Heterogeneous mechanical arm cluster synchronous cooperation method based on state coupling neural network and leader elimination competition

Abstract

The invention discloses a heterogeneous mechanical arm cluster synchronous cooperation method based on a state coupling neural network and leader elimination competition, and the method comprises the steps: constructing a cooperative motion planning problem of a heterogeneous mechanical arm cluster, and carrying out the modeling of the problem as a constrained quadratic programming problem; aiming at each mechanical arm sub-cluster in the mechanical arm cluster, constructing a leader elimination competition mechanism, and determining a leader mechanical arm based on a node connectivity evaluation function; based on a nonlinear complementary function corresponding to the quadratic programming problem and a state coupling relation between the mechanical arms, a state coupling neural network solver is constructed; a quadratic programming problem is solved based on a state coupling neural network solver, and the optimal joint angular velocity control quantity of each mechanical arm is output; and an expected joint angle is obtained according to the optimal joint angular velocity control quantity and converted into a control instruction, and the heterogeneous mechanical arm cluster is controlled to complete a synchronous cooperation task. Joint synchronization between the mechanical arms can be achieved, and robustness and continuity of a whole mechanical arm cluster cooperative task are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of robot cooperative control technology, specifically to a method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition. Background Technology

[0002] As task complexity increases, a single robotic arm is no longer sufficient for large-scale collaborative tasks, such as collaborative handling of large components and collaborative processing at multiple workstations. Therefore, research on the collaborative control of multi-robotic arm systems is of great significance. Multi-robotic arm systems are generally divided into homogeneous systems and heterogeneous systems. Heterogeneous systems consist of robotic arms with different structures or functions, also known as multi-robot clusters. Their modeling is more complex, the communication topology changes dynamically, and the design of control strategies is more challenging.

[0003] Existing research on multi-manipulator cooperative control largely focuses on the positional coordination of end effectors in the task space, neglecting the consistency of end effector orientation. For a six-DOF manipulator, simultaneously considering position and orientation in kinematics cannot satisfy functional redundancy, and directly controlling position and orientation in the task space increases the complexity of kinematic solutions. Furthermore, existing research on distributed multi-robot systems mostly assumes a fixed communication topology and a fixed leader. In real-world distributed systems, manipulators are connected via communication networks. If communication between some manipulators is interrupted, control strategies relying on a fixed leader may lead to system performance degradation or even task failure. Therefore, a cooperative control method for heterogeneous manipulator clusters that can maintain robustness in dynamic communication environments and effectively achieve end effector orientation synchronization is needed. Summary of the Invention

[0004] To overcome the defects and shortcomings of existing technologies, this invention provides a method for synchronous collaboration of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition. This invention achieves joint synchronization through the state coupling mechanism inside the state-coupled neural network, thereby indirectly achieving orientation synchronization of the end effector, avoiding the complexity of directly adding orientation constraints to the optimization problem. At the same time, the leader role is dynamically switched when the communication topology changes through the leader elimination competition mechanism, which improves the robustness of the distributed robotic arm cluster under communication failure.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] This invention provides a method for synchronous collaboration of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition, comprising the following steps:

[0007] The cooperative motion planning problem of a heterogeneous robotic arm cluster is constructed and modeled as a constrained quadratic programming problem;

[0008] For each sub-cluster of robotic arms in the robotic arm cluster, a leader elimination competition mechanism is constructed, and the leader robotic arm is determined based on the node connectivity evaluation function;

[0009] Based on the nonlinear complementary function corresponding to the quadratic programming problem and the state coupling relationship between the robotic arms, a state coupling neural network solver is constructed.

[0010] The quadratic programming problem is solved using a state-coupled neural network solver, which outputs the optimal joint angular velocity control value for each robotic arm.

[0011] The optimal joint angular velocity control value is used to obtain the desired joint angle, which is then converted into control commands to control the heterogeneous robotic arm cluster to complete synchronous collaborative tasks.

[0012] As a preferred technical solution, the constrained quadratic programming problem is expressed as:

[0013] ;

[0014] in, Let be the joint angular velocity vector of all robotic arms. , A diagonal matrix representing the leader's symbol. , The Laplace matrix is ​​used to describe the communication topology between robotic arms. Let Jacobian matrix be the system's Jacobian matrix. For positive kinematic functions, and For the reference trajectory and its derivative of the sub-cluster, A vector representing the relative positional relationship between robotic arms. For convergence parameters, and These are the constraint matrix and vector obtained from the physical amplitude limiting transformation of joint angles and angular velocities.

[0015] As a preferred technical solution, a leader elimination competition mechanism is constructed, and the leader robotic arm is determined based on a node connectivity evaluation function, specifically including:

[0016] For the robotic arms in the sub-cluster, calculate their connectivity evaluation function:

[0017] Determine the maximum value of the connectivity evaluation function for all robotic arms within the sub-cluster;

[0018] The leadership role is switched to the robotic arm with the highest evaluation value.

[0019] As a preferred technical solution, the connectivity evaluation function is expressed as follows:

[0020] ;

[0021] in, Indicates robotic arm Node degree within the sub-cluster Preserve the preference coefficient for the leader. For the indicator function, when the robotic arm The current leader The value is 1 if it is true, and 0 otherwise.

[0022] As the preferred technical solution, the leadership role is switched to the robotic arm with the highest evaluation value, specifically as follows:

[0023] ;

[0024] in, The maximum value of the connectivity evaluation function, This represents the current leader's evaluation score. Indicates the current leader, Indicating a new leader. No. A cluster of robotic arms Indicates the first robotic arms in a sub-cluster The connectivity evaluation function.

[0025] As a preferred technical solution, a state-coupled neural network solver is constructed based on the nonlinear complementary function corresponding to the quadratic programming problem and the state coupling relationship between the robotic arms, specifically including:

[0026] The quadratic programming problem is transformed into a system of nonlinear complementary equations based on the Fischer-Burmeister nonlinear complementary function.

[0027] The dynamic equations for constructing a state-coupled neural network are as follows:

[0028] ;

[0029] in, For network state variables, , It is an adjustable parameter. The energy function is constructed based on a set of nonlinear complementary equations. For activation function, For the feedback gain matrix, Let be the coupling matrix. , , This is the coefficient matrix derived from the problem model and the derivative of the energy function.

[0030] As a preferred technical solution, network state variables Represented as:

[0031] ;

[0032] in, Indicates joint angular velocity, Represents the equality-bound Lagrange multipliers. This represents the inequality constraint multiplier.

[0033] As a preferred technical solution, the optimal joint angular velocity control quantity output by the state-coupled neural network solver asymptotically converges to the optimal solution of the quadratic programming problem as the energy function approaches zero.

[0034] The present invention also provides a computer-readable storage medium storing a program that, when executed by a processor, implements the above-described method for synchronous collaboration of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition.

[0035] The present invention also provides a computer device, including a processor and a memory for storing processor-executable programs, wherein when the processor executes the program stored in the memory, it implements the above-described method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition.

[0036] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0037] (1) The present invention directly realizes joint synchronization between robotic arms through the coupling mechanism inside the state-coupled neural network. Joint synchronization ensures the consistency of the orientation of the end effector. This method avoids adding complex orientation constraints in the optimization problem and does not require direct orientation control in kinematics, thus reducing the complexity of problem modeling and solving.

[0038] (2) The leader elimination competition mechanism proposed in this invention can dynamically evaluate and elect the best leader based on the real-time communication topology. When the original leader experiences a communication failure, the leadership can be automatically and smoothly transferred to the robotic arm with better connectivity, thereby ensuring the robustness and continuity of the entire robotic arm cluster's collaborative tasks in the event of changes or failures in the communication link.

[0039] (3) The present invention uses a state-coupled neural network based on nonlinear complementary functions and coupled design to solve the collaborative optimization problem. It has high solution accuracy, fast convergence speed, and can generate the optimal control quantity of the robotic arm that satisfies multiple constraints in real time.

[0040] (4) This invention is applicable not only to heterogeneous robotic arm clusters, but also to homogeneous clusters, providing an effective control implementation scheme for the synchronous collaborative control of distributed robot clusters, and its application is wide. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating the synchronous collaboration method for heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition, as described in this invention.

[0042] Figure 2(a) is a schematic diagram of the average joint angular velocity before the switching of the first group of robotic arms in this invention;

[0043] Figure 2(b) is a schematic diagram of the average joint angular velocity after the first group of robotic arms is switched according to the present invention;

[0044] Figure 2(c) is a schematic diagram of the average joint angular velocity before the switching of the second group of robotic arms in this invention;

[0045] Figure 2(d) is a schematic diagram of the average joint angular velocity after the second group of robotic arms is switched according to the present invention;

[0046] Figure 3 This is a schematic diagram of the end effector trajectory tracking of the robotic arm cluster in a collaborative task according to the present invention;

[0047] Figure 4 An experimental diagram showing the collaborative drawing of a star-shaped trajectory by five robotic arms after applying the method of this invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] Example 1

[0050] like Figure 1 As shown, this embodiment provides a method for synchronous collaboration of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition, including the following steps:

[0051] S1: Construct the cooperative motion planning problem of heterogeneous robotic arm clusters, and model the cooperative motion planning problem of heterogeneous robotic arm clusters into a quadratic programming problem with equality and inequality constraints.

[0052] In this embodiment, consider a... A redundant robotic arm cluster system consists of several sub-clusters, each containing robotic arms with identical structures. The synchronization and cooperation problem of the cluster is modeled as the following constrained quadratic programming problem:

[0053] ;

[0054] in, Let be the joint angular velocity vector of all robotic arms. , A diagonal matrix representing the leader's symbol. , The Laplace matrix is ​​used to describe the communication topology between robotic arms. Let Jacobian matrix be the system's Jacobian matrix. For positive kinematic functions, and For the reference trajectory and its derivative of the sub-cluster, A vector representing the relative positional relationship between robotic arms. For convergence parameters, and These are the constraint matrices and vectors obtained from the physical amplitude limiting transformation of joint angles and angular velocities;

[0055] S2: For each robotic arm sub-cluster in the robotic arm cluster, build and run a leader elimination competition mechanism to dynamically elect or maintain the leader robotic arm of the sub-cluster based on the node connectivity evaluation function.

[0056] In this embodiment, the implementation process of the leader elimination competition mechanism specifically includes:

[0057] S21: For the first Sub-clusters robotic arm Calculate its connectivity evaluation function :

[0058] ;

[0059] in, Indicates robotic arm Node degree within the sub-cluster Preserve the preference coefficient for the leader. For indicator functions, when The current leader The value is 1 if the condition is met, and 0 otherwise.

[0060] S22: Determine the maximum value of the evaluation function for all robotic arms within the sub-cluster. ;

[0061] ;

[0062] S23: Determine the new leader according to the following decision-making rules. ;

[0063] ;

[0064] This mechanism will only switch the leadership role to the robotic arm with the highest evaluation value when the current leader's evaluation value is no longer the highest in the cluster, thus avoiding unnecessary leader oscillations caused by brief communication disturbances or identical scores.

[0065] S3: Based on the nonlinear complementary function corresponding to the quadratic programming problem and the state coupling relationship between the robotic arms, a state coupling neural network solver is constructed.

[0066] In this embodiment, constructing a state-coupled neural network solver specifically includes:

[0067] S31: Introduce Lagrange multipliers and use the Fischer-Burmeister nonlinear complementary function to transform the quadratic programming problem in step S1 into a system of nonlinear equations;

[0068] S32: Define the energy function associated with this system of nonlinear equations. ;

[0069] S33: The dynamic equations for the state-coupled neural network are as follows:

[0070] ;

[0071] in, These are network state variables, including joint angular velocities. Equality-bound Lagrange multipliers Inequality-bound multipliers , , It is an adjustable parameter. The energy function is constructed based on a system of nonlinear complementary equations. For activation function, For the feedback gain matrix, For a coupling matrix, its non-zero block structure makes the terms... Corresponding to mathematics It can couple the synchronization errors of joint speeds between robotic arms, thereby driving the joint movements of robotic arms within the cluster to tend towards synchronization. , , The coefficient matrix is ​​derived from the problem model and the derivative of the energy function;

[0072] S4: Using the state-coupled neural network solver built in step S3, solve the quadratic programming problem online. Its output is the optimal control quantity for each robotic arm, and its output vector contains the optimal joint angular velocity control quantity for each robotic arm.

[0073] In this embodiment, the optimal joint angular velocity control quantity output by the state-coupled neural network solver Through energy function It tends to zero and asymptotically converges to the theoretical optimal solution of the quadratic programming problem;

[0074] S5: Integrate the optimal joint angular velocity control quantity obtained in step S4 to obtain the desired joint angle, and convert it into a control command to drive and control the heterogeneous robotic arm cluster to complete the synchronous collaborative task in real time.

[0075] In this embodiment, converting into control commands means that the host computer sends the desired joint angle sequence obtained by integration to the underlying servo driver of each robotic arm through the communication interface, so as to realize real-time closed-loop position control of the robotic arm joints.

[0076] Example 2

[0077] This embodiment verifies the synchronous collaboration method of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition in Embodiment 1 through simulation in a simulation environment. In the simulation environment, a collaborative task model is established for a system containing two heterogeneous sub-clusters (e.g., the first group of 5 robotic arms R1-R5, and the second group of 5 robotic arms R6-R10). The task requires the two sub-clusters to track different preset trajectories (e.g., butterfly and star patterns) while maintaining consistent orientation of the end effectors within the clusters. This problem is modeled as a quadratic programming problem, where the Laplace matrix... Determined based on the initial communication topology (e.g., fully connected);

[0078] In this embodiment, leaders of two sub-clusters are initialized (e.g., robotic arm R1 for sub-cluster 1 and robotic arm R6 for sub-cluster 2). During system operation, a communication failure is artificially set at t=2.5 seconds to simulate the interruption of communication between leaders R1 and R6 and some followers. At this time, the leader elimination competition mechanism is triggered. Based on the real-time connectivity score, the leader of sub-cluster 1 is automatically switched to R2 and the leader of sub-cluster 2 is switched to R7.

[0079] In this embodiment, the parameters of the state-coupled neural network solver are set as follows: , , Activation function Use a linear function;

[0080] In this embodiment, the established quadratic programming problem, whose matrix is ​​dynamically updated with the change of leader, is input into a state-coupled neural network solver. This solver performs online parallel computation and outputs the optimal joint angular velocity vectors of all robotic arms in real time. ;

[0081] In this embodiment, for Numerical integration is performed to obtain the desired joint angle, which is then converted into control commands to drive the virtual robotic arm's movement. As shown in Figures 2(a)-2(d), the simulation results of the joint angular velocities before and after the robotic arm switching are obtained online from the state-coupled neural network solver. The results show that the joint angular velocities output by the solver are smooth and oscillating. Figure 3 As shown, the end effector trajectory tracking results of the robotic arm cluster in the collaborative task are obtained, indicating that even if a leader switch occurs at t=2.5 seconds, the two robotic arm clusters can still accurately and smoothly complete their respective preset trajectory tracking tasks.

[0082] Table 1 below shows the specific data on the synchronization errors of the joint angles and angular velocities of each robotic arm in the cluster before and after the Leader Elimination Competition (LEC) mechanism is triggered:

[0083] Table 1. Comparison of maximum values ​​of joint angle and speed synchronization errors of robotic arm groups before and after switching.

[0084]

[0085] After the Leader Elimination Competition (LEC) mechanism was triggered, the maximum joint synchronization error of each robotic arm changed significantly. For example, for sub-cluster one, the angular synchronization error of the original leader R1... from After reaching LEC Its angular velocity synchronization error from Down to Sub-cluster two, the original leader R6's angular synchronization error from Significantly improved to Its angular velocity synchronization error from Significantly reduced Meanwhile, the synchronization errors of some follower robotic arms (such as R3, R4, R5, R8, R9, and R10) also decreased significantly after LEC. These data indicate that the proposed mechanism not only maintains system stability after a leader change but also optimizes the overall synchronization accuracy of the cluster. Although the error values ​​of individual robotic arms (such as R2 and R7) remained stable before and after the change, this reflects a reasonable adjustment of the system state based on connectivity assessment. This invention can effectively absorb internal disturbances caused by changes in communication topology and quickly guide the system back to a high-performance synchronization state.

[0086] Furthermore, in this embodiment, in a laboratory environment, five Ufactory xArm6 six-DOF robotic arms were used to form a cluster to perform the task of collaboratively drawing a star-shaped trajectory. Initially, R1 was designated as the leader in the communication topology. During the experiment, a communication failure was artificially simulated, changing the communication topology. At this time, the leader elimination competition mechanism was triggered, automatically calculating and electing R2 as the new leader. The state-coupled neural network calculated the control variables in real time according to the new topology and distributed them to each robotic arm. Figure 4 As shown, the experimental results of five robotic arms collaboratively drawing a star-shaped trajectory after applying the method of the present invention were obtained. Even with the leader dynamically switching, the five robotic arms still successfully and synchronously completed the task of drawing a high-precision star-shaped trajectory, verifying the effectiveness and practicality of the present invention in a real physical system.

[0087] Example 3

[0088] This embodiment provides a storage medium, which may be a ROM, RAM, disk, optical disk, or other storage medium. The storage medium stores one or more programs. When the program is executed by the processor, it implements the heterogeneous robotic arm cluster synchronous cooperation method based on state-coupled neural network and leader elimination competition as described in Embodiment 1.

[0089] Example 4

[0090] This embodiment provides a computing device, which may be a desktop computer, laptop computer, smartphone, PDA handheld terminal, tablet computer or other terminal device with display function. The computing device includes a processor and a memory. The memory stores one or more programs. When the processor executes the program stored in the memory, it implements the heterogeneous robotic arm cluster synchronous cooperation method based on state-coupled neural network and leader elimination competition of Embodiment 1.

[0091] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for synchronous collaboration of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition, characterized in that, Includes the following steps: The cooperative motion planning problem of a heterogeneous robotic arm cluster is constructed and modeled as a constrained quadratic programming problem; For each sub-cluster of robotic arms in the robotic arm cluster, a leader elimination competition mechanism is constructed, and the leader robotic arm is determined based on the node connectivity evaluation function; Based on the nonlinear complementary function corresponding to the quadratic programming problem and the state coupling relationship between the robotic arms, a state coupling neural network solver is constructed. The quadratic programming problem is solved using a state-coupled neural network solver, which outputs the optimal joint angular velocity control value for each robotic arm. The optimal joint angular velocity control value is used to obtain the desired joint angle, which is then converted into control commands to control the heterogeneous robotic arm cluster to complete synchronous collaborative tasks.

2. The method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition as described in claim 1, characterized in that, The constrained quadratic programming problem is represented as: ； in, Let be the joint angular velocity vector of all robotic arms. , A diagonal matrix representing the leader's symbol. , The Laplace matrix is ​​used to describe the communication topology between robotic arms. Let Jacobian matrix be the system's Jacobian matrix. For positive kinematic functions, and For the reference trajectory and its derivative of the sub-cluster, A vector representing the relative positional relationship between robotic arms. For convergence parameters, and These are the constraint matrix and vector obtained from the physical amplitude limiting transformation of joint angles and angular velocities.

3. The method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition as described in claim 1, characterized in that, Construct a leader elimination competition mechanism and determine the leader robotic arm based on the node connectivity evaluation function, specifically including: For the robotic arms in the sub-cluster, calculate their connectivity evaluation function: Determine the maximum value of the connectivity evaluation function for all robotic arms within the sub-cluster; The leadership role is switched to the robotic arm with the highest evaluation value.

4. The method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition as described in claim 3, characterized in that, The connectivity evaluation function is expressed as: ； in, Indicates robotic arm Node degree within the sub-cluster Preserve the preference coefficient for the leader. For the indicator function, when the robotic arm The current leader The value is 1 if it is true, and 0 otherwise.

5. The heterogeneous robotic arm cluster synchronous cooperation method based on state-coupled neural network and leader elimination competition according to claim 3, characterized in that, The leadership role switches to the robotic arm with the highest evaluation value, specifically as follows: ； in, This represents the maximum value of the connectivity evaluation function. This represents the current leader's evaluation score. Indicates the current leader, Indicating a new leader. Indicates the first A cluster of robotic arms Indicates the first robotic arms in a sub-cluster The connectivity evaluation function.

6. The method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition as described in claim 1, characterized in that, Based on the nonlinear complementary function corresponding to the quadratic programming problem and the state coupling relationship between robotic arms, a state-coupled neural network solver is constructed, specifically including: The quadratic programming problem is transformed into a system of nonlinear complementary equations based on the Fischer-Burmeister nonlinear complementary function. The dynamic equations for constructing a state-coupled neural network are as follows: ； in, For network state variables, , It is an adjustable parameter. The energy function is constructed based on a set of nonlinear complementary equations. For activation function, For the feedback gain matrix, Let be the coupling matrix. , , This is the coefficient matrix derived from the problem model and the derivative of the energy function.

7. The method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition as described in claim 6, characterized in that, Network state variables Represented as: ； in, Indicates joint angular velocity, Represents the equality-bound Lagrange multipliers. This represents the inequality constraint multiplier.

8. The method for synchronous cooperation of heterogeneous robotic arm clusters based on state-coupled neural networks and leader elimination competition as described in claim 1, characterized in that, The optimal joint angular velocity control output by the state-coupled neural network solver converges asymptotically to the optimal solution of the quadratic programming problem as the energy function approaches zero.

9. A computer-readable storage medium storing a program, characterized in that, When the program is executed by the processor, it implements the heterogeneous robotic arm cluster synchronous cooperation method based on state-coupled neural network and leader elimination competition as described in any one of claims 1-8.

10. A computer device comprising a processor and a memory for storing a processor-executable program, characterized in that, When the processor executes the program stored in the memory, it implements the heterogeneous robotic arm cluster synchronous cooperation method based on state-coupled neural network and leader elimination competition as described in any one of claims 1-8.

Images