Dexterous hand collaborative grabbing control method and system oriented to modeling uncertainty

Abstract

The invention relates to the technical field of robot flexible operation and intelligent control, and discloses a modeling uncertainty-oriented dexterous hand collaborative grabbing control method and system, and the method comprises the following steps: S1, building a dexterous hand system dynamics model; s2, virtual coupling is introduced among multiple fingers of the dexterous hand system to construct a communication topological structure; a collaborative grabbing error is defined, and a corresponding potential energy function is constructed; s3, designing a gradient descent distributed control law based on time-varying gain; s4, motion gravity change is processed through a gravity item integral compensator, and kinematics parameter uncertainty is compensated based on a time-varying gain Jacobian matrix adaptive law; and S5, integrating the distributed control law in the step S3 and the adaptive processing mechanism in the step S4, so that the dexterous hand system realizes collaborative grabbing control within the preset time of the user. The method is accurate in control and flexible in response, and can realize efficient, robust and quick response control of the grabbing task in collaborative operation of small-batch customized products.

Description

Technical Field

[0001] The present invention relates to the technical field of robot flexible operation and intelligent control, and in particular to a dexterous hand collaborative grasping control method and system oriented to modeling uncertainty. Background Art

[0002] As a core pillar of the modern industrial system, the intelligent transformation of manufacturing is of strategic significance for enhancing a country's industrial competitiveness. Currently, industrial robots are widely used in high-end intelligent manufacturing fields such as shipbuilding and aerospace components, significantly improving production efficiency and process consistency. However, in flexible manufacturing and assembly scenarios, especially for small-batch customized products with complex structures, shapes, and dimensional variations, traditional rigid robotic systems have significant limitations in terms of degree of freedom configuration and environmental adaptability.

[0003] Multi-finger dexterous hands, due to their flexible structure and diverse operational capabilities, have become a research hotspot in the field of intelligent manufacturing. As a next-generation end-effector, they can effectively address flexible assembly requirements in unstructured environments, demonstrating excellent task adaptability and operational generalization. However, it should be noted that multi-finger dexterous hands still face key technical challenges in achieving complex tasks: the system must simultaneously meet operational robustness and transient responsiveness requirements under environmental uncertainty.

[0004] Existing research has attempted solutions based on physical models and data-driven approaches. While these approaches can address the problem of stable grasping, their practical application is still limited by the following technical bottlenecks: First, the computational complexity of the control algorithm increases exponentially, making it difficult to ensure real-time performance; second, the system's transient response characteristics and ability to adapt to dynamic environments are significantly insufficient; and most importantly, the control system's convergence time has significant uncertainty. These limitations severely restrict the application of multi-fingered dexterous hands in industrial scenarios with high real-time requirements and have become a key factor hindering the development of intelligent manufacturing technology.

[0005] In summary, a new distributed collaborative grasping control method for dexterous hands is urgently needed to achieve efficient, robust, and fast-response control of grasping tasks in collaborative operations of small-batch customized products. Summary of the Invention

[0006] The present invention aims to provide a collaborative grasping control method and system for dexterous hands oriented to modeling uncertainty, with precise control and flexible response. It can achieve efficient, robust and fast response control of grasping tasks in collaborative operations of small batch customized products, and provide verifiable real-time control guarantees and theoretical support for high-demand application scenarios such as flexible manufacturing.

[0007] To achieve the above objectives, the basic scheme provided by the present invention is as follows.

[0008] Option 1 The collaborative grasping control method of dexterous hands oriented to modeling uncertainty includes the following steps: S1, based on the motion state and nonlinear characteristics of the dexterous hand, establish the dynamic model of the dexterous hand system; S2, constructs a communication topology by introducing virtual coupling between multiple fingers of the dexterous hand system; defines the collaborative grasping error to quantify the collaborative state deviation, and constructs the corresponding potential energy function; S3, to address the uncertainty of the dynamic model, a gradient descent distributed control law based on time-varying gains is designed; S4, processes the motion gravity variation through the gravity term integral compensator, and compensates the kinematic parameter uncertainty based on the time-varying gain Jacobian matrix adaptive law; S5 integrates the distributed control law of S3 and the adaptive processing mechanism of S4, enabling the dexterous hand system to achieve collaborative grasping control within the user-preset time.

[0009] Option 2 A dexterous hand collaborative grasping control system for modeling uncertainty, used to execute the dexterous hand collaborative grasping control method for modeling uncertainty as described in Solution 1; comprising: System modeling module, used to establish the dynamic model of the dexterous hand system based on the motion state and nonlinear characteristics of the dexterous hand; A virtual coupling modeling module is used to introduce virtual coupling between multiple fingers of the dexterous hand system to build a communication topology, define the collaborative grasping error and construct the corresponding potential energy function; Distributed control module, used to design a time-varying gain-based gradient descent distributed control law to address the uncertainty of the dynamic model; Gravity compensation module, used to process the change of motion gravity through gravity term integral compensator; Jacobian estimation module, used to compensate for kinematic parameter uncertainty based on the adaptive law of the time-varying gain Jacobian matrix; The control law integration module is used to integrate the output content of the distributed control module, gravity compensation module and Jacobian estimation module, and form the control law of the entire dexterous hand system, so that the dexterous hand system can achieve collaborative grasping control within the time preset by the user.

[0010] The working principle and advantages of the present invention are: The present invention's collaborative grasping control method and system for dexterous hands with modeling uncertainty offers precise control and flexible response, enabling efficient, robust, and rapid response control of grasping tasks in collaborative operations for small-batch customized products. Key points: First, this approach addresses the limitations of traditional force and formation closure (which relies on precise environmental modeling) and data-driven approaches (which require global communication) in collaborative grasping tasks with dexterous hands. By introducing a distributed collaborative control architecture (corresponding to S2) that incorporates a virtual coupling mechanism at the fingertips, this approach achieves collaborative grasping solely through local information exchange between fingers. This effectively avoids the reliance of traditional centralized control on global communication links and high-performance computing resources, significantly improving the system's scalability and real-time performance. Adding new fingers requires only expanding the topological connections, without reconfiguring the global controller. This provides a foundation for the modular design of multi-fingered dexterous hands.

[0011] Second, compared to methods based on fuzzy logic systems (FLS) or system error disturbance compensation (the former suffers from rule-dependence, while the latter suffers from asymptotic convergence limitations), this approach designs a gravity compensation integrator (corresponding to S4) that incorporates a time-varying gain mechanism. A time-varying gain function is embedded in the gravity compensation integrator, allowing the compensation strength to automatically increase over time, synchronously offsetting the dynamic changes in gravity. Furthermore, under the traditional kinematic regression matrix, the time-varying gain is innovatively introduced, constructing an adaptive Jacobian matrix estimation law (corresponding to S4) with a preset time convergence. This approach significantly improves the dexterous hand system's response speed to modeling uncertainty disturbances, enhances grasping accuracy and transient control performance, and provides more robust and adaptive grasping capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of a method flow in an embodiment of a dexterous hand collaborative grasping control method and system for modeling uncertainty according to the present invention; Figure 2 A schematic diagram of the virtual coupling and topological communication relationship of the dexterous hand system in an embodiment of the dexterous hand collaborative grasping control method and system for modeling uncertainty according to the present invention; Figure 3 A schematic diagram of position and velocity simulation of each fingertip of a dexterous hand system according to an embodiment of a dexterous hand collaborative grasping control method and system for modeling uncertainty of the present invention; Figure 4 A schematic diagram of a collaborative grasping trajectory of a dexterous hand system under a preset time in an embodiment of a dexterous hand collaborative grasping control method and system for modeling uncertainty according to the present invention; Figure 5 A schematic diagram of the error of the dexterous hand collaborative grasping system under a preset time in an embodiment of the dexterous hand collaborative grasping control method and system for modeling uncertainty of the present invention; Figure 6 Schematic diagram of the online estimation change of the adaptive Jacobian matrix of the dexterous hand collaborative grasping control method and system embodiment for modeling uncertainty of the present invention. DETAILED DESCRIPTION

[0013] The following is a further detailed description through specific implementation methods: The embodiment is basically as shown in the attached Figure 1 Figure 2 shows a collaborative grasping control method for dexterous hands oriented to modeling uncertainty, including the following steps: S1, based on the motion state and nonlinear characteristics of the dexterous hand, establish a dynamic model of the dexterous hand system.

[0014] The constructed dynamic model of the dexterous hand system is shown as follows: ; in, They represent the first The generalized joint rotation angle, velocity and acceleration of the root finger; The first The inertia matrix of the root finger; Indicates the Coriolis and centrifugal force matrices of the root finger; Indicates the Finger gravity items; Indicates the Dynamical parameters of a known bounded compact set of root fingers; Indicates the Input control torque of the root finger.

[0015] S2, by introducing virtual coupling between multiple fingers of the dexterous hand system to construct a communication topology structure; and define the collaborative grasping error to quantify the collaborative state deviation, and construct the corresponding potential energy function.

[0016] Specifically, if Figure 2 As shown, the undirected graph Represents the communication topology graph between fingers, where the graph satisfies the infinitesimal rigidity property. Vertex set Represents the first Fingers. Indicates the communication connection relationship. For any edge , indicating fingers and There needs to be collaborative interaction between them. and Represented as the number of vertices and edges of an undirected graph. The correlation matrix Elements It can be expressed as: ; in, Represents a vertex For the The end of the edge, Represented as a vertex For the The head of the edge.

[0017] For the collaborative grasping task, the reference configuration of the force closure point is set as , based on which the set of expected stable configurations during the grasping process is defined as .

[0018] The raw grasping error is defined to quantify the collaborative state deviation as: ; in, Fingers representing the dexterous hand system Fingertips and fingers The actual distance of the fingertip, where For fingertips The position coordinates of or 3), Expressed as the second norm of the Euclidean distance, Indicates fingers and fingers The relative distance of the desired formation of fingertips, Represents an undirected graph Middle, Side At the moment The original error.

[0019] Furthermore, the original grasping error is scaled based on the time-varying control gain, which is expressed as: ; in, represents the time-varying gain function, Represents an undirected graph Middle, Side At the moment The original error, Represents an undirected graph Middle, Side At the moment Scaling error.

[0020] Time-varying gain function It can be expressed as: ; Where, represents the control time of the control system of the dexterous hand, denote the order and positive constant of the system respectively, Convergence time preset by the user; is the starting proportional coefficient of the transition interval; , are the upper limit of the gain function and the lower limit of the stable stage respectively.

[0021] Furthermore, based on distributed control theory and gradient optimization descent algorithm, the potential energy function of the entire dexterous hand system is constructed as: ; in, represents the total potential energy of the entire dexterous hand system, Represents an undirected graph The number of edges in the fingertip and fingertips Number of interactive connections); Represents an undirected graph Middle, Side Scaling error.

[0022] S3, to address the uncertainty of the dynamic model, a gradient descent distributed control law based on time-varying gains is designed.

[0023] The gradient descent distributed control law based on time-varying gain is expressed as: ; in, The first The distributed control law of the whole finger, Denoted as the first design gain parameter, represents the generalized adaptive estimated kinematic Jacobian matrix, Expressed as the scaling error for The gradient value of represents the second design control gain, Indicates the The joint speed of each finger, is the third control gain parameter, Expressed as Gravity compensation for each finger.

[0024] Further, Expressed as the scaling error for The gradient value can be expressed as: ; in, represents the time-varying gain function, is expressed as a normalized incidence matrix, is a diagonal matrix, Represents scaling error The stacked vector form.

[0025] S4, handles the motion gravity variation through the gravity term integral compensator, and compensates the kinematic parameter uncertainty based on the time-varying gain Jacobian matrix adaptive law.

[0026] Specifically, in this embodiment, based on the traditional internal integral compensator and the time-varying gain function, a gravity compensation law under time-varying gain is designed, that is, the gravity term integral compensator, which is expressed as: ; in, is the third control gain parameter, Expressed as The time-varying gain function at time t, Expressed as Gravity compensation for each finger, Expressed as Finger control input.

[0027] Based on the kinematic regression matrix, time-varying gain mechanism, and adaptive control theory, the adaptive law based on the time-varying gain Jacobian matrix is ​​designed, which can be expressed as: ; in, The first The update rate of the parameter estimates of each finger, represents the time-varying gain function, Indicates the The linear regression matrix of the fingers, Indicates the The gradient of the potential energy function of each finger, represents the fourth design gain parameter, Indicates the The generalized joint angles of the fingers, Indicates the The kinematic parameter estimation vector of each finger, Indicates the The joint velocity vectors of each finger.

[0028] S5 integrates the distributed control law of S3 and the adaptive processing mechanism of S4, enabling the dexterous hand system to achieve collaborative grasping control within the user-preset time.

[0029] Considering the dexterity of the hand The topological structure diagram between fingers when the dynamic model parameters are uncertain Satisfy infinitesimal rigidity and select appropriate gain parameters. , represents a set of real numbers; for the reference configuration The resulting force-closure configuration , the preset time control law (i.e. the control law of the entire dexterous hand system) can be obtained as follows: ; in, The first The total control input vector of each finger; Represent the first, second, third, and fourth design control gain parameters respectively; represents the generalized adaptive estimated kinematic Jacobian matrix, Indicates the The gradient vector of the finger potential energy function, Indicates the The generalized joint angles of the fingers, Expressed as The update rate of the gravity compensation term for each finger, Expressed as The actual gravity of each finger, Indicates the The joint velocity vectors of the fingers, Indicates the The update rate of the parameter estimates of each finger, represents the time-varying gain function, Indicates the The kinematic regression matrix of each finger, Indicates the The kinematic parameter estimation vector of each finger.

[0030] This embodiment further provides a dexterous hand collaborative grasping control system facing modeling uncertainty, which is used to execute the above-mentioned dexterous hand collaborative grasping control method facing modeling uncertainty; comprising: The system modeling module is used to establish a dynamic model of the dexterous hand system based on its motion state and nonlinear characteristics. Specifically, this module defines its generalized joint angles, velocities, and accelerations based on the motion state and nonlinear characteristics of the dexterous hand, and constructs a dynamic model of the dexterous hand system that includes uncertain parameters such as the inertia matrix, Coriolis force term, and gravity term.

[0031] The virtual coupling modeling module is used to introduce virtual coupling between multiple fingers of the dexterous hand system to build a communication topology, define the collaborative grasping error, and construct a corresponding potential energy function. Specifically, this module introduces virtual coupling between the fingertips, constructs a communication topology, and defines the collaborative grasping error as the deviation between the actual and expected distances between the fingertips. This potential energy function is then used to measure the energy state of the system.

[0032] The distributed control module is used to design a gradient descent distributed control law based on time-varying gains to address the uncertainty of the dynamic model and achieve rapid convergence of the grasping error.

[0033] The gravity compensation module is used to process the change of motion gravity through the gravity term integral compensator and dynamically correct the gravity deviation in the control input.

[0034] The Jacobian estimation module is used to compensate for kinematic parameter uncertainties based on the time-varying gain Jacobian matrix adaptive law to improve modeling accuracy and system robustness.

[0035] The control law integration module is used to integrate the output content of the distributed control module, gravity compensation module and Jacobian estimation module, and form the control law of the entire dexterous hand system, so that the dexterous hand system can achieve collaborative grasping control within the time preset by the user.

[0036] This embodiment provides a dexterous hand collaborative grasping control method and system for modeling uncertainty, which has precise control and flexible response. It can achieve efficient, robust and fast response control of grasping tasks in collaborative operations of small batch customized products, and provide verifiable real-time control guarantee and theoretical support for high-demand application scenarios such as flexible manufacturing.

[0037] In addition, in order to prove the effectiveness of this scheme, a simulation experiment was carried out through Matlab2024a for verification, as follows: The Barret Hand BHB series 3D dexterous hand is used as the simulation object. This dexterous hand has a three-fingered structure, with each finger containing two joints. Therefore, the dexterous hand system has a total of 6 independent degrees of freedom.

[0038] For each finger The joint state is defined as Indicates the The joint angle vectors of the fingers, represents the connecting rod length vector.

[0039] In the grasping task, the target object is set as a regular triangular pyramid with a side length of 0.4m and closed points distributed at three vertices. The topological relationship between the fingertip and the contact point can be expressed by the association matrix , specifically expressed as follows: ; In the simulation, the coordinate origin of the three-finger dexterous hand is set as , the initial joint angles of each finger are set to: 、 、 , and the initial joint velocity is set to 0. , the initial kinematic parameter estimates are set as: .

[0040] The simulation settings are as follows: the total simulation time is set to , the sampling period is , the user sets the system convergence time .

[0041] Other relevant simulation parameters are shown in Table 1.

[0042] Table 1 Single finger system parameters

[0043] The control gain parameter is set to 、 .

[0044] For the time-varying gain function, set ,in represent the order and constant term of the nonlinear system respectively.

[0045] To avoid gain function divergence, set the maximum gain , the minimum gain in steady state is , Used to adjust the time span of the safe transition interval.

[0046] The simulation results are as follows Figure 3 、 Figure 4 、 Figure 5 and Figure 6 As shown. Among them, Figure 3 The joint angles and velocities of all the dexterous hand's fingers are displayed, where a time-varying gain function effectively suppresses fluctuations in angles and velocities. The results show that each finger can successfully reach the target position within the user-set time and achieve rapid convergence of velocity to zero.

[0047] Figure 4 The trajectory path from the initial fingertip position to the target position during grasping is intuitively depicted.

[0048] Figure 5 The dynamic changes of the system grasping error under the preset time constraint are given.

[0049] The simulation results show that the preset time convergence mechanism proposed in this scheme can ensure that the grasping error converges to zero accurately within the preset time, fully verifying the high precision and fast response performance of the control method of this scheme.

[0050] Figure 6 The adaptive Jacobian matrix parameter estimation process is demonstrated. The results show that the preset time adaptive Jacobian matrix designed in this scheme can accurately estimate the kinematic parameters within the user-preset time, significantly enhancing the robustness and adaptability of the dexterous hand system under modeling uncertainty.

[0051] The above is only an embodiment of the present invention. Common knowledge such as the specific structure and characteristics of the scheme is not described in detail here. Ordinary technicians in the relevant field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the guidance of this application. Some typical well-known structures or well-known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A collaborative grasping control method for dexterous hands oriented to modeling uncertainty, characterized by: The following steps are involved: S1, based on the motion state and nonlinear characteristics of the dexterous hand, establish the dynamic model of the dexterous hand system; S2, constructs a communication topology by introducing virtual coupling between multiple fingers of the dexterous hand system; defines the collaborative grasping error to quantify the collaborative state deviation, and constructs the corresponding potential energy function; S3, to address the uncertainty of the dynamic model, a gradient descent distributed control law based on time-varying gains is designed; S4, processes the motion gravity variation through the gravity term integral compensator, and compensates the kinematic parameter uncertainty based on the time-varying gain Jacobian matrix adaptive law; S5 integrates the distributed control law of S3 and the adaptive processing mechanism of S4, enabling the dexterous hand system to achieve collaborative grasping control within the user-preset time.

2. The collaborative grasping control method for dexterous hands oriented to modeling uncertainty according to claim 1 is characterized in that: In S1, the dynamic model of the dexterous hand system is as follows: ； in, They represent the first The generalized joint rotation angle, velocity and acceleration of the root finger; The first The inertia matrix of the root finger; Indicates the Coriolis and centrifugal force matrices of the root finger; Indicates the Finger gravity items; Indicates the Dynamical parameters of a known bounded compact set of root fingers; Indicates the Input control torque of the root finger.

3. The collaborative grasping control method for dexterous hands oriented to modeling uncertainty according to claim 1 is characterized in that: In S2, the collaborative grasping error is: ； in, Fingers representing the dexterous hand system Fingertips and fingers The actual distance of the fingertip, where For fingertips The location coordinates of Expressed as the second norm of the Euclidean distance, Indicates fingers and fingers The relative distance of the desired formation of fingertips, Represents an undirected graph Middle, Side At the moment The original error.

4. The dexterous hand collaborative grasping control method for modeling uncertainty according to claim 1 is characterized in that: In S2, it also includes: scaling the original grasping error based on the time-varying control gain, expressed as: ； in, represents the time-varying gain function, Represents an undirected graph Middle, Side At the moment The original error, Represents an undirected graph Middle, Side At the moment Scaling error.

5. The dexterous hand collaborative grasping control method for modeling uncertainty according to claim 4 is characterized in that: In S2, it also includes: Based on distributed control theory and gradient optimization descent algorithm, the potential energy function of the entire dexterous hand system is constructed as: ； in, represents the total potential energy of the entire dexterous hand system, Represents an undirected graph The number of edges in ; Represents an undirected graph Middle, Side Scaling error.

6. The dexterous hand collaborative grasping control method for modeling uncertainty according to claim 1 is characterized in that: In S3, the gradient descent distributed control law based on time-varying gain is expressed as: ； in, The first The distributed control law of the whole finger, Denoted as the first design gain parameter, represents the generalized adaptive estimated kinematic Jacobian matrix, Expressed as the scaling error for The gradient value of represents the second design control gain, Indicates the The joint speed of each finger, is the third control gain parameter, Expressed as Gravity compensation for each finger.

7. The dexterous hand collaborative grasping control method for modeling uncertainty according to claim 1 is characterized in that: The gravity term integral compensator is expressed as: ； in, is the third control gain parameter, Expressed as The time-varying gain function at time t, Expressed as Gravity compensation for each finger, Expressed as Finger control input.

8. The dexterous hand collaborative grasping control method for modeling uncertainty according to claim 1 is characterized in that: The adaptive law based on the time-varying gain Jacobian matrix is ​​expressed as: ； in, The first The update rate of the parameter estimates of each finger, represents the time-varying gain function, Indicates the The linear regression matrix of the fingers, Indicates the The gradient of the potential energy function of each finger, represents the fourth design gain parameter, Indicates the The generalized joint angles of the fingers, Indicates the The kinematic parameter estimation vector of each finger, Indicates the The joint velocity vectors of each finger.

9. The dexterous hand collaborative grasping control method for modeling uncertainty according to claim 1 is characterized in that: In S5, the distributed control law of S3 and the adaptive processing mechanism of S4 are integrated to form the control law of the entire dexterous hand system: ； in, The first The total control input vector of each finger; Represent the first, second, third, and fourth design control gain parameters respectively; represents the generalized adaptive estimated kinematic Jacobian matrix, Indicates the The gradient vector of the finger potential energy function, Indicates the The generalized joint angles of the fingers, Expressed as The update rate of the gravity compensation term for each finger, Expressed as The actual gravity of each finger, Indicates the The joint velocity vectors of the fingers, Indicates the The update rate of the parameter estimates of each finger, represents the time-varying gain function, Indicates the The kinematic regression matrix of each finger, Indicates the The kinematic parameter estimation vector of each finger.

10. A collaborative grasping control system for dexterous hands oriented to modeling uncertainty, characterized by: A dexterous hand collaborative grasping control method for executing modeling uncertainty according to any one of claims 1 to 9; comprising: System modeling module, used to establish the dynamic model of the dexterous hand system based on the motion state and nonlinear characteristics of the dexterous hand; A virtual coupling modeling module is used to introduce virtual coupling between multiple fingers of the dexterous hand system to build a communication topology, define the collaborative grasping error and construct the corresponding potential energy function; Distributed control module, used to design a time-varying gain-based gradient descent distributed control law to address the uncertainty of the dynamic model; Gravity compensation module, used to process the change of motion gravity through gravity term integral compensator; Jacobian estimation module, used to compensate for kinematic parameter uncertainty based on the adaptive law of the time-varying gain Jacobian matrix; The control law integration module is used to integrate the output content of the distributed control module, gravity compensation module and Jacobian estimation module, and form the control law of the entire dexterous hand system, so that the dexterous hand system can achieve collaborative grasping control within the time preset by the user.

Images