obtain the pseudo-feedforward compensation transfer function and set the calculation logic to be converted to the time domain in order to obtain the feedforward control quantity. S50, set up the Smith prediction error calculation module to obtain signal error and feedback control quantity; S60 superimposes the feedforward control quantity and the feedback control quantity to obtain the total control quantity, and outputs the total control quantity after limiting the amplitude. S70 compares the deviation between the actual output of the robotic arm control system and the output of the prediction model in real time. When the deviation exceeds the threshold, the parameters of the prediction model are corrected online. S80 feeds back the actual output of the actuator to the Smith prediction error calculation module in real time, forming a closed control structure and synchronously monitoring whether the effect meets the standard.
2. The robotic arm control method based on motion prediction and delay compensation according to claim 1, characterized in that, Configure a signal transmission delay monitoring mechanism, including: Collect the following key timestamps from both the master and slave stations: the time the master station issues the command. The time when the main station receives instructions Feedback time from the station , Main site receives feedback time ; Calculate uplink one-way delay Downlink one-way delay : 、 ； By taking into account the bus symmetry and eliminating interference from slave processing time, the total transmission delay can be obtained. : ； Total transmission delay Combined with execution lag time Get the total latency : 。 3. The robotic arm control method based on motion prediction and delay compensation according to claim 2, characterized in that, Constructing the total lag model and the prediction model, including: The formula for obtaining the total delay time is used as the total lag model; The controlled robotic arm is used to describe the hysteresis-free dynamic transfer function. Using the prototype, the estimated lag time is taken. Total delay time Introducing a pure time delay element Obtain the prediction model : ; For the Laplace operator, It is a natural constant.
4. The robotic arm control method based on motion prediction and delay compensation according to claim 1, characterized in that, The pseudo-feedforward compensation transfer function is expressed by the following formula: ； For pseudo-feedforward compensation transfer function, For external disturbance channel transfer function, For the transfer function of the PD controller, This is the total delay time. For the Laplace operator.
5. The robotic arm control method based on motion prediction and delay compensation according to claim 4, characterized in that, Obtaining feedforward control values includes: The sensor collects the disturbance electrical signal and converts it into a physical quantity: ； The physical quantity that converts the disturbance electrical signal. This is the AD conversion value at the current sampling time. To disturb the physical range, For AD conversion range, For continuous time variables; The disturbance is smoothed by first-order filtering: ； This is the current filtered disturbance electrical signal. The filtered value from the previous time step. These are the filter coefficients; Substituting into the difference equation, calculate the feedforward control quantity at the current time: ； This is the feedforward control quantity at the current moment. For the dynamic attenuation coefficient, This refers to the static gain coefficient. The actuator range is limited, and the final output is a feedforward control quantity; ； It serves as the feedforward control quantity for the actuator and as the feedforward compensation quantity. The minimum / maximum compensation amount allowed by the actuator.
6. The robotic arm control method based on motion prediction and delay compensation according to claim 1, characterized in that, The Smith prediction error calculation module is expressed by the following formula: ； The feedback control value is obtained using the following formula: ； In the formula, To estimate the signal error between the model's output and the hysteresis-free predicted output, The target value for the control command of the robotic arm. This is the actual output of the robotic arm control system. This is the amount of delayed compensation. To predict the model's output, This is a forecast output without lag. For continuous time variables, For feedback control, This is the proportionality coefficient. is the differential coefficient.
7. The robotic arm control method based on motion prediction and delay compensation according to claim 1, characterized in that, The total control quantity is then limited and output, expressed by the following formula: ； In the formula, This is the total control input signal after amplitude limiting. For feedback control, This is the feedforward control variable. It is a saturation or limiting function.
8. The robotic arm control method based on motion prediction and delay compensation according to claim 1, characterized in that, Online correction of the parameters of the prediction model includes gain correction and time constant correction. The gain correction is as follows: ； The time constant is corrected to: ； In the formula, This is the corrected gain. The gain before correction. The deviation between the actual output of the robotic arm control system and the predicted model output. The corrected time constant. The time constant before correction is , , These are the correction step sizes used for gain correction and time constant correction, respectively.
9. The robotic arm control method based on motion prediction and delay compensation according to claim 1, characterized in that, Whether the monitoring effect meets the standard includes synchronously monitoring disturbance suppression performance indicators during closed-loop operation: suppression speed and steady-state error; when the suppression performance indicators are found to be below the predetermined standard, the delay monitoring parameters or compensation parameters are adjusted retrospectively.
10. A robotic arm control system based on motion prediction and delay compensation, characterized in that, The robotic arm control method based on motion prediction and delay compensation according to any one of claims 1-9 includes: Passive object and disturbance module, used to establish mathematical models of the controlled robotic arm and disturbance channels; The signal monitoring module is used to set up a signal transmission delay monitoring mechanism based on the mathematical model established in the passive object and disturbance module. The lag and prediction module is used to construct the total lag model and prediction model based on the signal transmission delay monitoring mechanism. The feedforward control module is used to obtain the pseudo-feedforward compensation transfer function and set the calculation logic to convert it to the time domain based on the mathematical model established in the passive object and disturbance module and the prediction model in the lag and prediction module, so as to obtain the feedforward control quantity. The feedback control module is used to set up the Smith prediction error calculation module and obtain signal error and feedback control quantity. The main control module is used to superimpose the feedforward control quantity and the feedback control quantity to obtain the total control quantity, and then output the total control quantity after limiting the amplitude. The prediction correction module is used to compare the deviation between the actual output of the robotic arm control system and the output of the prediction model in real time. When the deviation exceeds the threshold, the parameters of the prediction model are corrected online. The closed-loop control and monitoring module is used to feed back the actual output of the actuator to the Smith prediction error calculation module in real time, forming a closed control structure, and to monitor whether the effect meets the standard synchronously.