A mechanical arm control method and system based on neutral logic parallel neural network

Abstract

The application discloses a kind of mechanical arm control method and system based on neutral logic parallel neural network, the method of the present application includes obtaining input task target and posture constraint condition, the trajectory tracking problem of mechanical arm to task target is converted into the nonlinear equation set constrained by posture constraint condition;Zero neural network NLS based on neutral fuzzy logic system ZNN is used to solve and obtain the joint angle of mechanical arm: the objective function of nonlinear equation set constrained by posture constraint condition is generated output value using neutral fuzzy logic system NLS, and the nonlinear equation set of trajectory tracking problem is solved using zero neural network ZNN to obtain the joint angle of mechanical arm;The control command generated is issued to drive mechanical arm to complete task target.The present application aims to solve the technical bottleneck of traditional ZNN model in dynamic environment adaptability, real-time computing efficiency and anti-interference ability, realize high-precision trajectory tracking and stable control of mechanical arm in complex scene.

Description

Technical Field

[0001] This invention relates to the field of robotic arm control technology, specifically to a robotic arm control method and system based on a neutral logic parallel neural network. Background Technology

[0002] With the rapid development of industrial automation and intelligent robot technology, the application demand for robotic arms in precision assembly, trajectory tracking, and complex environment operations is increasing. Existing robotic arm control methods mostly employ zeroing neural network (ZNN) models, achieving motion control of the robotic arm through solving nonlinear equations. However, as the complexity of application scenarios increases, the limitations of traditional ZNN models in terms of environmental adaptability, computational efficiency, and control accuracy are gradually becoming apparent: In existing technologies, traditional ZNN models rely on fixed parameters, requiring frequent manual parameter tuning under different task scenarios. Although current research has improved the adaptability of ZNN models by introducing fuzzy logic systems (FLS) to construct fuzzy ZNN (FZNN) models, it still faces two major bottlenecks: First, insufficient environmental adaptability; fuzzy rules have limited ability to handle unstructured scenarios, making it difficult for FZNN models to handle more complex unstructured scenarios. Second, severe computational burden; the computational load of existing ZNN models is increasing, and the addition of fuzzy logic systems further increases the computational load of ZNN models, making it difficult to meet real-time requirements. Furthermore, existing ZNN models struggle to balance control accuracy and stability in scenarios involving noise interference or high-precision robotic arm control. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a robotic arm control method and system based on neutral logic parallel neural networks, which addresses the above-mentioned problems in the prior art. This invention aims to solve the technical bottlenecks of traditional ZNN models in terms of dynamic environment adaptability, real-time computing efficiency and anti-interference ability, and realize high-precision trajectory tracking and stable control of robotic arms in complex scenarios.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0005] A robotic arm control method based on a neutral logic parallel neural network includes the following steps:

[0006] S101, Input the task objective and attitude constraints, and transform the trajectory tracking problem of the robotic arm for the task objective into a set of nonlinear equations constrained by attitude constraints;

[0007] S102, Solve the nonlinear equations constrained by attitude conditions using a nullable neural network (NLS-ZNN) based on a neutral fuzzy logic system to obtain the joint angles of the robotic arm, including: ① Generating the output value of the objective function of the nonlinear equations constrained by attitude conditions using a neutral fuzzy logic system (NLS). ② Output value The nonlinear equations of the trajectory tracking problem are solved using a zero-form neural network (ZNN) to obtain the joint angles of the robotic arm;

[0008] S103 sends control commands generated based on the joint angles of the robotic arm to drive the robotic arm to complete the task objective.

[0009] Optionally, step S101 includes:

[0010] S201, Input the task objective and attitude constraints, wherein the task objective includes the desired trajectory of the robotic arm's end effector. and the expected pose quaternion The expected trajectory The coordinates are , ,in For attitude quaternions The scalar part represents the magnitude of rotation; For attitude quaternions The vector part represents the direction and magnitude of rotation; the functional expression of the attitude constraint is: ,in The joint angle of the robotic arm. Indicates the lower bound of the joint angle. Indicates the upper limit of the joint angle;

[0011] S202, with The joint angles of the robotic arm represent the desired trajectory. and the expected pose quaternion As the objective function of the nonlinear equation system Transform the attitude constraints into matrix inequalities. ,in The coefficient matrix, This represents the upper and lower bounds of the joint angles of the robotic arm. , , for An identity matrix of dimension 1 Representing the degrees of freedom of the robotic arm, we obtain the matrix inequality. A constrained system of nonlinear equations;

[0012] S203, subject to matrix inequalities The constrained nonlinear equations are transformed into solution equations using the Lagrange multiplier method and the Karush-Kuhn-Tucker conditions. The problem, among which Let be the objective function of the transformed nonlinear equation system. Let be the state vector to be solved. , They are slack variables; and we have:

[0013] ;

[0014] ;

[0015] in, ~ For slack variables The first to m elements in the array, where m is the slack variable. The number of elements in the data.

[0016] Optionally, in step S102, the objective function of the nonlinear equation system constrained by attitude constraints is used to generate an output value using a neutral fuzzy logic system (NLS). include:

[0017] S301, the objective function of the nonlinear equation system constrained by attitude constraints is used as the error function. ;

[0018] S302, the error function norm As input to a neutral fuzzy logic system ;

[0019] S303, input By providing a pre-defined true membership function Uncertain membership function and spurious membership functions Neutralize the process;

[0020] S304, the neutralization result is based on the "IF-THEN" rule set. Perform neutral fuzzy reasoning;

[0021] S305, Deneutralize the neutral fuzzy inference result to obtain the output value. .

[0022] Optionally, in step S304, the "IF-THEN" rule set The function expression is:

[0023] If the input value is larger than NE, then the output value is larger than NE.

[0024] If the input value is Moderate (ME), then the output value is Moderate (ME).

[0025] If the input value is smaller than SE, then the output value is smaller than SE.

[0026] in, These represent the "IF-THEN" rule set. The three rules, smaller SE, moderate ME, and larger NE, define three intervals for the joint angles of the robotic arm. "Describes the mapping relationship between inputs and outputs;"

[0027] The output value obtained by deneutralizing the neutral fuzzy inference in step S305 The function expression is:

[0028] ;

[0029] in, and These are preset constant parameters. This is a preset constant used to prevent the denominator from being zero. For the preset real membership function, For a pre-defined uncertain membership function, This is a pre-defined spurious membership function.

[0030] Optionally, in step S102, the output value will be... When using a nullable neural network (ZNN) to solve the nonlinear equations of a trajectory tracking problem to obtain the joint angles of a robotic arm, the functional expression of the ZNN is as follows:

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] in, The dynamic coefficient matrix of the system. for The first-order differential, The dynamic and constraint vectors of the system. The constant convergence factor For activation function, Let be the objective function of the transformed nonlinear equation system. The objective function of a system of nonlinear equations about Jacobian matrix, , for An identity matrix of dimension 1 Indicates the degrees of freedom of the robotic arm. ~ For slack variables The first to the second One element, For slack variables The number of elements in; for The first-order differential, for The first-order differential, for The first-order differential, , Indicates the joint angle of the robotic arm. Indicates the lower bound of the joint angle. Indicates the upper limit of the joint angle.

[0036] Optionally, the activation function is a linear activation function, a sign double power activation function, a modified finite-time activation function, or a fast predefined time-converged activation function, and the function expression of the linear activation function is:

[0037] ;

[0038] in, It is a linear activation function. This is the input to the activation function;

[0039] The function expression for the sign double power activation function is:

[0040] ;

[0041] ;

[0042] in, For the sign double power activation function, For symbolic functions, As a regulating factor, and ;

[0043] The improved finite-time activation function has the following expression:

[0044] ;

[0045] in, For an improved finite-time activation function, The nonlinear convergence gain coefficient is... Let be the linear convergent gain coefficient, and , ;

[0046] The function expression for a fast, predefined, time-converging activation function is:

[0047] ;

[0048] in, As a regulating factor, and Here is the gain coefficient of the exponential term, and , , .

[0049] Optionally, in step S102, the output value will be... When using a zero-dimensional neural network (ZNN) to solve the nonlinear equations for trajectory tracking to obtain the joint angles of a robotic arm, this includes applying the task time domain... Divided into partially overlapping task sub-intervals ,and ,in For the task time domain At the end of the day, , partially overlapping task sub-intervals include:

[0050] First section: ;

[0051] Middle section: ;

[0052] Ending interval: ;

[0053] in, The convergence parameters are set at a predetermined time; for each task sub-interval, the output value will be... The joint angles of the robotic arm are obtained by solving the nonlinear equations of the trajectory tracking problem using an independent nullable neural network (ZNN).

[0054] The present invention also provides a robotic arm control system based on a neutral logic parallel neural network, including a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the robotic arm control method based on the neutral logic parallel neural network.

[0055] The present invention also provides a computer-readable storage medium storing a computer program or instructions that are programmed or configured to execute the robotic arm control method based on a neutral logic parallel neural network via a processor.

[0056] The present invention also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the robotic arm control method based on a neutral logic parallel neural network via a processor.

[0057] Compared with the prior art, the present invention can mainly achieve the following beneficial effects: (1) The present invention includes solving the nonlinear equation system constrained by attitude constraints using the nullable neural network NLS-ZNN based on neutral fuzzy logic system to obtain the joint angle of the robotic arm. By introducing the neutral fuzzy logic system, a dynamic parameter adjustment mechanism is realized, which solves the problem of insufficient adaptability of the existing ZNN model; (2) The present invention includes solving the nonlinear equation system constrained by attitude constraints using the nullable neural network NLS-ZNN based on neutral fuzzy logic system to obtain the joint angle of the robotic arm. The neutral fuzzy logic system NLS-ZNN based on neutral fuzzy logic system NLS-ZNN and the nullable neural network NLS-ZNN based on neutral fuzzy logic system NLS-ZNN are used to solve the nonlinear equation system constrained by attitude constraints to obtain the joint angle of the robotic arm. The ZNN can adopt a serial or parallel architecture as needed to realize the serial and parallel computation of the ZNN model. In particular, the parallel computation can improve the computation efficiency of the zero-negation neural network ZNN and break through the bottleneck of serial computation efficiency. (3) The zero-negation neural network NLS-ZNN based on the neutral fuzzy logic system can be applied to the position and attitude control of various high-precision robotic arms, providing new solutions for industrial robots, aerospace and other fields. It can meet the requirements of real-time control and flexible scheduling, solve the technical bottlenecks of traditional ZNN models in terms of dynamic environment adaptability, real-time computation efficiency and anti-interference ability, and realize high-precision trajectory tracking and stable control of robotic arms in complex scenarios. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the basic process of the method in an embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the workflow of the neutral fuzzy logic system in an embodiment of the present invention.

[0060] Figure 3 This is a schematic diagram illustrating the working principle of the neutral fuzzy logic system in an embodiment of the present invention.

[0061] Figure 4 This is a comparison of errors under ideal conditions in the embodiments of the present invention.

[0062] Figure 5This is a comparison of errors under noise 3sin(t) interference in the embodiments of the present invention.

[0063] Figure 6 This refers to the joint angle variation of the robotic arm in this embodiment of the invention.

[0064] Figure 7 This is a 3D comparison diagram of the expected trajectory and the actual trajectory in an embodiment of the present invention.

[0065] Figure 8 This is a plane comparison diagram of the expected trajectory and the actual trajectory in an embodiment of the present invention.

[0066] Figure 9 This represents the error variation of the NLS-PZNN model in this embodiment of the invention. Detailed Implementation

[0067] To enable those skilled in the art to better understand the technical solution of the present invention, the following will take the 7-DOF KUKAiiwa14 robotic arm as the controlled object to achieve high-precision trajectory tracking and attitude control for planar carving tasks as an example, and in conjunction with the accompanying drawings in the embodiments of the present invention, to further describe the technical solution of the present invention in detail.

[0068] like Figure 1 As shown, the robotic arm control method based on a neutral logic parallel neural network in this embodiment includes the following steps:

[0069] S101, Input the task objective and attitude constraints, and transform the trajectory tracking problem of the robotic arm for the task objective into a set of nonlinear equations constrained by attitude constraints;

[0070] S102, Solve the nonlinear equations constrained by attitude conditions using a nullable neural network (NLS-ZNN) based on a neutral fuzzy logic system to obtain the joint angles of the robotic arm, including: ① Generating the output value of the objective function of the nonlinear equations constrained by attitude conditions using a neutral fuzzy logic system (NLS). ② Output value The nonlinear equations of the trajectory tracking problem are solved using a zero-form neural network (ZNN) to obtain the joint angles of the robotic arm;

[0071] S103 sends control commands generated based on the joint angles of the robotic arm to drive the robotic arm to complete the task objective.

[0072] In this embodiment, step S101 includes:

[0073] S201, Input the task objective and attitude constraints, wherein the task objective includes the desired trajectory of the robotic arm's end effector. and the expected pose quaternion The expected trajectory The coordinates are , ,in For attitude quaternions The scalar part represents the magnitude of rotation; For attitude quaternions The vector part represents the direction and magnitude of rotation; the functional expression of the attitude constraint is: ,in The joint angle of the robotic arm. Indicates the lower bound of the joint angle. Indicates the upper limit of the joint angle;

[0074] S202, with The joint angles of the robotic arm represent the desired trajectory. and the expected pose quaternion As the objective function of the nonlinear equation system :

[0075] ;

[0076] Transform the attitude constraints into matrix inequalities. ,in The coefficient matrix, This represents the upper and lower bounds of the joint angles of the robotic arm. , , for An identity matrix of dimension 1 Representing the degrees of freedom of the robotic arm, we obtain the matrix inequality. A constrained system of nonlinear equations;

[0077] S203, Introducing slack variables The matrix inequality will be applied. The constrained nonlinear equations are transformed into solution equations using the Lagrange multiplier method and the Karush-Kuhn-Tucker conditions. The problem, among which Let be the objective function of the transformed nonlinear equation system. Let be the state vector to be solved. , They are slack variables; and we have:

[0078] ;

[0079] ;

[0080] in, ~ For slack variables The first to m elements in the array, where m is the slack variable. The number of elements in the data.

[0081] like Figure 2 and Figure 3 As shown, in step S102 of this embodiment, the objective function of the nonlinear equation system constrained by attitude constraints is used to generate an output value using a neutral fuzzy logic system (NLS). include:

[0082] S301, the objective function of the nonlinear equation system constrained by attitude constraints is used as the error function. ,Right now:

[0083] ;

[0084] S302, the error function norm As input to a neutral fuzzy logic system ;

[0085] S303, input By providing a pre-defined true membership function Uncertain membership function and spurious membership functions Neutralize the process;

[0086] S304, the neutralization result is based on the "IF-THEN" rule set. Perform neutral fuzzy reasoning;

[0087] S305, Deneutralize the neutral fuzzy inference result to obtain the output value. .

[0088] In step S303 of this embodiment, the preset true membership function Uncertain membership function and spurious membership functions The function expression is:

[0089] ;

[0090] ;

[0091] ;

[0092] Real membership function exist The alpha value increases as the input value increases, thus it is an increasing function. (Undetermined membership function) exist The above approximates a Cauchy distribution. (Spurious membership function) exist The value increases as the input value increases, thus it is a decreasing function. That is, input .

[0093] In step S304 of this embodiment, the "IF-THEN" rule set The function expression is:

[0094] If the input value is larger than NE, then the output value is larger than NE.

[0095] If the input value is Moderate (ME), then the output value is Moderate (ME).

[0096] If the input value is smaller than SE, then the output value is smaller than SE.

[0097] in, These represent the "IF-THEN" rule set. The three rules, smaller SE, moderate ME, and larger NE, define three intervals for the joint angles of the robotic arm. "Describes the mapping relationship between input and output."

[0098] The function expression for deneutralizing neutral fuzzy inference in step S305 of this embodiment is:

[0099] ;

[0100] in, and These are preset constant parameters. This is a preset constant used to prevent the denominator from being zero. For the preset real membership function, For a pre-defined uncertain membership function, This is a pre-defined spurious membership function. In this function expression, It is mainly used as a gain or scaling factor to adjust the entire output range, amplifying or reducing the final control value; This is a sensitivity adjustment parameter, specifically used to adjust the system's sensitivity to uncertainty. The function expression incorporates the core idea of ​​neutral fuzzy logic systems—the measurement and management of uncertainty—into the final output of the neutral fuzzy logic system, making it suitable for handling complex, uncertain, and incomplete information problems in the real world.

[0101] According to the error function An error reduction formula can be constructed:

[0102] ,

[0103] in, The first derivative of the error function, A constant set, For the obtained output value, This is represented as an activation function; therefore, the model of the nullable neural network ZNN in the nullable neural network NLS-ZNN based on the neutral fuzzy logic system can be derived. In step S102 of this embodiment, the output value is... When using a nullable neural network (ZNN) to solve the nonlinear equations of a trajectory tracking problem to obtain the joint angles of a robotic arm, the functional expression of the ZNN is as follows:

[0104] ;

[0105] ;

[0106] ;

[0107] ;

[0108] in, The dynamic coefficient matrix of the system. for The first-order differential, For the system's dynamic constraint vector, The constant convergence factor For activation function, Let be the objective function of the transformed nonlinear equation system. The objective function of a system of nonlinear equations about Jacobian matrix, , for An identity matrix of dimension 1 Indicates the degrees of freedom of the robotic arm. ~ For slack variables The first to the second One element, For slack variables The number of elements in; for The first-order differential, for The first-order differential, for The first-order differential, , Indicates the joint angle of the robotic arm. Indicates the lower bound of the joint angle. Indicates the upper limit of the joint angle.

[0109] The activation function can be a linear activation function, a sign double power activation function, a modified finite-time activation function, or a fast predefined time-converged activation function. The expression for the linear activation function is as follows:

[0110] ;

[0111] in, It is a linear activation function. This is the input to the activation function;

[0112] The function expression for the sign double power activation function is:

[0113] ;

[0114] ;

[0115] in, For the sign double power activation function, For symbolic functions, As a regulating factor, and ;

[0116] The improved finite-time activation function has the following expression:

[0117] ;

[0118] in, For an improved finite-time activation function, The nonlinear convergence gain coefficient is... Let be the linear convergent gain coefficient, and , ;

[0119] The function expression for a fast, predefined, time-converging activation function is:

[0120] ;

[0121] in, As a regulating factor, and Here is the gain coefficient of the exponential term, and , , .

[0122] In step S102 of this embodiment, the output value will be... When using a nullable neural network (ZNN) to solve the nonlinear equations of a trajectory tracking problem to obtain the joint angles of a robotic arm, a serial ZNN model based on nullable logic systems (NLS) and a parallel ZNN model based on NLS can be used as needed. The serial ZNN model based on NLS involves using a neutral fuzzy logic system (NLS) to generate the output value of the objective function of the nonlinear equations constrained by attitude conditions from step ①. And step ② will output the value The nullable neural network (ZNN) is used to solve the nonlinear equations of the trajectory tracking problem to obtain the joint angles of the robotic arm for serial execution. The parallel ZNN model based on NLS is to use the neutral fuzzy logic system (NLS) to generate the output value of the objective function of the nonlinear equations constrained by attitude conditions in step ①. And step ② will output the value The zero-form neural network (ZNN) is used to solve the nonlinear equations of the trajectory tracking problem to obtain the joint angles of the robotic arm for parallel execution. For example, in step S102 of this embodiment, the output value is... When using a zero-dimensional neural network (ZNN) to solve the nonlinear equations for trajectory tracking to obtain the joint angles of a robotic arm, this includes applying the task time domain... Divided into partially overlapping task sub-intervals ,and ,in For the task time domain At the end of the day, , partially overlapping task sub-intervals include:

[0123] First section: ;

[0124] Middle section: ;

[0125] Ending interval: ;

[0126] in The convergence parameters are set at a predetermined time; for each task sub-interval, the output value will be... An independent nullable neural network (ZNN) is used to solve the nonlinear equations of the trajectory tracking problem to obtain the joint angles of the robotic arm. When using an NLS-based parallel ZNN model, the required number of parallel threads can be selected as needed. The output values... When using independent nullable neural networks (ZNNs) to solve the nonlinear equations of the trajectory tracking problem to obtain the joint angles of the robotic arm, the functional expressions of each independent ZNN can be expressed as:

[0127] ;

[0128] in, For the corresponding task sub-interval .against partially overlapping task sub-intervals For overlapping areas, select the joint angle corresponding to the smallest error.

[0129] To verify the robotic arm control method based on neutral logic parallel neural network in this embodiment, four types of activation functions were combined in this embodiment: linear activation function, sign double power activation function, improved finite-time activation function, and fast predefined time convergence activation function. In addition, the NLS-based serial ZNN model (NLS-ZNN) and three NLS-based parallel ZNN models (NLS-PZNN1 to NLS-PZNN3) were tested, and the results are shown in Table 1.

[0130] Table 1: Comparison of actual computation time required by different ZNN models to solve the same problem

[0131]

[0132] In Table 1, the actual computation time is in seconds. Since the computation time required for each run varies, all times in Table 1 are averages of 10 runs. NLS-ZNN is a serial ZNN model based on NLS, while NLS-PZNN1 to NLS-PZNN3 are three parallel ZNN models based on NLS. NLS-PZNN1 represents computation using 2 threads, NLS-PZNN2 represents computation using 4 threads, and NLS-PZNN3 represents computation using 8 threads. Table 1 shows that, generally, the more threads used in the NLS-based parallel ZNN models, the shorter the actual computation time.

[0133] The nullable neural network NLS-ZNN (NLS-ZNN model) based on neutral fuzzy logic system in this embodiment and the existing fuzzy ZNN model (FZNN) were experimentally compared on a KUKA iiwa14 robotic arm with 7 degrees of freedom. The results are as follows. Figures 4-5 As shown. Figure 4 and Figure 5 This demonstrates a comparison of the errors between the NLS-ZNN model and the fuzzy ZNN model using the method of this embodiment. Figure 4 and Figure 5 It is evident that the NLS-ZNN model of this embodiment outperforms the fuzzy ZNN model in both convergence and robustness. Figure 6 This embodiment demonstrates the joint angle variations generated by the NLS-ZNN model using the method described in this example. ~ For different joint angles generated by the NLS-ZNN model of the method in this embodiment, the joint angles solved using the NLS-ZNN model of the method in this embodiment are... ~ All are within joint constraints. Figure 7 and Figure 8 This embodiment demonstrates a comparison between the end effector trajectory of the robotic arm obtained from the NLS-ZNN model and the desired trajectory. Figure 7 and Figure 8 As can be seen, when the robotic arm uses the joint angle generated by the NLS-ZNN model of this embodiment, its end effector trajectory highly coincides with the expected trajectory. Figure 9 This demonstrates the error variation of the NLS-ZNN model (specifically, the parallelized NLS-ZNN model, i.e., the NLS-PZNN model) using the method of this embodiment, and its error stabilizes at... .

[0134] In summary, this embodiment presents a novel neutral fuzzy logic system for robotic arm control based on a neutral logic parallel neural network. Three different functions are used to represent true membership, false membership, and uncertain membership. Neutral fuzzy logic is then combined with the "IF-THEN" rule for neutral fuzzy inference, followed by deneutralization to obtain model parameters suitable for the current state. Secondly, a novel parallel nullification neural network is designed to efficiently solve nonlinear equations with inequality constraints. This model partitions the solution interval and allows for the selection of different thread numbers based on actual conditions. The model's properties guarantee convergence to zero within a predefined time in all solution threads. Finally, the robotic arm control problem is modeled as a system of nonlinear equations constrained by inequalities. The proposed neutral fuzzy logic parallel nullification neural network is used to calculate the joint control signals of the robotic arm, thereby achieving control of the robotic arm. This embodiment of the robotic arm control method based on neutral logic parallel neural networks first models the trajectory tracking task of the robotic arm, considering the joint angles and posture constraints of the robotic arm. The entire task is transformed into a system of nonlinear equations with inequality constraints. Then, relaxation variables are introduced, and the problem is transformed into solving a matrix equality problem through mathematical theory. Next, a neutral fuzzy logic system and a parallelized ZNN model are constructed. The introduction of the neutral fuzzy logic system brings excellent adaptability to the ZNN model and improves its convergence and robustness. Parallel technology improves the computational efficiency of the ZNN model, saving a significant amount of actual computation time. Finally, the corresponding matrix equality problem is solved using the NLS-PZNN model proposed in this embodiment of the robotic arm control method based on neutral logic parallel neural networks, obtaining the joint angles of the robotic arm. These joint angles are then input into the robotic arm through control commands, thereby achieving high-precision control of the robotic arm. Since the nonlinear equation system can consider most robotic arm control scenarios and constraints in the real world, and the neutral fuzzy logic system provides excellent adaptability for the NLS-PZNN model, this robotic arm control method based on neutral logic parallel neural networks can be widely applied to various task scenarios. The robotic arm control method based on neutral logic parallel neural networks in this embodiment can greatly accelerate the calculation speed of the null neural network model, shorten the calculation time, and improve the calculation efficiency. Through the neutral fuzzy logic system, the adaptability, convergence, and robustness of the null neural network model are further improved, and it can be widely applied to various robotic arm scenarios that can be modeled as nonlinear equation systems with inequality constraints.

[0135] Furthermore, this embodiment also provides a robotic arm control system based on a neutral logic parallel neural network, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the robotic arm control method based on the neutral logic parallel neural network. This embodiment also provides a computer-readable storage medium storing a computer program or instructions programmed or configured to execute the robotic arm control method based on the neutral logic parallel neural network via a processor. This embodiment also provides a computer program product, including a computer program or instructions programmed or configured to execute the robotic arm control method based on the neutral logic parallel neural network via a processor.

[0136] Those skilled in the art will understand that the technical solutions provided by this invention may take the form of a method, system, or computer program product. Therefore, this invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this invention may take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, produce an implementation of the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0137] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A robotic arm control method based on a neutral logic parallel neural network, characterized in that, Includes the following steps: S101, Input the task objective and attitude constraints, and transform the trajectory tracking problem of the robotic arm for the task objective into a system of nonlinear equations constrained by attitude constraints; S102, Solve the nonlinear equations constrained by attitude conditions using a nullable neural network (NLS-ZNN) based on a neutral fuzzy logic system to obtain the joint angles of the robotic arm, including: ① Generating the output value of the objective function of the nonlinear equations constrained by attitude conditions using a neutral fuzzy logic system (NLS). This includes: S301, which uses the objective function of the nonlinear equation system constrained by attitude constraints as the error function. S302, the error function norm As input to a neutral fuzzy logic system S303, will input By providing a pre-defined true membership function Uncertain membership function and spurious membership functions Perform neutralization processing; S304, base the neutralization processing result on the "IF-THEN" rule set. Perform neutral fuzzy reasoning, the "IF-THEN" rule set The function expression is: If the input value is "large", then the output value is "large". : If the input value is "Zhong", then the output value is "Zhong"; If the input value is "small", then the output value is "small". in, These represent the "IF-THEN" rule set. The three rules, "SE", "ME", and "NE", define the "small", "medium", and "large" ranges for the joint angles of the robotic arm, respectively. "Describe the mapping relationship between input and output; S305, deneutralize the neutral fuzzy inference result to obtain the output value." : ； in, and These are preset constant parameters. This is a preset constant used to prevent the denominator from being zero. For the preset real membership function, For a pre-defined uncertain membership function, ① A pre-defined spurious membership function; ② Output the value The nonlinear equations of the trajectory tracking problem are solved using a zero-form neural network (ZNN) to obtain the joint angles of the robotic arm; S103 sends control commands generated based on the joint angles of the robotic arm to drive the robotic arm to complete the task objective.

2. The robotic arm control method based on a neutral logic parallel neural network according to claim 1, characterized in that, Step S101 includes: S201, Input the task objective and attitude constraints, wherein the task objective includes the desired trajectory of the robotic arm's end effector. and the expected pose quaternion The expected trajectory The coordinates are , ,in For attitude quaternions The scalar part represents the magnitude of rotation; For attitude quaternions The vector part represents the direction and magnitude of rotation; the functional expression of the attitude constraint is: ,in The joint angle of the robotic arm. Indicates the lower bound of the joint angle. Indicates the upper limit of the joint angle; S202, with The joint angles of the robotic arm represent the desired trajectory. and the expected pose quaternion As the objective function of the nonlinear equation system Transform the attitude constraints into matrix inequalities. ,in The coefficient matrix, This represents the upper and lower bounds of the joint angles of the robotic arm. , , for An identity matrix of dimension 1 Representing the degrees of freedom of the robotic arm, we obtain the matrix inequality. A constrained system of nonlinear equations; S203, subject to matrix inequalities The constrained nonlinear equations are transformed into solution equations using the Lagrange multiplier method and the Karush-Kuhn-Tucker conditions. The problem, among which Let be the objective function of the transformed nonlinear equation system. Let be the state vector to be solved. , They are slack variables; and we have: ； ； in, ~ For slack variables The first to m elements in the array, where m is the slack variable. The number of elements in it.

3. The robotic arm control method based on a neutral logic parallel neural network according to claim 1, characterized in that, In step S102, the output value will be... When using a nullable neural network (ZNN) to solve the nonlinear equations of a trajectory tracking problem to obtain the joint angles of a robotic arm, the functional expression of the ZNN is as follows: ； ； ； ； in, The dynamic coefficient matrix of the system. for The first-order differential, For the system's dynamic constraint vector, The constant convergence factor For activation function, Let be the objective function of the transformed nonlinear equation system. The objective function of a system of nonlinear equations about Jacobian matrix, , for An identity matrix of dimension 1 Indicates the degrees of freedom of the robotic arm. ~ For slack variables The first to the second One element, For slack variables The number of elements in; for The first-order differential, for The first-order differential, for The first-order differential, , Indicates the joint angle of the robotic arm. Indicates the lower bound of the joint angle. Indicates the upper limit of the joint angle.

4. The robotic arm control method based on neutral logic parallel neural network according to claim 3, characterized in that, The activation function is a linear activation function, a sign double power activation function, a modified finite-time activation function, or a fast predefined time-converged activation function. The expression for the linear activation function is: ； in, It is a linear activation function. This is the input to the activation function; The function expression for the sign double power activation function is: ； ； in, For the sign double power activation function, For symbolic functions, As a regulating factor, and ; The improved finite-time activation function has the following expression: ； in, For an improved finite-time activation function, The nonlinear convergence gain coefficient is... Let be the linear convergent gain coefficient, and , ; The function expression for a fast, predefined, time-converging activation function is: ； in, As a regulating factor, and Here is the gain coefficient of the exponential term, and , , .

5. The robotic arm control method based on a neutral logic parallel neural network according to claim 1, characterized in that, In step S102, the output value will be... When using a zero-dimensional neural network (ZNN) to solve the nonlinear equations for trajectory tracking to obtain the joint angles of a robotic arm, this includes applying the task time domain... Divided into partially overlapping task sub-intervals ,and ,in For the task time domain At the end of the day, , partially overlapping task sub-intervals include: First section: ; Middle section: ; Ending interval: ; in, The convergence parameters are set at a predetermined time; for each task sub-interval, the output value will be... The joint angles of the robotic arm are obtained by solving the nonlinear equations of the trajectory tracking problem using an independent nullable neural network (ZNN).

6. A robotic arm control system based on a neutral logic parallel neural network, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to execute the robotic arm control method based on a neutral logic parallel neural network as described in any one of claims 1 to 5.

7. A computer-readable storage medium storing a computer program or instructions, characterized in that, The computer program or instructions are programmed or configured to execute the robotic arm control method based on a neutral logic parallel neural network as described in any one of claims 1 to 5 via a processor.

8. A computer program product, comprising a computer program or instructions, characterized in that, The computer program or instructions are programmed or configured to execute the robotic arm control method based on a neutral logic parallel neural network as described in any one of claims 1 to 5 via a processor.

Images