A robot fine operation control method based on contact state uncertainty

Abstract

The application discloses a robot fine operation control method based on contact state uncertainty, comprising the following steps: acquiring contact information between a robot end effector and an environment, and constructing a contact feature vector; obtaining the probability distribution of each state in the contact state set by using a contact state recognition model, calculating the contact state uncertainty, and then obtaining the normalized uncertainty; constructing an adaptive strategy generation function to generate a control strategy; dynamically adjusting the control parameters of the robot based on the control strategy, and realizing compliant control execution through a compliant control model; in the process of compliant control execution, a sliding error is constructed based on the robot end position error, and an adaptive update law of the control parameters is designed based on the Lyapunov method to adjust the stiffness and damping parameters online. The application can improve the operation success rate and system stability of the robot under different contact states.

Description

Technical Field

[0001] This invention relates to the field of robot intelligent control technology, and more specifically to a method for fine operation control of robots based on contact state uncertainty. Background Technology

[0002] With the continuous development of industrial and service robot technologies, the demand for robots to perform precision tasks in complex environments is increasing, such as precision assembly, insertion and removal operations, and contact-based work in automobile manufacturing. These tasks typically involve dynamic contact between the robot's end effector and the environment, placing high demands on operational accuracy, stability, and compliance.

[0003] In existing technologies, precise robot manipulation mainly relies on visual guidance or force-based control methods. Visual guidance methods are easily affected by occlusion, changes in lighting, and calibration errors during the contact phase, making it difficult to provide stable and reliable feedback. Force-based control methods, such as impedance control and admittance control, can achieve a certain degree of compliant interaction, but they usually use fixed control strategies or preset parameters, making it difficult to adapt to complex and changing contact states.

[0004] In actual operation, the contact state between the robot and the environment is characterized by uncertainty and dynamic change, such as gradually transitioning from a free space state to initial contact, edge contact, and stable contact. The optimal control strategy differs significantly under different contact states, and existing methods often lack quantitative assessment of the uncertainty of the contact state, leading to untimely or unreasonable switching of control strategies, thereby affecting the success rate of operation and system stability. Summary of the Invention

[0005] In view of this, the present invention provides a robot fine operation control method based on contact state uncertainty, in order to solve the problem that the existing technology lacks quantitative assessment of contact state uncertainty, which leads to untimely or unreasonable switching of control strategies, thereby affecting the success rate of operation and system stability.

[0006] A method for fine manipulation control of a robot based on contact state uncertainty, comprising: Step S1: Obtain contact information between the robot's end effector and the environment, and construct a contact feature vector; Step S2: Based on the contact feature vector, the probability distribution of each state in the contact state set is obtained using the contact state recognition model, and the contact state uncertainty is calculated. Then, the contact state uncertainty is normalized to obtain the normalized uncertainty. Step S3: Construct an adaptive policy generation function based on the contact state and normalized uncertainty to generate a control policy; Step S4: Dynamically adjust the robot's control parameters based on the control strategy, and implement compliant control execution through the compliant control model; Step S5: During the execution of compliant control, a slip error is constructed based on the robot end position error, and an adaptive update law for the control parameters is designed based on the Lyapunov method to adjust the stiffness and damping parameters online. Step S6: During task execution, the contact information is updated in real time, and steps S1 to S5 are executed repeatedly to achieve joint adaptive updates of the control strategy and control parameters until the task termination condition is met.

[0007] The robot fine operation control method based on contact state uncertainty provided by the present invention has the following beneficial effects: (1) By constructing a contact state recognition model and introducing a contact state uncertainty assessment mechanism, this invention achieves a quantitative description of the reliability of the contact state, and solves the problem that the prior art lacks a quantitative assessment of the uncertainty of the contact state. (2) By constructing a strategy generation function based on uncertainty, this invention realizes the continuous generation and smooth transition of control strategies, avoids the instability problem caused by traditional strategy switching, and effectively alleviates the problem of untimely or unreasonable control strategy switching. (3) By integrating exploratory control strategy and execution control strategy, this invention enables the robot to have adaptive decision-making ability in uncertain environments, and is suitable for various contact-driven tasks such as assembly and grinding, and has good robustness and adaptability. (4) The present invention combines a compliant control method to realize online adjustment of control parameters, thereby improving the stability and success rate of robot precision operation. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a robot fine manipulation control method based on contact state uncertainty, provided in an embodiment of the present invention. Detailed Implementation

[0009] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0010] Example 1 Please see Figure 1The first embodiment of the present invention provides a robot fine operation control method based on contact state uncertainty, including steps S1 to S6. This embodiment is based on the insertion operation task based on contact state uncertainty. This embodiment takes the robot performing a shaft hole insertion task as an example to describe the method of the present invention in detail. The robot includes a six-degree-of-freedom robotic arm, a six-dimensional force / torque sensor, and a controller. The force / torque sensor is installed between the end effector and the tool to obtain the contact information between the robot and the environment in real time.

[0011] Step S1: Obtain contact information between the robot's end effector and the environment, and construct a contact feature vector.

[0012] The contact force and torque between the robot's end effector and the environment, obtained in the tool coordinate system, are defined as follows:

[0013] in, To represent the original contact force and torque signals measured by the robot's end effector in the tool coordinate system, These represent the contact forces along the three coordinate axes, respectively. Indicates the torque in the corresponding direction. This indicates transpose.

[0014] To reduce the impact of sensor noise, the contact signal is low-pass filtered:

[0015] in, This is the generalized contact force vector composed of contact force and torque. This is a low-pass filter function.

[0016] Then calculate Rate of change (used to characterize the trend of contact change):

[0017] in, for The derivative with respect to time, express The differential, Indicates time The differential.

[0018] Simultaneously acquire the end-effector pose:

[0019] in, The end-effector pose vector. The coordinates representing the end position. Indicates the attitude angle.

[0020] Constructing contact feature vectors for:

[0021] in, This is the generalized contact force vector composed of contact force and torque. for The derivative with respect to time, The end-effector pose vector. This indicates transpose.

[0022] Step S2: Based on the contact feature vector, the probability distribution of each state in the contact state set is obtained using the contact state recognition model, and the contact state uncertainty is calculated. Then, the contact state uncertainty is normalized to obtain the normalized uncertainty.

[0023] First, a set of contact states between the robot's end effector and the environment is predefined. for:

[0024] in, , , , , , These are respectively the free space state, initial contact state, edge contact state, eccentric contact state, stuck state, and stable contact state.

[0025] Based on this, using contact feature vectors The probability distribution of each contact state is output using the contact state recognition model:

[0026] in, Indicates the first Contact state, For the first The probability distribution of contact states. For the first A function to determine the contact state of a class. For the first A function to determine the contact state of a class.

[0027] Then, based on the contact state probability distribution, information entropy is used to quantify the uncertainty. Specifically, the contact state uncertainty is calculated using the following formula. :

[0028] To facilitate subsequent control strategy design, the uncertainty is normalized to obtain the normalized uncertainty:

[0029] in, To normalize uncertainty, The number of contact states. ,like This indicates an ambiguous state (multiple mixed states); if This indicates that the state is clear.

[0030] Step S3: Construct an adaptive policy generation function based on the contact state and normalized uncertainty to generate a control policy.

[0031] The expression for the adaptive policy generation function is:

[0032]

[0033] in, For control strategies, Let be the weighting function with respect to uncertainty. This is an exploratory control strategy used to perform contact environment searches when uncertainty is high; This is an execution-oriented control strategy used to perform fine-grained operations when uncertainty is low; , To adjust the parameters.

[0034] In the Cartesian space of the robot's end effector Represented as:

[0035] in, These represent the velocity components of the robot's end effector in three directions.

[0036] The exploratory control strategy is a disturbance control strategy generated along the contact plane direction, which takes the form of periodic or random disturbance motion. Specifically, in this embodiment, the exploratory strategy is used to search for hole positions when the contact state uncertainty is high. It is defined as periodic disturbance motion within the contact plane, and its expression is:

[0037] in, For the disturbance amplitude, The frequency is the disturbance frequency, and the disturbance is applied to the contact plane (xy plane).

[0038] The execution-type control strategy is a control strategy along the target direction, which achieves fine operation through force control or trajectory tracking. Specifically, in this embodiment, the execution strategy is used to perform an insertion operation when the contact state is clear. It is defined as movement along the normal direction of the tool coordinate system, and its expression is:

[0039] in, Determined through force control:

[0040] in, For the desired contact force, For actual normal contact force, To control the gain.

[0041] Step S4: Dynamically adjust the robot's control parameters based on the control strategy, and implement compliant control execution through a compliant control model.

[0042] The compliance control model satisfies the following equation:

[0043] in, The equivalent inertia matrix, Indicates the actual position of the robot's end effector. Indicates the robot's end effector speed. Indicates the acceleration at the robot's end effector. Indicates the desired position. Here is the damping matrix. Here is the stiffness matrix. It is an external contact force.

[0044] In step S5, during the execution of compliant control, a slip error is constructed based on the robot's end-effector position error, and an adaptive update law for the control parameters is designed based on the Lyapunov method to adjust the stiffness and damping parameters online.

[0045] In order to ensure the stability of the robot under different contact states, this embodiment designs an adaptive update law for the control parameters based on the Lyapunov method.

[0046] First, define the position error. for:

[0047] Define slip error for:

[0048] in, for The derivative with respect to time, For error adjustment parameters, .

[0049] The adjustment of the damping matrix primarily affects the system's dynamic response, and its stability is already reflected in the dynamics term of the Lyapunov function; therefore, there is no need to construct a separate parameter error term. Constructing the Lyapunov function... , is an energy function used to describe system state error and parameter error.

[0050]

[0051] in, Represents the trace operation of a matrix. The current stiffness matrix, For the ideal stiffness matrix, It is a positive definite symmetric matrix used to adjust the parameter update rate.

[0052] Then, to make the Lyapunov function decrease, the stiffness update law is designed as follows:

[0053] in, This represents the rate of change of the stiffness matrix with respect to time. This is the uncertainty adjustment coefficient. This is an uncertain reference value. The first term in the stiffness update law. For error-driven adjustment, the second term Adjustment for uncertainty.

[0054] Next, the damping update law is designed as follows:

[0055] in, This represents the rate of change of the damping matrix with respect to time; It is a positive definite symmetric matrix used to adjust the parameter update rate; , To adjust the parameters, Here is the damping reference matrix.

[0056] Specifically, under the action of the aforementioned adaptive update law, the constructed Lyapunov function satisfies that its time derivative is non-positive, i.e. This ensures the stability of the system under different contact conditions.

[0057] Step S6: During task execution, the contact information is updated in real time, and steps S1 to S5 are executed repeatedly to achieve joint adaptive updates of the control strategy and control parameters until the task termination condition is met.

[0058] In this embodiment, the task termination condition is determined to be met and the task is completed when any of the following conditions are met: (1) the insertion force is stable and reaches the set threshold; (2) the end pose error is less than the preset accuracy.

[0059] Example 2 The second embodiment of the present invention is a surface polishing operation task based on contact state uncertainty assessment. This embodiment takes a robot performing a complex curved surface polishing task as an example to illustrate the method of the present invention, which includes steps S1 to S6. In this task, the robot's end tool maintains continuous contact with the workpiece surface and moves relative to it along a preset trajectory, requiring the maintenance of a constant contact force while ensuring contact stability.

[0060] Except for the differences in the definition of contact state, construction of control strategy and task objectives, the specific implementation of steps S2 to S6 in this embodiment is the same as that in embodiment one, and will not be repeated here.

[0061] Step S1: Obtain contact information between the robot's end effector and the environment, and construct a contact feature vector.

[0062] In this embodiment, the contact force between the robot end effector and the workpiece surface is obtained in the tool coordinate system. :

[0063] in, Indicates the normal contact force (perpendicular to the workpiece surface). , Indicates tangential contact force. This indicates transpose.

[0064] Step S2: Based on the contact feature vector, the probability distribution of each state in the contact state set is obtained using the contact state recognition model, and the contact state uncertainty is calculated. Then, the contact state uncertainty is normalized to obtain the normalized uncertainty.

[0065] In this embodiment, a set of contact states is defined based on the characteristics of the polishing task. for:

[0066] in, Indicates a non-contact state. This indicates a stable polishing state. This indicates an excessive contact force. This indicates an unstable contact state.

[0067] Step S3: Construct an adaptive policy generation function based on the contact state and normalized uncertainty to generate a control policy.

[0068] In this embodiment, the constructed exploration strategy and execution strategy are specifically as follows: (1) Exploratory control strategy

[0069] When the contact state is unstable or uncertain, an exploratory strategy is adopted to adjust the contact state, which takes the form of small perturbations in the tangential and normal directions:

[0070] in, Used for fine-tuning in the normal direction to restore or maintain contact. and Used for fine-tuning in the tangential direction to find a stable contact area, the fine-tuning amount can be generated by small random perturbation or low-frequency oscillation.

[0071] (2) Execution-oriented control strategy

[0072] When the contact state is clear, perform a grinding operation along the target trajectory:

[0073] in, The velocity is the velocity along the tangential direction of the curved surface. This refers to the normal velocity. The normal velocity is determined through force control. , For the desired contact force, For actual normal contact force, To control the gain.

[0074] Step S4: Dynamically adjust the robot's control parameters based on the control strategy, and implement compliant control execution through a compliant control model.

[0075] In step S5, during the execution of compliant control, a slip error is constructed based on the robot's end-effector position error, and an adaptive update law for the control parameters is designed based on the Lyapunov method to adjust the stiffness and damping parameters online.

[0076] In this embodiment, contact force is adjusted through a compliance control model during the polishing process, and control parameters are adaptively adjusted based on uncertainty. The specific control strategy is as follows: (1) When When the value is large (state is uncertain), reduce stiffness and increase damping; (2) When When the value is relatively small (state is stable), increase the stiffness to ensure trajectory accuracy; (3) When detected At this time, reduce the normal velocity to avoid overload.

[0077] Step S6: During task execution, the contact information is updated in real time, and steps S1 to S5 are executed repeatedly to achieve joint adaptive updates of the control strategy and control parameters until the task termination condition is met.

[0078] In this embodiment, the robot performs a grinding task along a preset curved surface trajectory and updates the contact feature vector in real time, cyclically executing steps S1 to S5. The task terminates when one of the following conditions is met: (1) trajectory execution is completed; (2) the set processing time is reached.

[0079] In summary, the robot fine operation control method based on contact state uncertainty according to the present invention has the following beneficial effects: (1) By constructing a contact state recognition model and introducing a contact state uncertainty assessment mechanism, this invention achieves a quantitative description of the reliability of the contact state, and solves the problem that the prior art lacks a quantitative assessment of the uncertainty of the contact state. (2) By constructing a strategy generation function based on uncertainty, this invention realizes the continuous generation and smooth transition of control strategies, avoids the instability problem caused by traditional strategy switching, and effectively alleviates the problem of untimely or unreasonable control strategy switching. (3) By integrating exploratory control strategy and execution control strategy, this invention enables the robot to have adaptive decision-making ability in uncertain environments, and is suitable for various contact-driven tasks such as assembly and grinding, and has good robustness and adaptability. (4) The present invention combines a compliant control method to realize online adjustment of control parameters, thereby improving the stability and success rate of robot fine operation.

[0080] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for fine-grained robot operation control based on contact state uncertainty, characterized in that, include: Step S1: Obtain contact information between the robot's end effector and the environment, and construct a contact feature vector; Step S2: Based on the contact feature vector, the probability distribution of each state in the contact state set is obtained using the contact state recognition model, and the contact state uncertainty is calculated. Then, the contact state uncertainty is normalized to obtain the normalized uncertainty. Step S3: Construct an adaptive policy generation function based on the contact state and normalized uncertainty to generate a control policy; Step S4: Dynamically adjust the robot's control parameters based on the control strategy, and implement compliant control execution through the compliant control model; Step S5: During the execution of compliant control, a slip error is constructed based on the robot end position error, and an adaptive update law for the control parameters is designed based on the Lyapunov method to adjust the stiffness and damping parameters online. Step S6: During the task execution process, update the contact information in real time and repeat steps S1 to S5 to achieve joint adaptive updates of control strategy and control parameters until the task termination condition is met. In step S2, the set of contact states for: in, , , , , , These are respectively: free space state, initial contact state, edge contact state, eccentric contact state, stuck state, and stable contact state; In step S3, the expression for the adaptive policy generation function is: in, For control strategies, Let be the weighting function with respect to uncertainty. For exploratory control strategies, For execution-oriented control strategies, , To adjust the parameters, To normalize uncertainty; In step S5, the adaptive update law of the control parameters satisfies the following equation: in, This represents the rate of change of the stiffness matrix with respect to time. This represents the rate of change of the damping matrix with respect to time. For positional error, for The derivative with respect to time, This refers to the sliding error; , It is a positive definite symmetric matrix used to adjust the parameter update rate; This is the uncertainty adjustment coefficient. , To adjust the parameters, This is a reference value for uncertainty. Here is the damping reference matrix. This is the error adjustment parameter.

2. The robot fine operation control method based on contact state uncertainty according to claim 1, characterized in that, In step S1, the contact feature vector for: in, This is the generalized contact force vector composed of contact force and torque. for The derivative with respect to time, The end-effector pose vector. This indicates transpose.

3. The robot fine operation control method based on contact state uncertainty according to claim 2, characterized in that, In step S2, the contact state uncertainty is calculated using the following formula. : in, Indicates the first Contact state, For the first The probability distribution of contact states. For the first A function to determine the contact state of a class. For the first A discriminant function for class contact states; The following formula is used to normalize the uncertainty of the contact state: in, The number of contact states.

4. The robot fine operation control method based on contact state uncertainty according to claim 3, characterized in that, The exploratory control strategy is a disturbance control strategy generated along the contact plane direction, which takes the form of periodic or random disturbance motion; the execution control strategy is a control strategy along the target direction, which achieves fine operation through force control or trajectory tracking.

5. The robot fine operation control method based on contact state uncertainty according to claim 4, characterized in that, In step S4, the compliance control model satisfies the following equation: in, The equivalent inertia matrix, Indicates the actual position of the robot's end effector. Indicates the robot's end effector speed. Indicates the acceleration at the robot's end effector. Indicates the desired position. Here is the damping matrix. Here is the stiffness matrix. It is an external contact force.

6. The robot fine operation control method based on contact state uncertainty according to claim 5, characterized in that, In step S5, under the action of the adaptive update law, the constructed Lyapunov function satisfies that its time derivative is non-positive, thereby ensuring system stability.

Images