Precise compliant force control method of polishing robot under environment uncertainty

Abstract

This invention relates to the field of intelligent control technology for industrial robots, and discloses a precise compliant force control method for a grinding and polishing robot under uncertain environmental conditions. The method includes establishing a dual-closed-loop control model of the environment and the robotic arm based on a set linear environment model and a robotic arm dynamics model; iteratively solving for the optimal control gain in a data-driven manner based on adaptive dynamic programming for the unknown environmental characteristics in the control model; designing a free-space control gain to achieve stable relative motion approaching the environment; designing a smooth transition mechanism for the free-contact space control gain; achieving model-free optimal control based on data iteration through adaptive dynamic programming; employing dual-closed-loop impedance control to achieve the impedance effect of the force; and utilizing contact transient impact modeling to dynamically adjust the control gain, ensuring the stability of the contact force and processing quality, making it suitable for various complex working environments.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for industrial robots, and more specifically to a method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions. Background Technology

[0002] With rapid technological advancements and intensifying competition in global manufacturing, industrial robots are playing an increasingly prominent role in precision manufacturing processes. Especially in high-precision machining scenarios such as grinding and polishing, industrial robots are gradually replacing traditional manual labor by achieving precise operation and stable quality assurance, significantly improving production efficiency and product quality. This not only aligns with the development trend of "intelligent manufacturing" but also serves as a key entry point for the intelligent upgrading of the manufacturing industry and technological innovation in industrial robots. In aerospace, marine engineering equipment, and automotive manufacturing, precision grinding and polishing operations place extremely stringent requirements on surface finish and shape accuracy. However, due to the complex and variable grinding and polishing environment and the numerous sources of error, existing grinding and polishing industrial robots cannot fully meet the high standards of product consistency and efficiency required by modern industry. Therefore, the need for precise and compliant force control of grinding and polishing industrial robots under uncertain environmental conditions is becoming increasingly prominent.

[0003] In the field of compliant force control for industrial robots, traditional sensor-based force / dynamic control methods (such as impedance control and admittance control) are widely used to handle the interaction between the robot and its environment. Impedance control, by modeling the dynamic relationship between the end-effector pose and contact force as a mass-spring-damped system, provides an easily adjustable dynamic response mechanism. However, traditional impedance control often struggles to achieve precise force control when facing unknown or time-varying environmental stiffness, easily leading to contact force overshoot or transient instability. In grinding and polishing tasks, this control deficiency can cause surface damage or reduced processing quality, severely impacting production efficiency and product consistency.

[0004] To address these issues, several improvement strategies have been proposed in recent years, including methods based on trajectory deformation and adaptive impedance parameter adjustment. Trajectory deformation methods dynamically adjust the reference motion trajectory to adapt to unknown environmental stiffness; however, this method has limited control bandwidth and struggles to handle scenarios with rapidly changing environmental stiffness. In contrast, adaptive impedance parameter adjustment methods update impedance parameters (such as stiffness and damping coefficients) online to adapt to environmental characteristics, but these methods typically require predefined offline parameter sets or high-bandwidth system support, limiting their application in complex industrial scenarios.

[0005] Furthermore, in grinding and polishing tasks involving flexible materials or fragile components, the requirements for contact force control are even more stringent. Over-force at the moment of contact can lead to material damage or a decrease in surface quality, while insufficient contact force can result in unsatisfactory processing results. Therefore, balancing the stability and accuracy of force control in both transient and continuous contact tasks is the core challenge facing compliant force control technology.

[0006] In recent years, compliant force control algorithms combining optimal control theory and online estimation of environmental parameters have become a research hotspot. For example, environmental stiffness estimation methods based on extended Kalman filters can dynamically adjust the control gain during contact to optimize the target interaction force. However, most of these methods are only effective under static or low-speed contact conditions, and their adaptability to highly dynamic and complex environments still needs further improvement.

[0007] Based on the above background, this invention proposes a precise compliant force control method for grinding and polishing robots, which combines adaptive impedance control and optimal control algorithm to address the impact of environmental uncertainties on contact force regulation and improve the accuracy and stability of grinding and polishing tasks. Summary of the Invention

[0008] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions, so as to solve the problems existing in the background art.

[0009] This invention provides the following technical solution: a method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions, comprising the following steps:

[0010] Step 1: Establish a dual closed-loop control model for the environment and robotic arm based on the established linear environment model and robotic arm dynamics model;

[0011] Step 2: For the unknown environmental characteristics in the control model, the optimal control gain is iteratively solved using adaptive dynamic programming in a data-driven manner.

[0012] Step 3: Design free-space control gain to achieve stable relative motion approaching the environment;

[0013] Step 4: Design a smooth transition mechanism for controllable gain in free-contact space.

[0014] Preferably, step one specifically comprises:

[0015] Based on the dynamic interaction between the robotic arm end effector and the complex environment, an impedance control model considering environmental uncertainties is established.

[0016] The linear environment model is represented as follows: In the formula, For environmental location, Indicates environmental speed. The environmental equilibrium position is defined; the inner loop impedance control defines the damping and stiffness of the environment as follows: and The contact force between the environment and the end effector of the robotic arm is ;

[0017] The dynamic model of the robotic arm is represented as follows: ;in, and These represent the actual end-effector position and the desired end-effector position, respectively. Indicates the robot's actual speed. This represents the robot's actual acceleration. The contact force at the end of the robotic arm; Indicates the robot's stiffness; Indicates the robot's mass; This indicates robot damping.

[0018] Preferably, the establishment of the environment-robotic arm dual closed-loop control model specifically involves:

[0019] Define a stable environmental equilibrium position Then the environmental position is equal to the robot's end-effector position, that is Furthermore, the force applied to the environment is equal to the contact force applied to the end effector of the robotic arm, i.e. ;in, This represents the force of direct contact between the robot and its environment.

[0020] For instantaneous force response, considering a first-order reference model, define... Reference force for the first-order reference model:

[0021] ;

[0022] Among them, parameters The bandwidth affects the settling time of a closed-loop control system. For the desired contact force, The first derivative of the reference force of the first-order reference model is given by the reference position from the input of the outer loop control system to the inner loop impedance system. Defined as:

[0023] ;

[0024] in, For robot end effector speed Feedback gain, For the robot's end position Feedback gain, Force tracking error The integral gain, where, , For model reference input force Feedforward gain; Indicates the integration time.

[0025] Preferably, the environment-robotic arm dual closed-loop control model is as follows:

[0026] ;

[0027] Rewritten as: ;in, For system status, The system state matrix, For the input matrix, For system input, The third derivative of the robot's position. The first derivative of the force tracking error. It is the second derivative of the reference force of the first-order reference model.

[0028] Preferably, the iterative solution of the optimal control gain based on adaptive dynamic programming in a data-driven manner specifically involves:

[0029] The optimal control gain is solved iteratively using the ADP algorithm, and the optimal control gain is obtained through data iteration using the policy-based ADP algorithm; the cost function in optimal control is also discussed. Represented as:

[0030] ;in, , The weights are the cost function weights. Indicates the integration time; Represents a diagonal matrix;

[0031] The objective function for optimal control is expressed as: ,and Minimum; among which, This represents the optimal control gain of the system.

[0032] Preferably, the specific steps for using the ADP algorithm based on policy iteration to iteratively solve for the optimal control gain are as follows:

[0033] Step a: Initialization: Given an initial control gain initial gain It can satisfy system stability;

[0034] Step b: Online data collection: Let Collect and output status, where Assuming random noise, calculate the iteration matrix. and ;in, Related to system state, control input, and control gain. Related to the system state and the weight matrix;

[0035] Step c: Strategy evaluation and improvement: Solve for the optimal control kernel matrix using the following formula. and control gain :

[0036] ;in, Represents matrix column vectorization. Indicates the control gain at the next moment;

[0037] Step d: If Then it converges, where The convergence threshold is used; otherwise, let... Repeat step cd until convergence.

[0038] Preferably, the design of the free space control gain to achieve stable relative motion approaching the environment specifically involves:

[0039] The control gain is optimized based on a collision impact model, using relative motion as a parameter, and the control gain is set accordingly. and The feedforward force is zero, and the feedforward force is... Set as The reference position in free space is represented as:

[0040] .

[0041] Preferably, the design-free-contact space control gain smooth transition mechanism specifically includes:

[0042] Design a smooth switching mechanism for control gain, the first Control gain After a time interval Free space control gain Smooth transition to contact space control gain This can be expressed as a formula:

[0043] ;in, express Corresponding filtering gain With filtration time The changing value.

[0044] The technical effects and advantages of this invention are as follows:

[0045] This invention, through steps two, three, and four, facilitates model-free optimal control based on data iteration using adaptive dynamic programming (ADP). It employs dual-closed-loop impedance control to achieve the impedance effect of force, and utilizes contact transient impact modeling to dynamically adjust the control gain, ensuring the stability of the contact force and processing quality. It is suitable for various complex working environments. By combining the dual-closed-loop force control framework with the optimal control algorithm based on ADP, it can achieve precise compliant force control for grinding and polishing tasks under uncertain environments, effectively improving processing accuracy and stability, while also taking into account the real-time performance and robustness of the system. Attached Figure Description

[0046] Figure 1 This is a flowchart of the precise compliance force control method for the grinding and polishing robot under uncertain environmental conditions according to the present invention.

[0047] Figure 2 This is a schematic diagram of the intelligent compliant force control method for the contact process between the grinding and polishing robotic arm and an uncertain environment according to the present invention.

[0048] Figure 3 The curve showing the change of the kernel matrix P matrix from the theoretically optimal P* matrix during data iteration in the ADP algorithm of this invention.

[0049] Figure 4 The curve showing the change of the control gain K matrix distance from the theoretically optimal K* matrix during data iteration in the ADP algorithm of this invention.

[0050] Figure 5 This is the contact force response curve of the force control algorithm of the present invention under the condition of sudden change in environmental characteristics.

[0051] Figure 6 This is the robot position change curve of the force control algorithm of the present invention under sudden changes in environmental characteristics. Detailed Implementation

[0052] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] like Figures 1-6 As shown, this invention provides a method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions, comprising the following steps:

[0054] Step 1: Establish a dual closed-loop control model for the environment and robotic arm based on the established linear environment model and robotic arm dynamics model;

[0055] Step 2: For the unknown environmental characteristics in the control model, the optimal control gain is iteratively solved using adaptive dynamic programming in a data-driven manner.

[0056] Step 3: Design free-space control gain to achieve stable relative motion approaching the environment; fully consider the uncertainty of the environmental position;

[0057] Step 4: Design a free-contact space control gain smooth transition mechanism to ensure that no overshoot force is generated at the moment of contact.

[0058] In this embodiment, it should be specifically explained that step one is as follows:

[0059] Based on the dynamic interaction between the robotic arm end effector and the complex environment, an impedance control model considering environmental uncertainties is established.

[0060] The linear environment model is represented as follows: In the formula, For environmental location, Indicates environmental speed. The environmental equilibrium position is defined; the inner loop impedance control defines the damping and stiffness of the environment as follows: and The contact force between the environment and the end effector of the robotic arm is ;

[0061] The dynamic model of the robotic arm is represented as follows: ;in, and These represent the actual end-effector position and the desired end-effector position, respectively. Indicates the robot's actual speed. This represents the robot's actual acceleration. The contact force at the end of the robotic arm; Indicates the robot's stiffness; Indicates the robot's mass; This indicates robot damping.

[0062] In this embodiment, it should be specifically explained that the establishment of the environment-robotic arm dual closed-loop control model is as follows:

[0063] Define a stable environmental equilibrium position Then the environmental position is equal to the robot's end-effector position, that is Furthermore, the force applied to the environment is equal to the contact force applied to the end effector of the robotic arm, i.e. ;in, This represents the force of direct contact between the robot and its environment.

[0064] Considering the feedforward feedback structure of the outer loop, to prevent the expectation force An excessively large instantaneous input can cause system overshoot. For the instantaneous force response, a first-order reference model is considered, defined as follows: Reference force for the first-order reference model:

[0065] ;

[0066] Among them, parameters The bandwidth affects the settling time of a closed-loop control system. For the desired contact force, The first derivative of the reference force of the first-order reference model is given by the reference position from the input of the outer loop control system to the inner loop impedance system. Defined as:

[0067] ;

[0068] in, For robot end effector speed Feedback gain, For the robot's end position Feedback gain, Force tracking error The integral gain, where, , For model reference input force Feedforward gain; Indicates the integration time;

[0069] The model of the feedforward feedback dual-loop impedance force control system, i.e., the environment-robotic arm dual-loop control model, can be obtained from the above formula:

[0070] ;

[0071] Rewritten as: ;in, For system status, The system state matrix, For the input matrix, For system input, The third derivative of the robot's position. The first derivative of the force tracking error. The second derivative of the reference force of the first-order reference model;

[0072] Assuming the robotic arm is in an inertial frame The axis direction is close to the environment; to verify the effectiveness of the method, the environment parameters are assumed to be... , The dynamic parameters of the robotic arm are , , The bandwidth of the reference force .

[0073] In this embodiment, it should be specifically explained that the iterative solution of the optimal control gain based on adaptive dynamic programming in a data-driven manner is as follows:

[0074] By collecting input and state data, the ADP algorithm is used to iterate the optimal control gain to ensure accurate tracking of the contact force trajectory.

[0075] The ADP algorithm is used for optimal control gain iteration, and the cost function in optimal control is... Represented as:

[0076] ;in, , The weights are the cost function weights. Indicates the integration time; Represents a diagonal matrix;

[0077] The objective function for optimal control is expressed as: ,and Minimum; among which, This represents the optimal control gain of the system;

[0078] Since environmental uncertainties prevent traditional model-based optimal control methods from being implemented, this embodiment uses the ADP algorithm based on policy iteration to iteratively solve for the optimal control gain.

[0079] In this embodiment, it should be specifically explained that the specific steps of using the policy iteration-based ADP algorithm to iteratively solve for the optimal control gain are as follows:

[0080] Step a: Initialization: Given an initial control gain initial gain It can satisfy system stability;

[0081] Step b: Online data collection: Let Collect and output status, where Assuming random noise, calculate the iteration matrix. and ;in, Related to system state, control input, and control gain. Related to the system state and the weight matrix;

[0082] Step c: Strategy evaluation and improvement: Solve for the optimal control kernel matrix using the following formula. and control gain :

[0083] ;in, Represents matrix column vectorization. Indicates the control gain at the next moment;

[0084] Step d: If Then it converges, where The convergence threshold is used; otherwise, let... Repeat step cd until convergence;

[0085] In this embodiment, the system simulation step size is taken as: Weighting coefficient ,matrix The initial value is set to the identity matrix, because The matrix satisfies Hurwitz, therefore the control gain The initial value is ;

[0086] First, based on the established known environment model, the theoretically optimal impedance control gain is calculated as follows: ;

[0087] Input via system Collect system data for 100 rounds, noise ,in The initial state of the system is set to The convergence threshold is set to Through strategy evaluation and improvement, the kernel matrix was improved in just 7 rounds of iteration. and control gain That is, converge to the optimal value, such as Figure 3 As shown, the optimal control gain obtained through optimal iteration is: This is consistent with model-based approaches.

[0088] In this embodiment, it should be specifically explained that the design of the free space control gain to achieve stable relative motion approaching the environment specifically means:

[0089] The control gain is optimized based on the collision and impact model to reduce the impact force during the contact transient and suppress the contact force overshoot.

[0090] Since the environmental location is unknown, only relative motion can be used. During free space motion, the contact force is zero, which will continuously generate force errors. Therefore, a control gain is set. and The feedforward force is zero, and the feedforward force is... Set as To achieve a stable feedforward effect, the reference position in free space is:

[0091] .

[0092] In this embodiment, it should be specifically explained that the design freedom-contact space control gain smooth transition mechanism is as follows:

[0093] Design a smooth switching mechanism for control gain to prevent overshoot of contact force at the moment of contact;

[0094] To ensure a smooth transition from free space to contact space, the smoothing property of the cosine function is utilized. Control gain After a time interval Free space control gain Smooth transition to contact space control gain :

[0095] ;in, express Corresponding filtering gain With filtration time The value of the change;

[0096] The proposed ADP-based dual closed-loop force control algorithm ultimately achieves the following results: Figure 4 and Figure 5 As shown.

[0097] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0098] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions, characterized in that: Includes the following steps: Step 1: Establish a dual closed-loop control model for the environment and robotic arm based on the established linear environment model and robotic arm dynamics model; Step one specifically involves: Based on the dynamic interaction between the robotic arm end effector and the complex environment, an impedance control model considering environmental uncertainties is established. The linear environment model is represented as follows: In the formula, For environmental location, Indicates environmental speed. The environmental equilibrium position is defined; the inner loop impedance control defines the damping and stiffness of the environment as follows: and The contact force between the environment and the end effector of the robotic arm is ; The dynamic model of the robotic arm is represented as follows: ;in, and These represent the actual end-effector position and the desired end-effector position, respectively. Indicates the robot's actual speed. This represents the robot's actual acceleration. The contact force at the end of the robotic arm; Indicates the robot's stiffness; Indicates the robot's mass; Indicates robot damping; The environment-robotic arm dual closed-loop control model is as follows: ； Rewritten as: ;in, For system status, The system state matrix, For the input matrix, For system input, The third derivative of the robot's position. The first derivative of the force tracking error. The second derivative of the reference force of the first-order reference model; Step 2: For the unknown environmental characteristics in the control model, the optimal control gain is iteratively solved using adaptive dynamic programming in a data-driven manner. Step 3: Design free-space control gain to achieve stable relative motion approaching the environment; The design of free-space control gain to achieve stable relative motion approaching the environment specifically involves: The control gain is optimized based on a collision impact model, using relative motion as a parameter, and the control gain is set accordingly. and The feedforward force is zero, and the feedforward force is... Set as The reference position in free space is represented as: ;in, For robot end effector speed Feedback gain, For model reference input force Feedforward gain; Step 4: Design a smooth transition mechanism for controllable gain in free-contact space; The design-free-contact-space control gain smooth transition mechanism is specifically as follows: Design a smooth switching mechanism for control gain, the first Control gain After a time interval Free space control gain Smooth transition to contact space control gain This can be expressed as a formula: ;in, express Corresponding filtering gain With filtration time The changing value.

2. The method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions according to claim 1, characterized in that: The establishment of the environment-robotic arm dual closed-loop control model is specifically as follows: Define a stable environmental equilibrium position Then the environmental position is equal to the robot's end-effector position, that is Furthermore, the force applied to the environment is equal to the contact force applied to the end effector of the robotic arm, i.e. ;in, This represents the force of direct contact between the robot and its environment. For instantaneous force response, considering a first-order reference model, define... Reference force for the first-order reference model: ； Among them, parameters The bandwidth affects the settling time of a closed-loop control system. For the desired contact force, The first derivative of the reference force of the first-order reference model is given by the reference position from the input of the outer loop control system to the inner loop impedance system. Defined as: ； in, For robot end effector speed Feedback gain, For the robot's end position Feedback gain, Force tracking error The integral gain, where, , For model reference input force Feedforward gain; Indicates the integration time.

3. The method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions according to claim 2, characterized in that: The specific steps of iteratively solving for the optimal control gain using adaptive dynamic programming in a data-driven manner are as follows: The optimal control gain is solved iteratively using the ADP algorithm, and the optimal control gain is obtained through data iteration using the policy-based ADP algorithm; the cost function in optimal control is also discussed. Represented as: ;in, , The weights are the cost function weights. Indicates the integration time; Represents a diagonal matrix; The objective function for optimal control is expressed as: ,and Minimum; among which, This represents the optimal control gain of the system.

4. The method for precise compliance force control of a grinding and polishing robot under uncertain environmental conditions according to claim 3, characterized in that: The specific steps for using the policy iteration-based ADP algorithm to iteratively solve for the optimal control gain are as follows: Step a: Initialization: Given an initial control gain initial gain It can satisfy system stability; Step b: Online data collection: Let Collect and output status, where Assuming random noise, calculate the iteration matrix. and ;in, Related to system state, control input, and control gain. Related to the system state and the weight matrix; Step c: Strategy evaluation and improvement: Solve for the optimal control kernel matrix using the following formula. and control gain : ;in, Represents matrix column vectorization. Indicates the control gain at the next moment; Step d: If Then it converges, where The convergence threshold is used; otherwise, let Repeat step cd until convergence.

Images