A method and system for correcting the kinematic parameters of a tandem robot based on remote motion centers and reinforcement learning (DH).

Abstract

This invention relates to a method for correcting the kinematic parameters (DH) of a serial robot based on remote center of motion (RCM) and reinforcement learning. The method involves establishing a DH parameter kinematic model to obtain the kinematic parameters; constructing a Q-leraning model based on the kinematic parameters and other data; modifying the current values ​​of the corresponding kinematic parameters through repeated training, recording the modified values ​​as undetermined values ​​for the next training iteration; determining whether the number of training iterations meets the requirements; if not, cyclically training; if yes, outputting the kinematic parameter with the smallest error as the optimal kinematic parameter value, and saving the Q-table obtained from the training. This method sets fixed RCM points and detects the offset of the RCM points after movement through image detection, eliminating the need for other measuring equipment to measure error values ​​and reducing equipment costs.

Description

Technical Field

[0001] This invention relates to the field of robot control, and more specifically, to a method and system for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning (DH). Background Technology

[0002] Robotic arms can replace humans in dangerous or highly repetitive tasks and are widely used in medical, aerospace, and manufacturing industries. Motion errors are unavoidable in robotic arm movement. Many factors influence these errors, generally categorized into two types: kinematic errors (such as link length deviations and joint offsets) and non-kinematic errors (such as thermal expansion and link deformation caused by ambient temperature). The deviation of actual kinematic parameters from their nominal values, generated during the manufacturing and assembly of the robotic arm, is called kinematic parameter error. Compared to kinematic parameter errors, non-kinematic errors have a smaller impact on mechanical motion control and can be ignored. To improve the accuracy of robotic arm movement, kinematic calibration is necessary before controlling the robotic arm to perform tasks such as obstacle avoidance and object grasping. Kinematic calibration is a critical issue in robotic arm applications, referring to modifying the robotic arm model parameters in the control program through software to make them as close as possible to the actual kinematic parameters.

[0003] Kinematic calibration of a robot typically involves four steps: (1) Modeling: This refers to using mathematical models to describe the structure and motion characteristics of the industrial robot, such as the Denavit-Hartenberg (DH) parametric model, quaternions, and exponential integral methods. (2) Error Measurement: This involves collecting data from the actual robot and using sensor devices or methods to measure information such as the position of each joint and the pose coordinates of the end effector. Examples include laser measurement systems, optical sensors, and stereo vision systems. (3) Parameter Identification: Based on the collected data and the functional relationship between the robot joint angles and the end effector pose, appropriate algorithms are used to identify the parameter values ​​in the kinematic model. Common algorithms include the Levenberg-Marquardt algorithm and the maximum likelihood estimation method. (4) Error Compensation: This involves compensating the identified parameter values ​​to the theoretical values, thus reducing the gap between the theoretical and actual values.

[0004] Existing technologies can use laser trackers to acquire the robot's end effector pose, identify and correct parameter values ​​in the kinematic model to improve the accuracy of the end effector. However, laser trackers are expensive, complex to operate, and require regular calibration. Alternatively, a binocular vision system can be used to acquire the robot's end effector's 3D spatial coordinates after calibrating the camera's intrinsic and extrinsic parameters, achieving error measurement. However, camera calibration introduces new errors, which are superimposed during the matrix calculation of forward and inverse kinematics, making the error model calculation more complex, increasing the difficulty of kinematic parameter calibration, and reducing the accuracy of error correction. For a specific robot model, with a large amount of robot data, a neural network model can be trained to fit the robot's kinematic parameters. However, the workload of data collection, cleaning, model training, and testing is enormous, and this approach is only applicable to a specific robot model. Summary of the Invention

[0005] To overcome at least one of the defects in the prior art, this invention provides a method for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning. This method not only reduces costs and simplifies calculations, but also corrects the kinematic parameters of the serial robot.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning, comprising the following steps:

[0007] Step 1: Establish the DH parameter kinematic model of the serial robot to obtain the robot's kinematic parameters;

[0008] Step 2: Select the kinematic parameters and set the initial values ​​of the kinematic parameters, the number of training iterations T, the initial state S0, and the reward function to establish the Q-leraning model;

[0009] Step 3: Randomly execute any action that modifies the kinematic parameters with probability ε, and execute the action that makes the Q-table take the maximum value in the current state with probability 1-ε. Modify the current value of the corresponding kinematic parameters and record the modified value as the undetermined value and use it as the current value of the kinematic parameters for the next training.

[0010] Step 4: Set the target point position, calculate multiple sets of joint angle solutions for the robot based on inverse kinematics, and use each set of joint angle solutions to control the end effector of the robot to move to the target point.

[0011] Step 5: Use a camera device to take a picture of the robot's actuator end effector and identify and obtain the coordinates of the actuator end effector; calculate the new state S' based on the changes in the coordinates of the actuator end effector and the target point position coordinates after the robot performs RCM, and calculate the reward r obtained by taking this action based on the reward function;

[0012] Step 6: In the new state S', assume that action a' is performed to maximize the Q-table, and update the Q-table using the Markov property;

[0013] Step 7: Determine if the number of training iterations has reached T. If not, repeat steps 3 to 6; if yes, proceed to step 8.

[0014] Step 8: Perform accuracy judgment on all recorded values ​​to be determined, select the values ​​of motion parameters that make the average positioning accuracy of the robot end effector meet the threshold requirement, and take the value of the kinematic parameter with the smallest error as the optimal kinematic parameter value, and save the Q table obtained from training.

[0015] In the above technical solution, modifying the robot's kinematic parameters is taken as the action; the end effector of the robot's actuator is taken as the RCM point, and the two-dimensional coordinates under the image captured by the camera device are taken as the state; the robot is rewarded or punished according to the Euclidean distance deviation of the RCM point under different joint solutions. Based on the above reinforcement learning model, the ε-greedy algorithm is used to update the Q-table. Through continuous training, a kinematic parameter correction model is obtained, which is the trained Q-table, stating that "when the RCM point has a certain deviation state, the corresponding action of modifying the kinematic parameters is taken to minimize the Euclidean distance deviation of the RCM fixed point under different joint solutions."

[0016] Preferably, the Q-leraning model is as follows:

[0017] Action Space: Sets the correction precision value and modification method for kinematic parameters. The modification method is to increase or decrease the correction precision value based on the current value of the kinematic parameter. Each motion parameter, whether increased, decreased, or unchanged, is set as an action in the action space.

[0018] Reward function:

[0019] In the formula, k is a constant, β is the decay factor, Threshold is the preset RCM threshold; Rcm err It represents the coordinate deviation of the RCM point at the execution end point under different joint angle solutions given the current kinematic parameters.

[0020] Preferably, the specific value of ε is:

[0021]

[0022] In the formula, i = 0, 1, 2...T, and T is the number of training iterations.

[0023] Preferably, in step six, the Q table is updated according to the following formula:

[0024] Q(s,a)=Q(s,a)+α(r+γQ(s′,a′)-Q(S,A))

[0025] In the formula, s is the robot's current state, a is the action taken, α and γ are decay factors, r is the reward the robot receives from the environment after taking the action, s' is the state at the next moment, and a' is the action assumed to be taken under s'.

[0026] Preferably, in step two, when selecting kinematic parameters, some kinematic parameters can be modified while assuming that other kinematic parameters remain unchanged; after the modification of some kinematic parameters is completed, the modified kinematic parameters are used as the new kinematic parameters of the robot and remain unchanged, and steps two to eight are repeated to modify the other unmodified parameters until all kinematic parameters are modified.

[0027] Preferably, in step two, when selecting kinematic parameters, some kinematic parameters can be selected for correction while assuming that other kinematic parameters remain unchanged;

[0028] Step eight is followed by step nine, which specifically includes:

[0029] S9.1: After completing the partial kinematic parameter correction, merge the corrected kinematic parameter values ​​into auxiliary correction values, and repeat steps two to eight to correct the auxiliary correction values ​​and at least one uncorrected kinematic parameter.

[0030] S9.2: Repeat S9.1 again until all kinematic parameters have been corrected.

[0031] Preferably, the kinematic parameters are corrected in the following order: first, the kinematic parameters closest to the robot's end effector are corrected.

[0032] Preferably, in step four, Harris corner detection is used to obtain the coordinates of the end of the actuator.

[0033] Preferably, the true values ​​of the kinematic parameters corresponding to the target point position coordinates are obtained by back-deriving different sets of joint angles by setting the target point position. The coordinates of the actuator end effector after the robot moves according to different sets of joint angles correspond to the values ​​of the kinematic parameters with errors. Therefore, the specific formula for calculating the RCM error is as follows:

[0034]

[0035] In the formula, X1 and Z1 are the pixel coordinates of the RCM point at the end of the actuator of the first joint of the robot; X2 and Z2 are the pixel coordinates of the RCM point at the end of the actuator of the second joint of the robot.

[0036] A system for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning, used to implement the aforementioned method for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning; comprising a robot, a camera device for capturing images of the robot's end effector, and a controller electrically connected to the robot and the camera device respectively.

[0037] A computer-readable storage medium for storing a computer program that, when executed by a processor, implements the above-described method for correcting the kinematic parameters of a DH-based serial robot using remote centers of motion and reinforcement learning.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows: This method sets a fixed RCM point and detects the offset of the RCM point after movement by image detection, eliminating the need for other measuring equipment to measure error values ​​and reducing equipment costs; it uses the end of the actuator as a fixed RCM point to constrain it, and combines reinforcement learning to eliminate the need for coordinate system transformation from robot base to end effector, calibration of camera intrinsic and extrinsic parameters, hand-eye calibration, etc., simplifying the calculation process; by reducing parameter calibration time and labor costs and reducing the repetition of parameter calibration, the operational threshold for parameter calibration is lowered. Attached Figure Description

[0039] Figure 1 This is a flowchart of a method for correcting the kinematic parameters of a serial robot (DH) based on remote center of motion and reinforcement learning, according to the present invention. Detailed Implementation

[0040] The accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent. To better illustrate this embodiment, some components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings. The positional relationships described in the drawings are for illustrative purposes only and should not be construed as limiting this patent.

[0041] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "long," and "short" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present patent. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0042] The technical solution of the present invention will be further described in detail below through specific embodiments and in conjunction with the accompanying drawings:

[0043] Example 1

[0044] like Figure 1 The illustration shows an embodiment of a DH kinematic parameter correction method for a serial robot based on remote center of motion and reinforcement learning, comprising the following steps:

[0045] Step 1: Establish the DH parameter kinematic model of the serial robot to obtain the kinematic parameters of the robot; in this embodiment, a 7-axis serial robot is used as an example. The DH parameter kinematic model has four link length parameters: L1, L2, L3, and L4.

[0046] Step 2: Select the kinematic parameters and set the initial values ​​of the kinematic parameters, the number of training iterations T, the initial state S0, and the reward function to establish the Q-leraning model;

[0047] In this embodiment, when selecting kinematic parameters, some kinematic parameters can be modified while assuming that other kinematic parameters remain unchanged; after modifying some kinematic parameters, the other parameters are modified. The modification order of kinematic parameters is to modify the kinematic parameters closer to the robot's end effector first. In this embodiment, L1 and L2 are kinematic parameters closer to the robot's end effector, so L1 and L2 are modified first.

[0048] The Q-leraning model is as follows:

[0049] Action Space: Sets the correction precision value and modification method for kinematic parameters. The modification method is to increase or decrease the correction precision value based on the current value of the kinematic parameter. Each increase, decrease, or no change of kinematic parameter is set as an action in the action space.

[0050] Reward function:

[0051] In the formula, k is a constant, β is the decay factor, Threshold is the preset RCM threshold; Rcm err It represents the coordinate deviation of the RCM point at the execution end point under different joint angle solutions given the current kinematic parameters.

[0052] Taking L1 and L2 as examples, let the RCM point of the end effector be moved to the target point P(X0,Y0,Z0). Using a camera device, the pixel coordinates (x0,z0) of the end effector in the image taken by the industrial camera can represent the position of the RCM point. Let the values ​​corresponding to L1 and L2 be the true kinematic parameter values. Then the values ​​of the kinematic parameters with errors are L1' = L1 + L2. error L2' = L2 + L2error Based on inverse kinematics, the positions of the end RCM points P1(x1,z1) and P2(x2,z2) under the two sets of joint solutions can be calculated. Due to the existence of errors, both sets of pixel coordinate points have corresponding joint angles q1 and q2, but the two coordinate points will not be exactly the same. According to kinematic analysis, the mathematical expression for calculating the RCM error under the two sets of joint solutions is:

[0053]

[0054] In the above formula (x0, z0) are the set coordinates of the end RCM point.

[0055] Under certain conditions, RCMerr will be minimized (set to 0) if and only if the parameters L1' and L2' are the true values ​​L1 and L2, respectively. However, due to the existence of errors, RCMerr is difficult to be 0. In the actual training of this embodiment, RCMerr can be approximated by taking end images with a camera.

[0056]

[0057] Step 3: Randomly execute any action that modifies the kinematic parameters with probability ε, and execute the action that makes the Q-table take the maximum value in the current state with probability 1-ε. Modify the current value of the corresponding kinematic parameters and record the modified value as the undetermined value and use it as the current value of the kinematic parameters for the next training.

[0058] The specific value of ε is:

[0059]

[0060] In the formula, i = 0, 1, 2...T, and T is the number of training iterations.

[0061] Preferably, in step six, the Q table is updated according to the following formula:

[0062] Q(S,a)=Q(S,a)+α(r+γQ(s′,a′)-Q(s,a))

[0063] In the formula, s is the robot's current state, a is the action taken, α and γ are decay factors, r is the reward the robot receives from the environment after taking the action, s' is the state at the next moment, and a' is the action assumed to be taken under s'.

[0064] Step 4: Set the target point position, calculate multiple sets of joint angle solutions for the robot based on inverse kinematics, and use each set of joint angle solutions to control the end effector of the robot to move to the target point.

[0065] Step 5: Take a picture of the robot's actuator end effector using a camera device, and obtain the coordinates of the actuator end effector using Harris corner detection; calculate the new state S' based on the position changes of the actuator end effector coordinates under different joint solutions after the robot performs RCM motion, and calculate the reward r obtained by taking this action based on the reward function;

[0066] Step 6: In the new state S', assume that action a' is performed to maximize the Q-table, and update the Q-table using the Markov property;

[0067] Step 7: Determine if the number of training iterations has reached T. If not, repeat steps 3 to 6; if yes, proceed to step 8.

[0068] Step 8: Perform accuracy assessment on all recorded undetermined values, select the kinematic parameter values ​​that ensure the robot's end-effector average positioning accuracy meets the threshold requirement, and then take the kinematic parameter value that minimizes the RCM error RCMerr as the optimal kinematic parameter value. Save the trained Q-table. After multiple training iterations, the undetermined values ​​obtained can all be used as parameter values ​​for L1 and L2. The desired effect of the correction is to minimize the RCM error value RCMerr, that is, to make L1 as equal to L1' as much as possible, and L2 as equal to L2' as much as possible.

[0069] Step 9: S9.1: After completing the correction of some kinematic parameters, merge the corrected kinematic parameter values ​​into auxiliary correction values, and repeat steps 2 to 8 to correct the auxiliary correction values ​​and at least one uncorrected kinematic parameter; S9.2: Repeat step 9.1 until all kinematic parameters have been corrected.

[0070] In this embodiment, after correcting L1 and L2, the corrected values ​​of L1 and L2 are merged into an auxiliary correction value L. * Repeat steps two through eight for L. * The L* value is corrected by L3 to obtain the corrected L3 value. If the corrected L* value is not equal to the combined corrected L1 and L2 value after correcting L1 and L2, the average of the two is taken as the true value of the combined L1 and L2 value. Then, the corrected L1 and L2 values ​​are adjusted according to the ratio of the initial corrected L1 and L2 values. After correcting L1, L2, and L3, the corrected L1, L2, and L3 values ​​are combined into an auxiliary correction value L. ** Repeat steps two through eight for L. ** The L4 value is then corrected, and finally, the L1, L2, L3, and L4 values ​​are corrected. By correcting the kinematic parameters segment by segment, the errors of parameters L1, L2, L3, and L4 can be minimized.

[0071] Another method for correcting other kinematic parameters is provided. After correcting the selected kinematic parameters in step two, the corrected kinematic parameters are used as the robot's new kinematic parameters and remain unchanged. Steps two through eight are then repeated to correct the remaining uncorrected parameters until all kinematic parameters are corrected. That is, after correcting L1 and L2, the corrected values ​​of L1 and L2 are used as the robot's kinematic parameters and remain unchanged. Then, steps two through eight are repeated to correct L3 and L4. This method is faster, but if the corrected values ​​of L1 and L2 differ significantly from the true values, the accuracy of L3 and L4 corrections will decrease. In other words, the errors in L1 and L2 parameters will be added to the corrections of L3 and L4 parameters. Therefore, this method is fast but has lower parameter correction accuracy.

[0072] In this embodiment, modifying the robot's kinematic parameters is taken as the action; the end effector of the robot's actuator is taken as the RCM point, and the two-dimensional coordinates under the image captured by the camera device are taken as the state; the robot is rewarded or punished according to the Euclidean distance deviation of the RCM point under different joint solutions. Based on the above reinforcement learning model, the Q-table is updated using the ε-greedy algorithm. Through continuous training, a kinematic parameter correction model is obtained, which states that "when the RCM point has a certain deviation state, the corresponding action of modifying the kinematic parameters is taken to minimize the Euclidean distance deviation of the RCM fixed point under different joint solutions." This is the trained Q-table.

[0073] The beneficial effects of this embodiment are as follows: This method sets a fixed RCM point and detects the offset of the RCM point after movement through image detection, eliminating the need for other measuring equipment to measure error values ​​and reducing equipment costs; it uses the end of the actuator as a fixed RCM point to constrain it, and combines reinforcement learning to eliminate the need for coordinate system transformation from robot base to end effector, calibration of camera intrinsic and extrinsic parameters, hand-eye calibration, etc., simplifying the calculation process; by reducing parameter calibration time and labor costs, and reducing the repetition of parameter calibration, it lowers the operational threshold for parameter calibration.

[0074] Example 2

[0075] An embodiment of a system for correcting robot kinematic parameters, for performing the DH kinematic parameter correction method for a serial robot based on remote center of motion and reinforcement learning as described in Embodiment 1, includes a robot, a camera device for capturing images of the robot's end effector, and a controller electrically connected to the robot and the camera device, respectively.

[0076] Example 3

[0077] A computer-readable storage medium for storing a computer program that, when executed by a processor, implements the method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning, as described in Embodiment 1.

[0078] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning, characterized in that, Includes the following steps: Step 1: Establish the DH parameter kinematic model of the serial robot to obtain the robot's kinematic parameters; Step 2: Select the kinematic parameters and set the initial values ​​of the kinematic parameters, the number of training iterations T, the initial state S0, and the reward function to establish the Q-leraning model; Step 3: Randomly execute any action that modifies the kinematic parameters with probability ε, and execute the action that makes the Q-table take the maximum value in the current state with probability 1-ε. Modify the current value of the corresponding kinematic parameters and record the modified value as the undetermined value and use it as the current value of the kinematic parameters for the next training. Step 4: Set the target point position, calculate multiple sets of joint angle solutions for the robot based on inverse kinematics, and use each set of joint angle solutions to control the end effector of the robot to move to the target point. Step 5: Take a picture of the robot's actuator end effector using a camera device and identify the coordinates of the actuator end effector; calculate the new state S' based on the position changes of the actuator end effector coordinates under different joint solutions after the robot performs RCM motion, and calculate the reward r obtained by taking this motion based on the reward function; Step 6: In the new state S', assume that action a' is performed to maximize the Q-table, and update the Q-table using the Markov property; Step 7: Determine if the number of training iterations has reached T. If not, repeat steps 3 to 6; if yes, proceed to step 8. Step 8: Perform accuracy judgment on all recorded undetermined values, select the values ​​of motion parameters that make the average positioning accuracy of the robot end effector meet the threshold requirements, and take the kinematic parameter with the smallest RCM error value as the optimal kinematic parameter value, and save the Q table obtained from training.

2. The method for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning according to claim 1, characterized in that, The Q-leraning model is as follows: Action Space: Sets the correction precision value and modification method for kinematic parameters. The modification method is to increase or decrease the correction precision value based on the current value of the kinematic parameter. Each motion parameter, whether increased, decreased, or unchanged, is set as an action in the action space. Reward function: reward In the formula, k is a constant. The attenuation factor, Threshold, is the preset RCM threshold; It represents the coordinate deviation of the RCM point at the execution end point under different joint angle solutions given the current kinematic parameters.

3. The method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning according to claim 2, characterized in that, The specific value of ε is: In the formula, i = 0, 1, 2...T, and T is the number of training iterations.

4. The method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning according to claim 3, characterized in that, In step six, the Q table is updated according to the following formula: In the formula, s is the robot's current state, a is the action taken, α and γ are decay factors, r is the reward the robot receives from the environment after taking the action, s' is the state at the next moment, and a' is the action assumed to be taken under s'.

5. The method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning according to claim 2, characterized in that, In step two, when selecting kinematic parameters, some kinematic parameters are selected for correction while assuming that other kinematic parameters remain unchanged. After the correction of some kinematic parameters is completed, the corrected kinematic parameters are used as the new kinematic parameters of the robot and remain unchanged. Steps two to eight are executed again to correct the other uncorrected parameters until all kinematic parameters are corrected.

6. The method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning according to claim 2, characterized in that, In step two, when selecting kinematic parameters, some kinematic parameters are selected for correction while assuming that other kinematic parameters remain unchanged; Step eight is followed by step nine, which specifically includes: S9.1: After completing the partial kinematic parameter correction, merge the corrected kinematic parameter values ​​into auxiliary correction values, and repeat steps two to eight to correct the auxiliary correction values ​​and at least one uncorrected kinematic parameter. S9.2: Repeat S9.1 again until all kinematic parameters have been corrected.

7. The method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning according to claim 5 or 6, characterized in that, The kinematic parameters are corrected in the following order: first, correct the kinematic parameters closest to the robot's end effector.

8. The method for correcting the kinematic parameters of a DH robot based on remote center of motion and reinforcement learning according to claim 1, characterized in that, The true values ​​of the kinematic parameters corresponding to the target point's position coordinates are obtained by inversely deducing different sets of joint angles by setting the target point's position. The coordinates of the actuator's end effector after the robot moves according to different sets of joint angles correspond to the values ​​of the kinematic parameters with errors. Therefore, the specific formula for calculating the RCM error is as follows: In the formula, , These are the pixel coordinates of the RCM point at the end of the actuator of the first joint of the robot. , These are the pixel coordinates of the RCM point at the end of the actuator of the second joint of the robot.

9. The method for correcting the kinematic parameters of a DH (Democratic Hierarchical Organization) robot based on remote center of motion and reinforcement learning according to claim 1, characterized in that, In step four, Harris corner detection is used to obtain the coordinates of the end of the actuator.

10. A system for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning, characterized in that, The method for correcting the kinematic parameters of a serial robot based on remote center of motion and reinforcement learning, as described in any one of claims 1-9, includes a robot, a camera device for capturing images of the robot's end effector, and a controller electrically connected to the robot and the camera device, respectively.

Images