Obstacle avoidance method and system for dual-arm robot based on weighted potential function and DMP

Abstract

The invention discloses a dual-arm robot obstacle avoidance method and system based on weighted potential function and DMP, which comprises the following steps: obtaining a teaching trajectory of the dual-arm robot, establishing a DMP model of the mechanical arm, inputting the teaching trajectory into the DMP model for learning and parameterizing the DMP model, constructing a weighted combination equipotential function and setting a dynamic potential function based on the equipotential function, adding the negative gradient of the dynamic potential function as a first coupling term into the DMP model, obtaining an obstacle avoidable DMP model, constructing a second coupling term with reference to the position and speed of the right mechanical arm, adding the second coupling term into the obstacle avoidable DMP model of the left mechanical arm, obtaining an autonomous obstacle avoidance DMP model of the left mechanical arm, the obstacle avoidance DMP model of the right mechanical arm and the autonomous obstacle avoidance DMP model of the left mechanical arm together forming an autonomous obstacle avoidance system of the dual-arm robot, and the motion trajectory obtained by the method and the system can avoid collision between each mechanical arm and surrounding obstacles and collision between the two mechanical arms.

Description

Technical Field

[0001] The present invention relates to the technical field of robot arm trajectory planning, and in particular to a dual-arm robot obstacle avoidance method and system based on weighted potential function and DMP. Background Art

[0002] With the rapid development of industrial automation and intelligent manufacturing, the application of robotics, especially dual-arm robots, in production lines has become increasingly important. Dual-arm robots can complete more complex and sophisticated tasks by simulating human two-handed operations. However, with the increasing complexity of application scenarios, such as automated assembly, precision machining, and complex material handling, the challenges faced by dual-arm robots have also increased. In these scenarios, although dual-arm robots are usually demonstrated in an obstacle-free environment, the actual operating environment is often filled with moving operators, other mechanical equipment, and temporary obstacles generated during the execution of tasks. Therefore, when performing tasks, dual-arm robots must not only maintain the motion characteristics of the teaching trajectory, but also have the ability to flexibly avoid obstacles to ensure the safety and efficiency of operations.

[0003] Among many obstacle avoidance technologies, the artificial potential field method combined with dynamic potential function is widely used because of its simplicity and efficiency. This method constructs a virtual potential field so that the robot can avoid obstacles under the guidance of the potential field. However, the traditional potential function method has the problem of local minimum, which may cause the robot to be trapped around obstacles and unable to escape. At the same time, some existing individual obstacle avoidance algorithms, such as the super-quadratic potential function method, are prone to generate paths that are not smooth and compact, which is unacceptable for dual-arm robots that need to perform delicate operations, because it may lead to reduced accuracy during operation, damage to the robot itself and the surrounding environment, and waste of robot workspace. In addition, considering the dynamics of the dual-arm robot operation, the two robotic arms may also collide during movement, and it is incomplete to consider the obstacle avoidance of each arm to the surrounding obstacles separately. Therefore, a dual-arm robot obstacle avoidance method based on DMP (Dynamic Movement Primitives) and dynamic potential function is proposed to achieve the obstacle avoidance of the dual-arm robot system to the surrounding obstacles and the avoidance of collision between the two arms under the premise of considering the dynamic characteristics of the robotic arms. Summary of the invention

[0004] The present invention aims to solve the problems of obstacle avoidance of surrounding obstacles and collision between the two arms of a dual-arm robot during task execution, and designs a dual-arm robot obstacle avoidance method and system based on weighted potential function and DMP.

[0005] On one hand, the present invention provides a dual-arm robot obstacle avoidance method based on weighted potential function and DMP, the method comprising the following steps:

[0006] S1. Obtain the teaching trajectory of the manipulator of the dual-arm robot in an obstacle-free environment, establish an initial DMP model of the manipulator, input the teaching trajectory of the manipulator in the obstacle-free environment into the initial DMP model for learning and obtaining a weight coefficient, and use the weight coefficient to update the initial DMP model to obtain a parameterized DMP model;

[0007] S2. Setting an adaptive change weighting coefficient according to the distance between the robot arm and the nearest obstacle, and constructing a weighted combination potential function according to the adaptive change weighting coefficient, the super quadratic potential function and the Gaussian potential function;

[0008] S3, setting a dynamic potential function according to the weighted combination of potential functions, the current position and speed of the robotic arm, calculating the negative gradient of the dynamic potential function, and adding the negative gradient of the dynamic potential function as the first coupling term to the parameterized DMP model, thereby obtaining an obstacle avoidable DMP model of the robotic arm, the obstacle avoidable DMP model including an obstacle avoidable DMP model of the left robotic arm and an obstacle avoidable DMP model of the right robotic arm;

[0009] S4. Construct a second coupling term based on the van der Pol oscillator, and add the second coupling term to the obstacle avoidance DMP model of the left robotic arm with reference to the position and velocity of the right robotic arm, so as to obtain the autonomous obstacle avoidance DMP model of the left robotic arm. The obstacle avoidance DMP model of the right robotic arm and the autonomous obstacle avoidance DMP model of the left robotic arm together constitute the autonomous obstacle avoidance model of the dual-arm robot;

[0010] S5. Set the starting point and end point of each robotic arm in the real scene, input the starting point and end point of each robotic arm in the real scene and obstacle information into the autonomous obstacle avoidance model of the dual-arm robot for processing, and output the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

[0011] Preferably, the initial DMP model of the robot arm in S1 can be specifically expressed as:

[0012] ;

[0013] in, ;

[0014] In the formula, is the time scaling factor, and Respectively represent the speed and position of the robot at the current moment, and are the stiffness matrix and damping matrix of the DMP model, and Respectively represent the target position and initial position of the robot arm, represents the phase variable, represents the first-order derivative, For the forced term, is the weight, For the A nonlinear radial basis function, , is the total number of nonlinear radial basis functions.

[0015] Preferably, the weighted combination equipotential function in S2 can be specifically expressed as:

[0016] ;

[0017] in, ;

[0018] ;

[0019] ;

[0020] In the formula, is the weighted combination equipotential function, is the equipotential function of the superquadratic potential function, The equipotential function of the Gaussian function, Adaptively change the weighting coefficient, and Respectively represent the current position of the robot arm The coordinate values ​​on the x, y and z axes, , and are real-valued functions that determine the scales of the x, y, and z axes, respectively. is a constant that determines the order of the superquadratic potential function, is the amplitude of the Gaussian distribution, is the mean vector of the Gaussian distribution, is the covariance matrix, is a constant, is the distance between the robot arm and the nearest obstacle, is the distance threshold, is the control coefficient.

[0021] Preferably, the dynamic potential function in S3 can be specifically expressed as:

[0022] ;

[0023] in, ;

[0024] In the formula, represents the dynamic potential function, and is a constant gain, is an equipotential function, is an equipotential function The gradient of is the speed of the robot arm The angle between the position of the robot arm and the direction of the nearest obstacle, Represents the gradient of the equipotential function and the velocity of the robotic arm The standard scalar product between Represents the norm.

[0025] Preferably, the obstacle avoidance DMP model of the robot arm in S3 can be specifically expressed as:

[0026] ;

[0027] in, ;

[0028] In the formula, represents the first coupling term, Represents the dynamic potential function gradient.

[0029] Preferably, in S4, a second coupling term based on a van der Pol oscillator is constructed, and the second coupling term can be specifically expressed as:

[0030] ;

[0031] In the formula, represents the second coupling term, is the oscillation coefficient.

[0032] Preferably, the DMP model of the left robotic arm in S4 that can autonomously avoid obstacles can be specifically expressed as:

[0033] ;

[0034] In the formula, and Represent the speed and position of the left robotic arm respectively, and They represent the stiffness matrix and damping matrix of the DMP model for the left robot arm to avoid obstacles. and Respectively represent the target position and initial position of the left robotic arm, represents the forcing term of the DMP model for the left robot to avoid obstacles, and represent the position and speed of the right robotic arm respectively, represents the second coupling term referenced to the position and velocity of the right manipulator.

[0035] On the other hand, the present invention provides a dual-arm robot obstacle avoidance system, which adopts the dual-arm robot obstacle avoidance method based on weighted potential function and DMP to avoid obstacles. The system includes a dual-arm robot and an image acquisition unit and an autonomous obstacle avoidance model which are arranged on the dual-arm robot body and are connected in sequence, wherein:

[0036] The image acquisition unit is used to obtain the position information of surrounding obstacles and send the obstacle position information to the autonomous obstacle avoidance model;

[0037] The autonomous obstacle avoidance model receives and processes the preset initial position and target position of each robotic arm as well as the obstacle position information to obtain the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

[0038] The dual-arm robot obstacle avoidance method and system based on weighted potential function and DMP first obtains the teaching trajectory of the dual-arm robot's mechanical arm in an obstacle-free environment, establishes an initial DMP model of the mechanical arm, inputs the teaching trajectory into the initial DMP model for learning, obtains the weight coefficient and parameterizes the initial DMP model, then constructs a weighted combination of equipotential functions and sets a dynamic potential function on this basis, adds the negative gradient of the dynamic potential function as the first coupling term to the parameterized DMP model, obtains an obstacle avoidable DMP model of the mechanical arm, specifically including an obstacle avoidable DMP model of the left mechanical arm and an obstacle avoidable DMP model of the right mechanical arm, then constructs a second coupling term based on the van der Pol oscillator, and uses the position of the right mechanical arm as the first coupling term. The position and speed are used as references, and the second coupling term is added to the obstacle avoidance DMP model of the left robotic arm to obtain the autonomous obstacle avoidance DMP model of the left robotic arm. The obstacle avoidance DMP model of the right robotic arm and the autonomous obstacle avoidance DMP model of the left robotic arm together constitute the autonomous obstacle avoidance model of the dual-arm robot. Finally, the autonomous obstacle avoidance model is applied to the real scene, and the starting point and end point of each robotic arm in the real scene and the obstacle information are input into the autonomous obstacle avoidance model of the dual-arm robot for processing, and the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other is output. The motion trajectory of the two robotic arms obtained by this method can avoid the collision of each robotic arm with the surrounding obstacles and the collision between the two robotic arms. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of a dual-arm robot obstacle avoidance method based on weighted potential function and DMP in one embodiment of the present invention;

[0040] Figure 2 is a diagram showing the effect of the robot arm avoiding obstacles when the teaching trajectory in two-dimensional space is a straight line and the obstacle is close in one embodiment of the present invention, wherein: Figure 2 (a) is the overall obstacle avoidance effect diagram of the robot arm when the teaching trajectory in two-dimensional space is a straight line. Figure 2(b) is the obstacle avoidance effect diagram of the robot arm on the x-axis and y-axis when the teaching trajectory is a straight line in two-dimensional space;

[0041] Figure 3 is a diagram showing the obstacle avoidance effect of a robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is far away in one embodiment of the present invention, wherein: Figure 3 (a) is the overall obstacle avoidance effect diagram of the robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is far away. Figure 3 (b) is a diagram showing the obstacle avoidance effect of the robot arm on the x-axis and y-axis when the teaching trajectory in two-dimensional space is a curve and the obstacle is far away;

[0042] Figure 4 is a diagram showing the obstacle avoidance effect of a robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is relatively close in one embodiment of the present invention, wherein: Figure 4 (a) is the overall obstacle avoidance effect diagram of the robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is close. Figure 4 (b) is a diagram showing the obstacle avoidance effect of the robot arm on the x-axis and y-axis when the teaching trajectory in two-dimensional space is a curve and the obstacle is close;

[0043] Figure 5 FIG. 1 is a diagram showing the effect of preventing the two arms of a dual-arm robot from colliding with each other in a three-dimensional space in one embodiment of the present invention, wherein Figure 5 (a) is a diagram showing the effect of preventing the two arms from colliding with each other when the target point of the left arm of the dual-arm robot in three-dimensional space is located at a certain point on the motion trajectory of the right arm. Figure 5 (b) is a diagram showing the effect of preventing the two arms from colliding with each other when the target point of the left arm and the target point of the right arm of the dual-arm robot in three-dimensional space are the same. DETAILED DESCRIPTION

[0044] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0045] See also Figure 1 , a dual-arm robot obstacle avoidance method based on weighted potential function and DMP includes the following steps:

[0046] S1. Obtain the teaching trajectory of the manipulator of the dual-arm robot in an obstacle-free environment, establish an initial DMP model of the manipulator, input the teaching trajectory of the manipulator in the obstacle-free environment into the initial DMP model for learning and obtain the weight coefficient, and use the weight coefficient to update the initial DMP model to obtain a parameterized DMP model.

[0047] Specifically, the teaching trajectory includes the position and speed of the left robotic arm in an obstacle-free environment, and the position and speed of the right robotic arm in an obstacle-free environment. The initial DMP (Dynamic Movement Primitives) model of the robotic arm is established. The DMP model is based on the spring-damper system and uses nonlinear force terms to learn and approximate the teaching trajectory to obtain the weight coefficient.

[0048] Furthermore, the initial DMP model of the robot arm can be expressed as:

[0049] (1)

[0050] In the formula, is the time scaling factor that controls the duration of the motion, and Respectively represent the speed and position of the robot at the current moment, and are diagonal matrices, corresponding to the stiffness matrix and damping matrix of the DMP model, respectively. and Respectively represent the target position and initial position of the robot arm, represents the first-order derivative.

[0051] In order to ensure that the DMP model eventually converges to the target , assuming and The relationship between them is: , is a phase variable that changes from 0 to , Will decrease from 1 to 0. It can be specifically expressed by the following canonical system equation:

[0052] (2)

[0053] represents the forcing term, which is a linear combination of N nonlinear radial basis functions and is used to adjust the DMP model to reproduce the skills learned from the teaching trajectory. It can be expressed as:

[0054] (3)

[0055] in, (4)

[0056] In the formula, For the A nonlinear radial basis function, and Respectively represent Nonlinear Radial Basis Function The center and height of , , is the total number of nonlinear radial basis functions, are weights, learned from the teaching trajectory.

[0057] The position and velocity parameters of the teaching trajectory are input into the DMP model, and the weights are determined by local weighted regression. , the process is as follows:

[0058] (5)

[0059] In the formula,

[0060] (5) -1

[0061] (5) -2

[0062] (5) -3

[0063] (5) -4

[0064] In the formula, is the coefficient matrix, To teach the trajectory, is the time point of teaching trajectory, It's the forgetting factor. is the coefficient matrix Middle The transpose of a column vector, Indicates The inverse covariance matrix of the prediction error is used to update the weights, is the initial value, , is the identity matrix, Track duration for the taught track.

[0065] Get weight Then, the forcing term can be calculated using formula (3): , and selectively set the starting position and target position of the robot arm. By solving the DMP model, a trajectory similar to the teaching trajectory motion style can be obtained.

[0066] S2. An adaptive change weighting coefficient is set according to the distance between the robot arm and the nearest obstacle, and a weighted combination potential function is constructed according to the adaptive change weighting coefficient, the super quadratic potential function and the Gaussian potential function.

[0067] Specifically, the equipotential function of the superquadratic potential function is constructed based on the superquadratic potential function:

[0068] (6)

[0069] In the formula, is the equipotential function of the superquadratic potential function, , and are real-valued functions that determine the scales of the x, y, and z axes, respectively. is the order of the superquadratic potential function and is a constant.

[0070] Superquadratic potential functions are used in obstacle avoidance strategies due to their shape adaptability. By adjusting the parameters of potential functions such as superquadratic potential functions, superquadratic potential fields can better match obstacles of different shapes and provide more accurate obstacle avoidance guidance for the robot arm. However, the complexity and potential sharp edges of superquadratic potential functions may cause unnecessary sharp turns in the obstacle avoidance path, thus affecting the smoothness and coherence of the motion.

[0071] Construct the equivalent potential function of the Gaussian potential function based on the Gaussian potential function:

[0072] (7)

[0073] In the formula, is the equipotential function of the Gaussian potential function, is the amplitude of the Gaussian distribution, is the mean vector of the Gaussian distribution, representing the center of the distribution, is the covariance matrix, which controls the shape and direction of the Gaussian distribution, is a constant that determines the height of the equipotential surfaces of the Gaussian potential function.

[0074] Gaussian potential function is widely used in obstacle avoidance algorithms due to its mathematical properties and smoothness. The Gaussian potential field is centered on the obstacle and radiates outward, which can guide the robot to smoothly bypass the obstacle. However, the rapid decay characteristics and symmetry of the Gaussian function show limitations when dealing with obstacles with complex shapes or uneven distribution, especially when the obstacle shape does not match the Gaussian potential field, which may lead to an obstacle avoidance path that is not compact enough or inefficient.

[0075] In order to improve the obstacle avoidance efficiency and adaptability of the DMP model, it is also necessary to set a weighting coefficient. If the weighting coefficient is set to a fixed value, it may cause some sudden and uncompact obstacle avoidance behaviors. At the same time, the specific value of the weighting coefficient required to be set for different obstacles is also uncertain. Therefore, in this embodiment, an adaptive weighting coefficient is set according to the distance between the robot arm and the nearest obstacle. , which can be expressed as:

[0076] (8)

[0077] In the formula, is the adaptive change weighting coefficient, , is the distance between the robot arm and the center of the nearest obstacle, It is the distance threshold that defines the center point of weight change, by setting The value of can be used to adapt to obstacles of different shapes and sizes while allowing the adaptive weighting coefficient to be controlled The changes are more concentrated on specific input values. is the control coefficient that controls the steepness of weight change.

[0078] The equipotential function of the superquadratic potential function is weighted combined with the equipotential function of the Gaussian function to obtain the weighted combined equipotential function :

[0079] (9)

[0080] In the formula, is the weighted combination equipotential function, is the adaptive change weighting coefficient, The equipotential functions of the superquadratic potential function are: is the equipotential function of the Gaussian potential function.

[0081] The obstacle avoidance strategy of the combination of the equation function of the super-quadratic potential function and the equation function of the Gaussian potential function aims to utilize the smoothness of the equipotential function of the Gaussian potential function and the shape adaptability of the equipotential function of the super-quadratic potential function, and dynamically adjust the weighting coefficients of the two to adapt to different obstacle avoidance requirements. In particular, at different distances from obstacles, by adaptively adjusting the weighting coefficients, it can not only ensure the smoothness of the path when far away from obstacles, but also provide accurate obstacle avoidance guidance when approaching obstacles.

[0082] S3. Set the dynamic potential function according to the weighted combination of equipotential functions, the current position and speed of the robotic arm, calculate the negative gradient of the dynamic potential function, and add the negative gradient of the dynamic potential function as the first coupling term to the parameterized DMP model, thereby obtaining the obstacle avoidable DMP model of the robotic arm, which includes the obstacle avoidable DMP model of the left robotic arm and the obstacle avoidable DMP model of the right robotic arm.

[0083] Specifically, the DMP model needs to be given an obstacle avoidance function. The artificial potential field method is a commonly used method in the field of robot obstacle avoidance. This method usually does not consider the robot speed or dynamic properties, and often leads to a non-smooth robot obstacle avoidance trajectory. Therefore, in this scheme, a dynamic potential function is constructed based on the weighted combination of equipotential functions, the current position and speed of the robot. The dynamic potential function can be specifically expressed as follows:

[0084] (10)

[0085] in,

[0086] (10) -1

[0087] In the formula, represents the dynamic potential function, and is a constant gain, is an equipotential function, is an equipotential function The gradient of is the speed of the robot arm The angle between the position of the robot arm and the direction of the nearest obstacle, Represents the gradient of the equipotential function and the velocity of the robotic arm The standard scalar product between Represents the norm.

[0088] The negative gradient of the above dynamic potential function (10) is taken as the first coupling term, that is:

[0089] (11)

[0090] In the formula, is the first coupling term.

[0091] The first coupling term (11) is added to the parameterized DMP model (the expressions of the parameterized model and the initial DMP model are the same, but the parameter values ​​are different) to obtain an obstacle avoidance DMP model, which enables the robot arm to have obstacle avoidance function.

[0092] Furthermore, the obstacle avoidance DMP model of the robot arm can be specifically expressed as:

[0093] (12)

[0094] It should be noted that the obstacle avoidable DMP model in the above formula includes the obstacle avoidable DMP model of the left robotic arm and the obstacle avoidable DMP model of the right robotic arm. Specifically, the relevant parameters in formula (12) are replaced with the corresponding parameters of the left robotic arm to obtain the obstacle avoidable DMP model of the left robotic arm, and the relevant parameters in formula (12) are replaced with the corresponding parameters of the right robotic arm to obtain the obstacle avoidable DMP model of the right robotic arm, where:

[0095] The obstacle avoidance DMP model of the left robotic arm can be expressed as:

[0096] (12) -1

[0097] In the formula, and Represent the speed and position of the left robotic arm respectively, and They represent the stiffness matrix and damping matrix of the DMP model for the left robot arm to avoid obstacles. and Respectively represent the target position and initial position of the left robotic arm, represents the phase variable, represents the first-order derivative, Represents the forcing term of the DMP model for the left robot to avoid obstacles.

[0098] The obstacle avoidance DMP model of the right robotic arm can be expressed as:

[0099] (12) -2

[0100] In the formula, and Respectively represent the speed and position of the right robotic arm at the current moment, and are the stiffness matrix and damping matrix of the obstacle avoidable DMP model of the right manipulator, and Respectively represent the target position and initial position of the right robotic arm, represents the phase variable, represents the first-order derivative, is the forcing term of the obstacle avoidable DMP model of the right robot arm.

[0101] After the above settings, both the right and left robotic arms can imitate and generalize the teaching trajectory through their corresponding obstacle avoidance DMP models and have the function of avoiding obstacles in the environment. In addition, setting the phase variables of the left and right robotic arms to be the same is equivalent to the two robotic arms having the same regularization time.

[0102] S4. Construct a second coupling term based on the van der Pol oscillator, and add the second coupling term to the obstacle avoidance DMP model of the left robotic arm with the position and velocity of the right robotic arm as a reference to obtain the autonomous obstacle avoidance DMP model of the left robotic arm. The obstacle avoidance DMP model of the right robotic arm and the autonomous obstacle avoidance DMP model of the left robotic arm together constitute the autonomous obstacle avoidance model of the dual-arm robot.

[0103] Specifically, since the dual-arm robot needs to avoid obstacles around it during operation, a collision may occur between the two arms. In order to avoid a collision between the two arms during operation, a van der Pol oscillator is first introduced in this embodiment, where the expression of the van der Pol oscillator is:

[0104] (13)

[0105] Then, a second coupling term based on the van der Pol oscillator is established. The second coupling term is used to prevent the two manipulators from colliding with each other. The second coupling term can be specifically expressed as:

[0106] (14)

[0107] in, represents the second coupling term, is the oscillation coefficient used to control the oscillation amplitude.

[0108] In the above formula (14), when It is the ordinary simple harmonic oscillation equation; when It is damped vibration, and the amplitude gradually decays to 0; when When the self-excited vibration occurs, the amplitude gradually increases until it reaches a stable value.

[0109] For the second coupling term of formula (14), and Set them as the position and speed of the right robot arm respectively, and take the oscillation coefficient , and then add it to the obstacle avoidance DMP model of the left robot arm to obtain the autonomous obstacle avoidance DMP model of the left robot arm. Through such operation, when the distance between the right robot arm and the left robot arm is too close or a collision may occur, the self-excited vibration effect of the right robot arm will be transmitted to the left robot arm through the dual-arm robot autonomous obstacle avoidance coupling term, so that the left robot arm deviates from the motion trajectory of the right robot arm and avoids a collision between the two robots.

[0110] Furthermore, the DMP model of the left robotic arm in S4 that can autonomously avoid obstacles can be specifically expressed as:

[0111] (15)

[0112] in,

[0113] ;

[0114]

[0115] In the formula, represents the second coupling term referenced to the position and velocity of the right manipulator.

[0116] The autonomous obstacle avoidance DMP model of the left robotic arm has the function of avoiding obstacles in the environment and preventing the right robotic arm from colliding with each other based on imitating and generalizing the teaching trajectory.

[0117] After obtaining the obstacle avoidance DMP model of the right robotic arm (corresponding to formula (12)-2) and the autonomous obstacle avoidance DMP model of the left robotic arm (corresponding to formula (15)), the two are combined to form the autonomous obstacle avoidance model of the dual-arm robot.

[0118] S5. Set the starting point and end point of each robotic arm in the real scene, input the starting point and end point of each robotic arm in the real scene and obstacle information into the autonomous obstacle avoidance model of the dual-arm robot for processing, and output the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

[0119] In one embodiment, a dual-arm robot obstacle avoidance system adopts the dual-arm robot obstacle avoidance method based on weighted potential function and DMP to avoid obstacles. The system includes a dual-arm robot and an image acquisition unit and an autonomous obstacle avoidance model that are arranged on the dual-arm robot body and are sequentially connected, wherein:

[0120] The image acquisition unit is used to obtain the position information of surrounding obstacles and send the obstacle position information to the autonomous obstacle avoidance model;

[0121] The autonomous obstacle avoidance model receives and processes the preset initial position and target position of each robotic arm as well as the obstacle position information to obtain the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

[0122] Specifically, after the autonomous obstacle avoidance model receives the preset initial position and target position of each robotic arm as well as the obstacle position information, for each time step, it first uses the autonomous obstacle avoidance model of the dual-arm robot (the obstacle avoidance DMP model of the right robotic arm (corresponding to formula (12)-2) and the autonomous obstacle avoidance DMP model of the left robotic arm (corresponding to formula (15))) to calculate the acceleration of each robotic arm at the current time step, uses the acceleration to update the velocity, and uses the velocity to update the position of each robotic arm at the current time step, thereby obtaining the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

[0123] For the specific definition of a dual-arm robot obstacle avoidance system, please refer to the definition of the dual-arm robot obstacle avoidance method based on weighted potential function and DMP in the above text, which will not be repeated here.

[0124] Furthermore, the dual-arm robot obstacle avoidance method based on weighted potential function and DMP proposed in the present invention is verified through experiments.

[0125] See also Figure 2 , Figure 2 This is a diagram showing the obstacle avoidance effect of the robot arm when the teaching trajectory in two-dimensional space is a straight line in one embodiment of the present invention, wherein: Figure 2 (a) is the overall obstacle avoidance effect diagram of the robot arm when the teaching trajectory in two-dimensional space is a straight line. Figure 2(b) is the obstacle avoidance effect diagram of the robot arm on the x-axis and y-axis when the teaching trajectory is a straight line in two-dimensional space;

[0126] exist Figure 2 In (a), the teaching trajectory (corresponding to Figure 2 The obstacle-free teaching trajectory in (a) is a straight line, and the starting point and end point are set to be the same as the teaching trajectory, and obstacles are added to test the trajectory imitation and obstacle avoidance functions of a single robot arm. In addition, for comparison, in addition to using the adaptive change weighting coefficient (corresponding to Figure 2 In addition to the adaptive weights in (a), two sets of adaptive change weight coefficients were manually set, corresponding to Figure 2 (a) and :when When , the equipotential function of the superquadratic potential function is dominant and has good adaptability to the obstacle shape, but there are some sudden obstacle avoidance trajectories, which is unfriendly to the movement of the robot arm; when When , the equipotential function of the Gaussian potential function dominates, and the obstacle avoidance trajectory has good smoothness, but there is a problem of low overlap with the teaching trajectory motion, which reduces the performance of DMP. Figure 2 (b), Figure 2 (b) shows the changes in the x-axis and y-axis and the differences from the teaching trajectory in detail. Figure 2 It can be seen that the method proposed in the present invention can combine the shape adaptability of the equipotential function of the superquadratic potential function and the smoothness of the equipotential function of the Gaussian potential function, and can achieve good adaptability and smoothness to the obstacle shape while avoiding the generation of sudden and low-overlapping trajectories.

[0127] Figure 3 This is a diagram showing the obstacle avoidance effect of the robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is far away in one embodiment of the present invention, wherein: Figure 3 (a) is the overall obstacle avoidance effect diagram of the robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is far away. Figure 3 (b) is a diagram showing the obstacle avoidance effect of the robotic arm on the x-axis and y-axis when the taught trajectory in two-dimensional space is a curve and the obstacle is far away.

[0128] exist Figure 3 In (a), the teaching trajectory (corresponding to Figure 3 (a) The obstacle-free teaching trajectory is a curve and the obstacle is far away. Similarly, the same starting point and end point as the teaching trajectory are set, and obstacles are added to test the trajectory imitation and obstacle avoidance functions of a single robot arm. Similarly, for comparison, two sets of adaptive change weighting coefficients are manually set, which are and The results show that when the obstacle is far away and does not affect the normal movement of the robot arm, the robot arm has a good imitation effect on the teaching trajectory.

[0129] Figure 4 This is a diagram showing the obstacle avoidance effect of the robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is close in one embodiment of the present invention, wherein: Figure 4 (a) is the overall obstacle avoidance effect diagram of the robot arm when the teaching trajectory in two-dimensional space is a curve and the obstacle is close. Figure 4 (b) is a diagram showing the obstacle avoidance effect of the robotic arm on the x-axis and y-axis when the taught trajectory in two-dimensional space is a curve and the obstacle is close.

[0130] exist Figure 4 In (a), the teaching trajectory (corresponding to Figure 4 (a) The obstacle-free teaching trajectory is a curve and the obstacle is close. Similarly, the same starting point and end point as the teaching trajectory are set, and obstacles are added to test the trajectory imitation and obstacle avoidance functions of a single robot arm. In addition, for comparison, two sets of adaptive change weighting coefficients are manually set, which are and The results show that the method proposed in the present invention also has good imitation and obstacle avoidance effects for the teaching trajectory of the curve.

[0131] Figure 5 This is a diagram showing the effect of preventing the two arms of a dual-arm robot from colliding with each other in a three-dimensional space in one embodiment of the present invention, wherein Figure 5 (a) is a diagram showing the effect of preventing the two arms from colliding with each other when the target point of the left arm of the dual-arm robot in three-dimensional space is located at a certain point on the motion trajectory of the right arm. Figure 5 (b) is a diagram showing the effect of preventing the two arms from colliding with each other when the target point of the left arm and the target point of the right arm of the dual-arm robot in three-dimensional space are the same.

[0132] exist Figure 5 In order to avoid clutter, the teaching trajectory in an obstacle-free environment is not shown. Figure 5 The lines in are respectively the right arm DMP imitation trajectory, the left arm DMP imitation trajectory and the left arm DMP imitation trajectory to prevent mutual collision. Figure 5 (b) The target point of the left arm is the same as that of the right arm because the left and right arms move in opposite directions. The results show that the DMP imitation trajectory of the left arm without adding the coupling term (i.e., the second coupling term) to prevent mutual collision between the two arms will directly overlap with the DMP imitation trajectory of the right arm, which means that a collision between the left and right arms will occur. The DMP imitation trajectory of the left arm that can autonomously avoid obstacles with the coupling term (i.e., the second coupling term) added will produce an offset at the original position under the influence of the position and speed of the right arm to avoid a collision between the two arms.

[0133] The dual-arm robot obstacle avoidance method and system based on weighted potential function and DMP first obtains the teaching trajectory of the dual-arm robot's mechanical arm in an obstacle-free environment, establishes an initial DMP model of the mechanical arm, inputs the teaching trajectory into the initial DMP model for learning, obtains the weight coefficient and parameterizes the initial DMP model, then constructs a weighted combination of equipotential functions and sets a dynamic potential function on this basis, adds the negative gradient of the dynamic potential function as the first coupling term to the parameterized DMP model, obtains an obstacle avoidable DMP model of the mechanical arm, specifically including an obstacle avoidable DMP model of the left mechanical arm and an obstacle avoidable DMP model of the right mechanical arm, then constructs a second coupling term based on the van der Pol oscillator, and uses the position of the right mechanical arm as the first coupling term. The position and speed are used as references, and the second coupling term is added to the obstacle avoidance DMP model of the left robotic arm to obtain the autonomous obstacle avoidance DMP model of the left robotic arm. The obstacle avoidance DMP model of the right robotic arm and the autonomous obstacle avoidance DMP model of the left robotic arm together constitute the autonomous obstacle avoidance model of the dual-arm robot. Finally, the autonomous obstacle avoidance model is applied to the real scene, and the starting point and end point of each robotic arm in the real scene and the obstacle information are input into the autonomous obstacle avoidance model of the dual-arm robot for processing, and the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other is output. The motion trajectory of the two robotic arms obtained by this method can avoid the collision of each robotic arm with the surrounding obstacles and the collision between the two robotic arms.

[0134] The above is a detailed introduction to the dual-arm robot obstacle avoidance method and system based on weighted potential function and DMP provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the core idea of ​​the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. Obstacle avoidance method for dual-arm robot based on weighted potential function and DMP, It is characterized in that The method comprises the following steps: S1. Obtain the teaching trajectory of the manipulator of the dual-arm robot in an obstacle-free environment, establish an initial DMP model of the manipulator, input the teaching trajectory of the manipulator in the obstacle-free environment into the initial DMP model for learning and obtaining a weight coefficient, and use the weight coefficient to update the initial DMP model to obtain a parameterized DMP model; S2. Setting an adaptive change weighting coefficient according to the distance between the robot arm and the nearest obstacle, and constructing a weighted combination potential function according to the adaptive change weighting coefficient, the super quadratic potential function and the Gaussian potential function; S3, setting a dynamic potential function according to the weighted combination of potential functions, the current position and speed of the robotic arm, calculating the negative gradient of the dynamic potential function, and adding the negative gradient of the dynamic potential function as the first coupling term to the parameterized DMP model, thereby obtaining an obstacle avoidable DMP model of the robotic arm, the obstacle avoidable DMP model including an obstacle avoidable DMP model of the left robotic arm and an obstacle avoidable DMP model of the right robotic arm; S4. Construct a second coupling term based on the van der Pol oscillator, and add the second coupling term to the obstacle avoidance DMP model of the left robotic arm with reference to the position and velocity of the right robotic arm, so as to obtain the autonomous obstacle avoidance DMP model of the left robotic arm. The obstacle avoidance DMP model of the right robotic arm and the autonomous obstacle avoidance DMP model of the left robotic arm together constitute the autonomous obstacle avoidance model of the dual-arm robot; S5. Set the starting point and end point of each robotic arm in the real scene, input the starting point and end point of each robotic arm in the real scene and obstacle information into the autonomous obstacle avoidance model of the dual-arm robot for processing, and output the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

2. The dual-arm robot obstacle avoidance method based on weighted potential function and DMP as claimed in claim 1, It is characterized in that The initial DMP model of the robot arm in S1 can be specifically expressed as: ； in, ; In the formula, is the time scaling factor, and Respectively represent the speed and position of the robot at the current moment, and are the stiffness matrix and damping matrix of the DMP model, and Respectively represent the target position and initial position of the robot arm, represents the phase variable, represents the first-order derivative, For the forced term, is the weight, For the A nonlinear radial basis function, , is the total number of nonlinear radial basis functions.

3. The dual-arm robot obstacle avoidance method based on weighted potential function and DMP as claimed in claim 2, It is characterized in that The weighted combination equipotential function in S2 can be specifically expressed as: ； in, ; ； ； In the formula, is the weighted combination equipotential function, is the equipotential function of the superquadratic potential function, The equipotential function of the Gaussian function, Adaptively change the weighting coefficient, and Respectively represent the current position of the robot arm The coordinate values ​​on the x, y and z axes, , and are real-valued functions that determine the scales of the x, y, and z axes, respectively. is a constant that determines the order of the superquadratic potential function, is the amplitude of the Gaussian distribution, is the mean vector of the Gaussian distribution, is the covariance matrix, is a constant, is the distance between the robot arm and the nearest obstacle, is the distance threshold, is the control coefficient.

4. The dual-arm robot obstacle avoidance method based on weighted potential function and DMP as claimed in claim 3, It is characterized in that The dynamic potential function in S3 can be specifically expressed as: ； in, ; In the formula, represents the dynamic potential function, and is a constant gain, is an equipotential function, is an equipotential function The gradient of is the speed of the robot arm The angle between the position of the robot arm and the direction of the nearest obstacle, Represents the gradient of the equipotential function and the velocity of the robotic arm The standard scalar product between Represents the norm.

5. The dual-arm robot obstacle avoidance method based on weighted potential function and DMP as claimed in claim 4, It is characterized in that The obstacle avoidance DMP model of the robot arm in S3 can be specifically expressed as: ； in, ; In the formula, represents the first coupling term, Represents the dynamic potential function gradient.

6. The dual-arm robot obstacle avoidance method based on weighted potential function and DMP as claimed in claim 5, It is characterized in that In S4, a second coupling term based on the van der Pol oscillator is constructed. The second coupling term can be specifically expressed as: ； In the formula, represents the second coupling term, is the oscillation coefficient.

7. The dual-arm robot obstacle avoidance method based on weighted potential function and DMP as claimed in claim 6, It is characterized in that The DMP model of the left robotic arm in S4 that can autonomously avoid obstacles can be specifically expressed as: ； In the formula, and Respectively represent the speed and position of the left robotic arm, and They represent the stiffness matrix and damping matrix of the DMP model for the left robot arm to avoid obstacles. and Respectively represent the target position and initial position of the left robotic arm, represents the forcing term of the DMP model for the left robot to avoid obstacles, and represent the position and speed of the right robotic arm respectively, represents the second coupling term referenced to the position and velocity of the right manipulator.

8. A dual-arm robot obstacle avoidance system, using the dual-arm robot obstacle avoidance method based on weighted potential function and DMP as described in any one of claims 1 to 7 to avoid obstacles, It is characterized in that The system includes a dual-arm robot, an image acquisition unit and an autonomous obstacle avoidance model which are arranged on the dual-arm robot body and are connected in sequence, wherein: The image acquisition unit is used to obtain the position information of surrounding obstacles and send the obstacle position information to the autonomous obstacle avoidance model; The autonomous obstacle avoidance model receives and processes the preset initial position and target position of each robotic arm as well as the obstacle position information to obtain the motion trajectory of the dual-arm robot that can avoid obstacles and prevent the two robotic arms from colliding with each other.

Images