Compliant control method for multiple components of composite robot

Abstract

The invention relates to the technical field of composite robot component control, and discloses a composite robot multi-component compliance control method which comprises the following steps: performing deformation analysis on an image of a to-be-grabbed target to obtain the deformation degree of the to-be-grabbed target; target control parameters of a mechanical arm in the composite robot are generated; a self-adaptive hydraulic impedance compliant control objective function is constructed, and hydraulic impedance parameters of the composite robot are obtained through calculation; and the torque of the composite robot is generated through the hydraulic impedance parameters, and torsional force adjustment is conducted on a mechanical arm in the composite robot. The deformation degree of the to-be-grabbed target is calculated based on the edge gradient change value of the continuous time image, the target control parameters for preventing the to-be-grabbed target from being deformed and completing the operation task as soon as possible are selected in a reinforcement learning mode and converted into the torque of the composite robot, the mechanical arm is subjected to torsional force adjustment, and the mechanical arm is controlled. The stability of force applied by the mechanical arms is improved, and compliant control over the multiple mechanical arms in the composite robot is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of composite robot component control, and in particular to a composite robot multi-component compliant control method. Background Art

[0002] The rapid development of robotics technology has penetrated into many fields such as manufacturing, service, medical, and agriculture. Among them, as a new type of robot form, the compliant control of multiple components of composite robots is one of the key technologies to achieve precise control and safe operation of robots. Traditional rigid control methods are difficult to adapt to complex and changing environments, especially when collaborating with humans or contacting uncertain objects, which can easily cause collisions or operational failures. In order to meet the complex needs of composite robots in industrial collaboration and life service scenarios, compliant control technology has emerged. By introducing the control coordination of force and position, compliant control technology enables robots to actively adapt to changes in the external environment during operation, thereby achieving efficient, safe, and stable task execution. Existing compliant control methods are mainly based on traditional rigid combination adjustment control. The motion and operation coupling relationship between components significantly increases the control difficulty, resulting in the failure of the flexible control effect. Summary of the invention

[0003] In view of this, the present invention proposes a multi-component compliant control method for a composite robot, which enables the robot to perceive and respond to feedback in the environment by integrating multi-scale visual detection technology, achieve flexible contact and operation, and reduce potential damage to the environment and mission objectives.

[0004] To achieve the above object, the present invention provides a method for compliant control of multiple components of a composite robot, comprising the following steps:

[0005] S1: Real-time acquisition of the image of the target to be grasped by the composite robot, and deformation analysis to obtain the deformation degree of the target to be grasped;

[0006] S2: constructing a composite robot multi-component collaborative control model to generate target control parameters of the manipulator in the composite robot, wherein the composite robot multi-component collaborative control model takes the deformation degree of the target to be grasped and the current control parameters of the manipulator as inputs and takes the target control parameters for reducing the deformation degree as outputs;

[0007] S3: Construct an adaptive hydraulic impedance compliance control objective function to calculate the hydraulic impedance parameters of the composite robot;

[0008] S4: The torque of the composite robot is generated by using the hydraulic impedance parameter, and the torsional force of the mechanical arm in the composite robot is adjusted.

[0009] As a further improvement method of the present invention:

[0010] Optionally, the composite robot is composed of:

[0011] The composite robot is composed of multiple parts, each part corresponds to a mechanical arm, the mechanical arm is composed of multiple mechanical arm joints and an end effector, the mechanical arm joints can perform rotation and translation operations, the end effector is a device installed at the end of the mechanical arm, the end effector deploys a gripper, and can directly interact with the target to be grasped, wherein the interactive behavior is the grasping behavior;

[0012] The target to be grasped is an object to be operated by the composite robot, and the target to be grasped has an operation task.

[0013] Optionally, the step S1 of performing deformation analysis on the image of the target to be captured includes:

[0014] The image of the target to be captured at time t is I t ;

[0015] Image I t The deformation analysis process is as follows:

[0016] For image I t Perform binarization to obtain image I t The binary pixel matrix A t ;

[0017] Binarized pixel matrix A t Perform filtering and noise reduction processing to obtain the pixel matrix A′ after filtering and noise reduction processing t ;

[0018] Calculate the pixel matrix A′ t The gradient magnitude of each pixel in the image is calculated, and the pixels with gradient magnitude higher than the preset gradient threshold are regarded as edge pixels of the target to be captured. The pixel area surrounded by the edge pixels is regarded as the image position of the target to be captured at time t, and the radius R of the pixel area is calculated. t , extract the SIFT feature descriptor of edge pixels;

[0019] The SIFT feature descriptor of the edge pixel at time t is calculated similarly with the SIFT feature descriptor of the edge pixel at time t-1, and two edge pixels whose similarity is higher than a preset similarity threshold are regarded as a group of successfully matched edge pixels, i.e., edge pixels at the same contour edge position of the target to be captured at different times, and the similarity calculation method is cosine similarity;

[0020] Based on the gradient change value of each group of edge pixels and the radius of the pixel area, the deformation degree of the target to be grasped at time t is calculated.

[0021]

[0022] in:

[0023] R t Represents the radius of the pixel area of ​​the target to be grasped at time t;

[0024] (grad t (d),grad t―1 (d)) represents the gradient magnitude of the dth group of edge pixels, D represents the number of groups of edge pixels, and grad t (d),grad t―1 (d) represents the gradient amplitudes of the edge pixels at time t and the edge pixels at time t-1 in the d-th group of edge pixels.

[0025] Optionally, constructing a multi-component collaborative control model of a composite robot in step S2 includes:

[0026] The multi-component collaborative control model of the composite robot includes an action strategy device and a state value device. The action strategy device includes all control parameters that the robot arm can take. The control parameters of the robot arm are the position vector, velocity vector and contact force vector of the end effector of the robot arm. The state value device is used to evaluate the value of all control parameters in the action strategy device based on the current state of the robot arm, and select the control parameter with the highest value as the target control parameter of the robot arm.

[0027] The current state of the robotic arm includes the current control parameters of the robotic arm and the deformation degree of the grasped target;

[0028] The values ​​of different control parameters in the state valuer under different states are parameters to be solved, which are solved based on the reward function;

[0029] The target control parameters of the manipulators in the composite robot are generated by using the multi-component cooperative control model of the composite robot, where the current control parameter of the nth manipulator in the composite robot at time t is f t (n), the target control parameters of the nth manipulator at time t generated by the multi-component collaborative control model of the composite robot are:

[0030] Optionally, the reward function is expressed as:

[0031] Reward(f|U)=w1·λ1(f|U)+w2·λ2(f|U)―w3·λ3(f|U)

[0032]

[0033] λ2(f|U)=1―δ(f|U)

[0034] λ3(f|U)=Max(f|U)

[0035] in:

[0036] Reward(f|U) represents the value of the control parameter f in state U, λ1(f|U) represents the operation value, λ2(f|U) represents the deformation value, λ3(f|U) represents the force balance value, and w1, w2, and w3 represent the value weights of the three values;

[0037] δ(f|U) represents the normalized result of the maximum deformation degree of the target to be grasped after taking the control parameter f;

[0038] Max(f|U) represents the maximum contact force difference between all the manipulators that contact the target to be grasped after taking the control parameter f.

[0039] Optionally, constructing the adaptive hydraulic impedance compliance control objective function in step S3 includes:

[0040] The adaptive hydraulic impedance compliance control objective function takes the current control parameters and target control parameters of the robot arm as inputs and the hydraulic impedance parameters as outputs. The expression of the adaptive hydraulic impedance compliance control objective function taking the current control parameters and target control parameters of the nth robot arm at time t as inputs is:

[0041]

[0042] in:

[0043] θ1, θ2, θ3 are the inertia matrix, damping matrix and stiffness matrix respectively; the heuristic algorithm is used to solve the inertia matrix, damping matrix and stiffness matrix;

[0044] f t (n) are the current control parameters of the nth robot at time t, is the target control parameter of the nth robot at time t; are the actual position vector, actual velocity vector, and actual contact force vector of the end effector of the nth robot arm at time t, The expected position vector, expected velocity vector and expected contact force vector of the end effector of the nth manipulator at time t output by the multi-component collaborative control model of the composite robot;

[0045] is the hydraulic impedance parameter of the nth manipulator in the composite robot at time t, where n∈[1,N], and N represents the number of manipulators in the composite robot.

[0046] Optionally, the process of using a heuristic algorithm to solve the inertia matrix, damping matrix and stiffness matrix is:

[0047] Initialize H groups of matrix parameters, where each group of matrix parameters includes inertia matrix, damping matrix and stiffness matrix. The initialized hth group of matrix parameters is Denotes the matrix parameters θ in turn 0 Inertia matrix, damping matrix and stiffness matrix in (h);

[0048] Set the maximum number of iterations for each set of matrix parameters to Max, then the rth iteration result of the hth set of matrix parameters is θ r r∈[1,Max];

[0049] Construct a fitness function with matrix parameters as input, substitute the matrix parameters into the fitness function, and obtain the fitness function value of the matrix parameters. The matrix parameters θ r (h) The corresponding fitness function value is Loss(θ r (h)):

[0050]

[0051] in:

[0052] Loss(·) represents the fitness function;

[0053] ||·||2 represents the L2 norm;

[0054] F(m) represents the true hydraulic impedance parameter of the mth group of training samples, F(m; θ r (h)) represents the matrix parameter θ of the mth group of training samples r (h) Simulated hydraulic impedance parameters;

[0055] The fitness weight of each group of matrix parameters is calculated, and all matrix parameters obtained in the r-th iteration are weighted. The weighted mean of the 10% matrix parameters with the largest fitness weights is selected as the elite center obtained in the r-th iteration. The matrix parameters are iterated based on the elite center until the maximum number of iterations is reached. The matrix parameters with the lowest fitness function value after reaching the maximum number of iterations are selected as the solution results of the inertia matrix, damping matrix and stiffness matrix.

[0056] Optionally, in step S4, the torque of the composite robot is generated by using the hydraulic impedance parameter to adjust the torsion force of the mechanical arm in the composite robot, including:

[0057] The torque of the nth manipulator in the composite robot at time t is G t (n):

[0058]

[0059] in:

[0060] T stands for transpose;

[0061] G t (n) is a matrix with Q rows and 3 columns, where Q represents the number of joints of each manipulator in the composite robot, and G t The qth line in (n) is the torsional force of the nth manipulator in the composite robot in the Cartesian space in the three coordinate axis directions, q∈[1,Q];

[0062] J represents the Jacobian matrix of 3 rows and Q columns, which describes the relationship between the joint space velocity and the linear velocity and angular velocity of the end effector in Cartesian space;

[0063] The torsional force of the Q joints of the manipulator in the composite robot is extracted from the torque, and the Q joints of the manipulator are driven to twist according to the torsional force, thereby changing the movement direction and speed of the joints, and then controlling the end effector of the manipulator, so that the manipulator can complete the operation task of the target to be grasped without damaging the target to be grasped.

[0064] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0065] A memory storing at least one instruction;

[0066] Communication interface, enabling electronic equipment to communicate; and

[0067] The processor executes the instructions stored in the memory to implement the above-mentioned multi-component compliant control method of the composite robot.

[0068] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned multi-component compliant control method of a composite robot.

[0069] Compared with the prior art, the present invention proposes a multi-component compliance control method for a composite robot, which has the following advantages:

[0070] First, this scheme proposes a control parameter generation method based on target deformation. The image of the target to be grasped is collected, and a double filtering method is used for filtering and denoising. The noise information in the image is filtered out and the edge information is retained to the maximum extent. The edge gradient amplitude of the image is obtained, and the deformation degree of the target to be grasped is calculated based on the edge gradient change value of the image at continuous moments. The reinforcement learning method is used to select the target control parameters that avoid the deformation of the target to be grasped and complete the operation task as soon as possible. In the reinforcement learning process, a reward function is constructed based on whether the operation task of the target to be grasped is completed, the deformation of the target to be grasped, and the force balance degree of the robotic arm. The lower the force balance degree of the robotic arm, the more balanced the forces applied by multiple robotic arms are, so as to avoid unnecessary deformation or damage.

[0071] At the same time, this scheme proposes a compliant control method for the robot arm, constructs an adaptive hydraulic impedance compliant control objective function, and uses a heuristic algorithm to solve it. In the solution process, the weighted center based on the optimal 10% matrix parameters is used as the elite center. Since only 10% of the matrix parameters are selected, when the global optimum falls into a local extreme value, the remaining matrix parameters still have the ability to update, then the weighted center will continue to update, and the algorithm will not fall into the local extreme value trap. This group guidance method based on the 10% optimal matrix parameters replaces the individual guidance method of the optimal individual, which can better maintain the population diversity, obtain the hydraulic impedance parameters of the composite robot, use the hydraulic impedance parameters to generate the torque of the composite robot, and adjust the torsion force of the robot arm in the composite robot, so that the robot arm can apply a stable force when in contact with an object, and can reduce the applied force when in contact with a person or other objects, thereby improving safety and realizing the compliant control of multiple robot arms in the composite robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic flow chart of a method for compliant control of multiple components of a composite robot provided in one embodiment of the present invention.

[0073] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0074] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0075] The embodiment of the present application provides a method for compliant control of multiple components of a composite robot. The execution subject of the method for compliant control of multiple components of a composite robot includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for compliant control of multiple components of a composite robot can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0076] Reference Figure 1 , Embodiment 1 of the present invention is:

[0077] A method for compliant control of multiple components of a composite robot comprises the following steps:

[0078] S1: Collect images of the target to be grasped by the composite robot in real time, and perform deformation analysis to obtain the deformation degree of the target to be grasped.

[0079] The composite robot is composed of:

[0080] The composite robot is composed of multiple components, each component corresponds to a robotic arm, and the robotic arm is composed of multiple robotic arm joints and an end effector. The robotic arm joints can perform rotation and translational movements. The end effector is a device installed at the end of the robotic arm. The end effector deploys a gripper and can directly interact with the target to be grasped, wherein the interactive behavior is the grasping behavior.

[0081] The deformation analysis of the image of the target to be grasped includes:

[0082] The image of the target to be captured at time t is I t ;

[0083] Image I t The deformation analysis process is as follows:

[0084] For image I t Perform binarization to obtain image I t The binary pixel matrix A t ;

[0085] Binarized pixel matrix A t Perform filtering and noise reduction processing to obtain the pixel matrix A′ after filtering and noise reduction processing t As a preferred embodiment of the present invention, a double filtering method is used to perform filtering and noise reduction processing, filter the noise information in the image and greatly retain the edge information. The binary pixel matrix A t The filtering and denoising formula for the pixel in the xth row and yth column is:

[0086]

[0087] in:

[0088] A′ t (x,y) represents the binary pixel matrix A t The filtering and denoising result of the pixel in the x-th row and y-th column, x∈[1,X],y∈[1,Y], X represents the number of pixel rows in the binary pixel matrix, and Y represents the number of pixel columns in the binary pixel matrix;

[0089] β(x,y;i,j) represents the binary pixel matrix A t where the filter coefficient of the pixel in the xth row and yth column at the neighborhood coefficient (i, j), β1(x, y; i, j) represents the neighborhood position filter coefficient, β2(x, y; i, j) represents the neighborhood pixel filter coefficient, σ1 represents the neighborhood position control parameter, σ2 represents the neighborhood pixel control parameter, ||·|| represents the L1 norm, and exp(·) represents the exponential function with a natural constant as the base;

[0090] Calculate the pixel matrix A′ t The gradient magnitude of each pixel in the image is calculated, and the pixels with gradient magnitude higher than the preset gradient threshold are regarded as edge pixels of the target to be captured. The pixel area surrounded by the edge pixels is regarded as the image position of the target to be captured at time t, and the radius R of the pixel area is calculated. t , extract the SIFT feature descriptor of edge pixels; specifically, based on the calculation of gradients in the horizontal and vertical directions by the traditional 2×2 convolution template, add the calculation of gradients in the directions of 45 degrees and 135 degrees to improve the continuity and comprehensiveness of the gradients;

[0091] The SIFT feature descriptor of the edge pixel at time t is calculated similarly with the SIFT feature descriptor of the edge pixel at time t-1, and two edge pixels whose similarity is higher than a preset similarity threshold are regarded as a group of successfully matched edge pixels, i.e., edge pixels at the same contour edge position of the target to be captured at different times, and the similarity calculation method is cosine similarity;

[0092] Based on the gradient change value of each group of edge pixels and the radius of the pixel area, the deformation degree of the target to be grasped at time t is calculated.

[0093]

[0094] in:

[0095] R t Represents the radius of the pixel area of ​​the target to be grasped at time t;

[0096] (grad t(d),grad t―1 (d)) represents the gradient magnitude of the dth group of edge pixels, D represents the number of groups of edge pixels, and grad t (d),grad t―1 (d) represents the gradient amplitudes of the edge pixels at time t and the edge pixels at time t-1 in the d-th group of edge pixels.

[0097] S2: Construct a multi-component collaborative control model of a composite robot to generate target control parameters of the manipulator in the composite robot. The multi-component collaborative control model of the composite robot takes the deformation degree of the target to be grasped and the current control parameters of the manipulator as input, and takes the target control parameters for reducing the deformation degree as output.

[0098] The step S2 constructs a multi-component collaborative control model of a composite robot, including:

[0099] The multi-component collaborative control model of the composite robot includes an action strategy device and a state value device. The action strategy device includes all control parameters that the robot arm can take. The control parameters of the robot arm are the position vector, velocity vector and contact force vector of the end effector of the robot arm. The state value device is used to evaluate the value of all control parameters in the action strategy device based on the current state of the robot arm, and select the control parameter with the highest value as the target control parameter of the robot arm.

[0100] The current state of the robotic arm includes the current control parameters of the robotic arm and the deformation degree of the grasped target;

[0101] As an embodiment of the present invention, the position vector, velocity vector and contact force vector are all in the form of a matrix with three rows and one column, corresponding to the position, velocity and contact force of the end effector on the target to be grasped in the three coordinate axes XYZ directions;

[0102] The values ​​of different control parameters in the state valuer under different states are parameters to be solved, which are solved based on the reward function;

[0103] The target control parameters of the manipulators in the composite robot are generated by using the multi-component cooperative control model of the composite robot, where the current control parameter of the nth manipulator in the composite robot at time t is f t (n), the target control parameters of the nth manipulator at time t generated by the multi-component collaborative control model of the composite robot are:

[0104] The expression of the reward function is:

[0105] Reward(f|U)=w1·λ1(f|U)+w2·λ2(f|U)―w3·λ3(f|U)

[0106]

[0107] λ2(f|U)=1―δ(f|U)

[0108] λ3(f|U)=Max(f|U)

[0109] in:

[0110] Reward(f|U) represents the value of the control parameter f in state U, λ1(f|U) represents the operation value, λ2(f|U) represents the deformation value, λ3(f|U) represents the force balance value, and w1, w2, and w3 represent the value weights of the three values;

[0111] δ(f|U) represents the normalized result of the maximum deformation degree of the target to be grasped after taking the control parameter f;

[0112] Max(f|U) represents the maximum contact force difference between all the manipulators that contact the target to be grasped after the control parameter f is adopted. As a preferred embodiment of the present invention, a reward function is constructed based on whether the operation task of the target to be grasped is completed, the deformation of the target to be grasped, and the force balance degree of the manipulator, wherein the lower the force balance degree of the manipulator, the more balanced the forces applied by the multiple manipulators are, so as to avoid unnecessary deformation or damage, and obtain the target control parameters that avoid deformation of the target to be grasped and complete the operation task as soon as possible; specifically, the proximal strategy optimization method can be adopted to solve the value of different control parameters in different states based on the reward function.

[0113] S3: Construct an adaptive hydraulic impedance compliant control objective function to calculate the hydraulic impedance parameters of the composite robot.

[0114] The adaptive hydraulic impedance compliance control objective function is constructed in step S3, including:

[0115] The adaptive hydraulic impedance compliance control objective function takes the current control parameters and target control parameters of the robot arm as inputs and the hydraulic impedance parameters as outputs. The expression of the adaptive hydraulic impedance compliance control objective function taking the current control parameters and target control parameters of the nth robot arm at time t as inputs is:

[0116]

[0117] in:

[0118] θ1, θ2, θ3 are the inertia matrix, damping matrix and stiffness matrix respectively; the heuristic algorithm is used to solve the inertia matrix, damping matrix and stiffness matrix;

[0119] f t (n) are the current control parameters of the nth robot at time t, is the target control parameter of the nth robot at time t; are the actual position vector, actual velocity vector, and actual contact force vector of the end effector of the nth robot arm at time t, The expected position vector, expected velocity vector and expected contact force vector of the end effector of the nth manipulator at time t output by the multi-component collaborative control model of the composite robot;

[0120] is the hydraulic impedance parameter of the nth manipulator in the composite robot at time t, where n∈[1,N], and N represents the number of manipulators in the composite robot.

[0121] The process of solving the inertia matrix, damping matrix and stiffness matrix using the heuristic algorithm is as follows:

[0122] Initialize H groups of matrix parameters, where each group of matrix parameters includes inertia matrix, damping matrix and stiffness matrix. The initialized hth group of matrix parameters is Denotes the matrix parameters θ in turn 0 Inertia matrix, damping matrix and stiffness matrix in (h);

[0123] Set the maximum number of iterations for each set of matrix parameters to Max, then the result of the rth iteration of the hth set of matrix parameters is r∈[1,Max];

[0124] Construct a fitness function with matrix parameters as input, substitute the matrix parameters into the fitness function, and obtain the fitness function value of the matrix parameters. The matrix parameters θ r (h) The corresponding fitness function value is Loss(θ r (h)):

[0125]

[0126] in:

[0127] Loss(·) represents the fitness function;

[0128] ||·||2 represents the L2 norm;

[0129] F(m) represents the true hydraulic impedance parameter of the mth group of training samples, F(m; θ r (h)) represents the matrix parameter θ of the mth group of training samples r (h) simulated hydraulic impedance parameters; specifically:

[0130]

[0131] f1 (m),f 2 (m),f 3 (m) are the actual position vector, actual velocity vector and actual contact force vector of the end effector of the robot in the mth group of training samples, respectively. is the expected position vector, expected velocity vector and expected contact force vector of the end effector of the manipulator output by the multi-component collaborative control model of the composite robot in the mth group of training samples;

[0132] The fitness weight of each group of matrix parameters is calculated, and all matrix parameters obtained in the r-th iteration are weighted. The weighted mean of the 10% matrix parameters with the largest fitness weights is selected as the elite center obtained in the r-th iteration. The matrix parameters are iterated based on the elite center until the maximum number of iterations is reached. The matrix parameters with the lowest fitness function value after reaching the maximum number of iterations are selected as the solution results of the inertia matrix, damping matrix and stiffness matrix.

[0133] The matrix parameter θ r The fitness weight of (h) is w r (h):

[0134]

[0135] in:

[0136] Loss r Represents the minimum fitness function value of the H group matrix parameters obtained in the rth iteration;

[0137] The elite center obtained in the rth iteration is

[0138] The matrix parameter θ r The iterative formula of (h) is:

[0139]

[0140] in:

[0141] rand(0,1) represents a random number between 0 and 1.

[0142] As a preferred embodiment of the present invention, a weighted center based on the optimal 10% of matrix parameters is used as the elite center. Since only 10% of the matrix parameters are selected, when the global optimum falls into a local extreme value, the remaining matrix parameters still have the ability to be updated, and the weighted center will continue to be updated, and the algorithm will not fall into the local extreme value trap. This group guidance method based on the 10% optimal matrix parameters replaces the individual guidance method of the optimal individual, and can better maintain population diversity.

[0143] S4: The torque of the composite robot is generated by using the hydraulic impedance parameter, and the torsional force of the mechanical arm in the composite robot is adjusted.

[0144] In the step S4, the torque of the composite robot is generated by using the hydraulic impedance parameter to adjust the torsion force of the mechanical arm in the composite robot, including:

[0145] The torque of the nth manipulator in the composite robot at time t is G t (n):

[0146]

[0147] in:

[0148] T stands for transpose;

[0149] G t (n) is a matrix with Q rows and 3 columns, where Q represents the number of joints of each manipulator in the composite robot, and G t The qth line in (n) is the torsional force of the nth manipulator in the composite robot in the Cartesian space in the three coordinate axis directions, q∈[1,Q];

[0150] J represents a Jacobian matrix of 3 rows and Q columns, describing the relationship between the joint space velocity and the linear velocity and angular velocity of the end effector in Cartesian space; as an embodiment of the present invention, the Jacobian matrix is ​​a matrix of known matrix values, which is calculated based on the kinematic model of the robot arm;

[0151] The torsional force of the Q joints of the manipulator in the composite robot is extracted from the torque, and the Q joints of the manipulator are driven to twist according to the torsional force, thereby changing the movement direction and speed of the joints, and then controlling the end effector of the manipulator, so that the manipulator can complete the operation task of the target to be grasped without damaging the target to be grasped.

[0152] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0153] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0154] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0155] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for compliant control of multiple components of a composite robot, characterized in that: The method comprises: S1: Real-time acquisition of the image of the target to be grasped by the composite robot, and deformation analysis to obtain the deformation degree of the target to be grasped; S2: constructing a composite robot multi-component collaborative control model to generate target control parameters of the manipulator in the composite robot, wherein the composite robot multi-component collaborative control model takes the deformation degree of the target to be grasped and the current control parameters of the manipulator as inputs and takes the target control parameters for reducing the deformation degree as outputs; S3: Construct an adaptive hydraulic impedance compliance control objective function to calculate the hydraulic impedance parameters of the composite robot; S4: The torque of the composite robot is generated by using the hydraulic impedance parameter, and the torsional force of the mechanical arm in the composite robot is adjusted.

2. A method for compliant control of multiple components of a composite robot as claimed in claim 1, characterized in that: The composite robot is composed of: The composite robot is composed of multiple parts, each part corresponds to a mechanical arm, the mechanical arm is composed of multiple mechanical arm joints and an end effector, the mechanical arm joints can perform rotation and translation operations, the end effector is a device installed at the end of the mechanical arm, the end effector deploys a gripper, and can directly interact with the target to be grasped, wherein the interactive behavior is the grasping behavior; The target to be grasped is an object to be operated by the composite robot, and the target to be grasped has an operation task.

3. A method for compliant control of multiple components of a composite robot as claimed in claim 2, characterized in that: In the step S1, deformation analysis is performed on the image of the target to be grasped, including: The image of the target to be captured at time t is I t ; Image I t The deformation analysis process is as follows: For image I t Perform binarization to obtain image I t The binary pixel matrix A t ; Binarized pixel matrix A t Perform filtering and noise reduction processing to obtain the pixel matrix A′ after filtering and noise reduction processing t ; Calculate the pixel matrix A′ t The gradient magnitude of each pixel in the image is calculated, and the pixels with gradient magnitude higher than the preset gradient threshold are regarded as edge pixels of the target to be captured. The pixel area surrounded by the edge pixels is regarded as the image position of the target to be captured at time t, and the radius R of the pixel area is calculated. t , extract the SIFT feature descriptor of edge pixels; The SIFT feature descriptor of the edge pixel at time t is calculated similarly with the SIFT feature descriptor of the edge pixel at time t-1, and two edge pixels whose similarity is higher than a preset similarity threshold are regarded as a group of successfully matched edge pixels, i.e., edge pixels at the same contour edge position of the target to be captured at different times, and the similarity calculation method is cosine similarity; Based on the gradient change value of each group of edge pixels and the radius of the pixel area, the deformation degree of the target to be grasped at time t is calculated. in: R t Represents the radius of the pixel area of ​​the target to be grasped at time t; (grad t (d),grad t―1 (d)) represents the gradient magnitude of the dth group of edge pixels, D represents the number of groups of edge pixels, and grad t (d),grad t―1 (d) represents the gradient amplitudes of the edge pixels at time t and the edge pixels at time t-1 in the d-th group of edge pixels.

4. A method for compliant control of multiple components of a composite robot as claimed in claim 1, characterized in that: The step S2 constructs a multi-component collaborative control model of a composite robot, including: The multi-component collaborative control model of the composite robot includes an action strategy device and a state value device. The action strategy device includes all control parameters that the robot arm can take. The control parameters of the robot arm are the position vector, velocity vector and contact force vector of the end effector of the robot arm. The state value device is used to evaluate the value of all control parameters in the action strategy device based on the current state of the robot arm, and select the control parameter with the highest value as the target control parameter of the robot arm. The current state of the robotic arm includes the current control parameters of the robotic arm and the deformation degree of the grasped target; The values ​​of different control parameters in the state valuer under different states are parameters to be solved, which are solved based on the reward function; The target control parameters of the manipulators in the composite robot are generated by using the multi-component cooperative control model of the composite robot, where the current control parameter of the nth manipulator in the composite robot at time t is f t (n), the target control parameters of the nth manipulator at time t generated by the multi-component collaborative control model of the composite robot are:

5. A method for compliant control of multiple components of a composite robot as claimed in claim 4, characterized in that: The expression of the reward function is: Reward(f|U)=w1·λ1(f|U)+w2·λ2(f|U)―w3·λ3(f|U) λ2(f|U)=1―δ(f|U) λ3(f|U)=Max(f|U) in: Reward(f|U) represents the value of the control parameter f in state U, λ1(f|U) represents the operation value, λ2(f|U) represents the deformation value, λ3(f|U) represents the force balance value, and w1, w2, and w3 represent the value weights of the three values; δ(f|U) represents the normalized result of the maximum deformation degree of the target to be grasped after taking the control parameter f; Max(f|U) represents the maximum contact force difference between all the manipulators that contact the target to be grasped after taking the control parameter f.

6. A method for compliant control of multiple components of a composite robot as claimed in claim 1, characterized in that: The adaptive hydraulic impedance compliance control objective function is constructed in step S3, including: The adaptive hydraulic impedance compliance control objective function takes the current control parameters and target control parameters of the robot arm as inputs and the hydraulic impedance parameters as outputs. The expression of the adaptive hydraulic impedance compliance control objective function taking the current control parameters and target control parameters of the nth robot arm at time t as inputs is: in: θ1, θ2, θ3 are the inertia matrix, damping matrix and stiffness matrix respectively; the heuristic algorithm is used to solve the inertia matrix, damping matrix and stiffness matrix; f t (n) are the current control parameters of the nth robot at time t, is the target control parameter of the nth robot at time t; are the actual position vector, actual velocity vector, and actual contact force vector of the end effector of the nth robot arm at time t, The expected position vector, expected velocity vector and expected contact force vector of the end effector of the nth manipulator at time t output by the multi-component collaborative control model of the composite robot; is the hydraulic impedance parameter of the nth manipulator in the composite robot at time t, where n∈[1,N], and N represents the number of manipulators in the composite robot.

7. A method for compliant control of multiple components of a composite robot as claimed in claim 6, characterized in that: The process of solving the inertia matrix, damping matrix and stiffness matrix using the heuristic algorithm is as follows: Initialize H groups of matrix parameters, where each group of matrix parameters includes inertia matrix, damping matrix and stiffness matrix. The initialized hth group of matrix parameters is Denotes the matrix parameters θ in turn 0 Inertia matrix, damping matrix and stiffness matrix in (h); Set the maximum number of iterations for each set of matrix parameters to Max, then the result of the rth iteration of the hth set of matrix parameters is Construct a fitness function with matrix parameters as input, substitute the matrix parameters into the fitness function, and obtain the fitness function value of the matrix parameters. The matrix parameters θ r (h) The corresponding fitness function value is Loss(θ r (h)): in: Loss(·) represents the fitness function; ||·||2 represents the L2 norm; F(m) represents the true hydraulic impedance parameter of the mth group of training samples, F(m; θ r (h)) represents the matrix parameter θ of the mth group of training samples r (h) Simulated hydraulic impedance parameters; The fitness weight of each group of matrix parameters is calculated, and all matrix parameters obtained in the r-th iteration are weighted. The weighted mean of the 10% matrix parameters with the largest fitness weights is selected as the elite center obtained in the r-th iteration. The matrix parameters are iterated based on the elite center until the maximum number of iterations is reached. The matrix parameters with the lowest fitness function value after reaching the maximum number of iterations are selected as the solution results of the inertia matrix, damping matrix and stiffness matrix.

8. A method for compliant control of multiple components of a composite robot as claimed in claim 1, characterized in that: The S4 step includes: The torque of the nth manipulator in the composite robot at time t is G t (n): in: T stands for transpose; G t (n) is a matrix with Q rows and 3 columns, where Q represents the number of joints of each manipulator in the composite robot, and G t The qth line in (n) is the torsional force of the nth manipulator in the composite robot in the Cartesian space in the three coordinate axis directions, q∈[1,Q], n∈[1,N], N represents the number of manipulators in the composite robot; J represents the Jacobian matrix of 3 rows and Q columns, which describes the relationship between the joint space velocity and the linear velocity and angular velocity of the end effector in Cartesian space; The torsional force of the Q joints of the manipulator in the composite robot is extracted from the torque, and the Q joints of the manipulator are driven to twist according to the torsional force, thereby changing the movement direction and speed of the joints, and then controlling the end effector of the manipulator, so that the manipulator can complete the operation task of the target to be grasped without damaging the target to be grasped.

Images