Adaptive evaluation control method for double-link robot arm based on event trigger

Abstract

The present application relates to a kind of self-adapting evaluation control method of event-triggered double-link mechanical arm, belong to double-link mechanical arm system tracking control field.The method utilizes the omnipotence approximation theorem of neural network and the design method of backstepping method, realizes the construction of adaptive neural network controller.At the same time, introduce event-triggered mechanism, reduce the communication overhead between controller and actuator, improve the utilization efficiency of network resources.Evaluation network and execution network are combined, utilize smooth utility function and optimal tracking controller, improve the control performance and fault tolerance of system.In addition, actuator fault compensation term is designed to suppress the influence of actuator fault on system performance.The present application can overcome the difficulty of accurate modeling in conventional method and the problem of limited network resources, realize the efficient control of double-link mechanical arm system and the improvement of fault tolerance.The present application has wide application potential, is applicable to industrial automation, robot and other fields.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of double-link robot arm system tracking control, and relates to an adaptive evaluation control method for a double-link robot arm based on event triggering. BACKGROUND

[0002] In the field of double-link robot arm system control, traditional control methods are usually based on accurate modeling and continuous control strategies. However, due to the existence of system parameter uncertainty, external disturbances, and actuator faults, accurate modeling and continuous control strategies have certain limitations in practical applications.

[0003] In order to solve these problems, in recent years, some adaptive control methods have appeared, one of which is a neural network-based adaptive control method. This method uses the powerful approximation ability of neural networks to achieve adaptive adjustment of the controller by learning the nonlinear mapping relationship of the actual system. However, due to the need for a large amount of data and computing resources for neural network training, as well as the limitations of network communication, traditional neural network-based control methods have certain difficulties in practical applications.

[0004] In addition, in view of the limited network communication resources, an event-triggered control method is proposed. This method introduces an event-triggered mechanism on the communication link between the controller and the actuator, and only communicates and transmits control instructions when specific events occur, thereby reducing network load and communication overhead.

[0005] However, the current adaptive fault-tolerant control method for double-link robot arm systems has less application in event triggering. Therefore, an adaptive evaluation fault-tolerant control method for a double-link robot arm system based on event triggering is urgently needed to overcome the limitations of traditional methods and effectively improve the control performance and fault tolerance of the system. SUMMARY

[0006] Therefore, the purpose of the present application is to provide an adaptive evaluation control method for a double-link robot arm based on event triggering, which can overcome the shortcomings of traditional methods in accurate modeling and network resource utilization, and improve the control performance and fault tolerance of the double-link robot arm system. The method of the present application combines neural network control and event triggering mechanism to overcome the shortcomings of traditional methods in accurate modeling and network resource utilization, improve the control performance, real-time performance and fault tolerance of the double-link robot arm system, and has wide application prospects.

[0007] To achieve the above purpose, the present application provides the following technical solutions:

[0008] The application discloses an adaptive evaluation control method for a double-link mechanical arm based on event triggering, and first, a dynamic model of the double-link mechanical arm is established, and is discretized for controller design and evaluation network construction.

[0009] The method specifically comprises the following steps:

[0010] S1: a dynamic model of a double-link mechanical arm and a desired tracking trajectory are established, specifically comprising: a dynamic model of the double-link mechanical arm is established with joint angle positions q and angular velocities of the mechanical arm as state variables, and a desired tracking trajectory of joint angle positions q1 and q2 of the double-link mechanical arm is established;

[0011] S2: the dynamic model of the double-link mechanical arm is discretized by using an Euler method;

[0012] S3: an event triggering mechanism for a network channel between a sensor and a controller is designed;

[0013] S4: an evaluation network performance index is designed;

[0014] S5: an evaluation network based on event triggering is designed;

[0015] S6: an adaptive neural network controller based on event triggering is designed, specifically comprising: an event triggering mechanism is introduced into a network channel between the controller and an actuator, and an adaptive neural network controller based on event triggering is constructed by using a backstepping design method and Lyapunov stability theory;

[0016] S7: an execution network based on event triggering is designed;

[0017] S8: design an actuator fault compensation term.

[0018] Further, in step S1, a dynamics model of the double-link mechanical arm is established as:

[0019]

[0020]

[0021]

[0022]

[0023] wherein q = [q1, q2] T and represent the angular position and angular velocity of the joints of the double-link mechanical arm, M(q) and represent the moment of inertia matrix and the Coriolis force matrix, respectively, G(q) ∈ R 2 represents the gravity acceleration vector, τ is the input torque, τ d is an external disturbance. a2 = 2l1l2, a4 = (m1+m2)l1g, a5 = m2l2g, m1, m2 represent the mass of the link, l1 and l2 represent the length of the link, and g represents the gravity acceleration.

[0024] Further, in step S1, the expected tracking trajectory of the joint angular positions q1 and q2 of the double-link mechanical arm is established as:

[0025] y d1 = f1(q1)

[0026] y d2 = f2(q2)

[0027] wherein y d1 , y d2 are the expected tracking trajectories of q1 and q2, respectively, and f1(q1), f2(q2) are known functions.

[0028] Further, step S2 specifically includes: letting x 1，1 = q1, x 2，1 = q2, x = [x j,1 , x j,2 ] T , using a period of T, and using the Euler method to discretize the dynamics model of the mechanical arm can obtain:

[0029]

[0030] wherein j = 1, 2, and g(xj (k)) = TM -1 (x j (k)), u(k) = τ(k), d j (k) = g(x j (k))τ d (k), f(x(k)) = -TM -1 (x j,1 (k)) [C(x j,1 (k), x j,2 (k)) x j,2 (k) + G(x j,1 (k))], T denotes the period, and M denotes the moment of inertia matrix.

[0031] Further, the step S3 specifically comprises: defining the signal transmission error of the network channel between the sensor and the controller as:

[0032]

[0033] wherein k t denotes the last event trigger time, k t+1 is the next event trigger time, x(k) is the current state of the system, x(k t ) is the last transmitted system state;

[0034] Define τ j (k) as the event trigger state indicator function of the jth subsystem:

[0035]

[0036] Design the trigger condition of the transmission signal of the network channel:

[0037] k t+1 = {k ∈ N | ET1 ∨ ET2 ∨ … ET N}

[0038] wherein ET j is designed as:

[0039]

[0040] wherein A j > 0, and C j > 0 are the to-be-determined trigger condition parameters, e j,1 (k) is the system error, W j is the execution network weight update law, j = 1, 2; r j,2 is a design parameter, L s is a lipschitz constant, is g j(x j The upper bound of (k));

[0041] In (k) t ,k t+1 Within the time interval, when the triggering condition is met, x(k) t The time will be updated to the trigger time x(k). t+1 The value of x(k) is not specified; otherwise, x(k) is considered. t It will be held to the value at the time of the previous event trigger by the zero-order hold.

[0042] Furthermore, step S4 specifically includes: defining the systematic error equation:

[0043] e 1，1 (k)=x 1,1 (k)-y d1 (k)

[0044] e 2,1 (k)=x 2,1 (k)-y d2 (k)

[0045] Define utility function q j (k) and long-term performance index function Q j (k) are respectively:

[0046]

[0047] Q j (k-1)=ζ j Q j (k)+q j (k)

[0048] Where, 0 < η j ,0<ζ j <1 is an adjustable parameter.

[0049] Furthermore, step S5 specifically includes: using an evaluation network to evaluate the long-term performance index function Q from step S4. j (k) Approximation:

[0050] Q j (k)=W cj T (k)S cj (Z cj (k t ))+ε j (k)

[0051] Where, ε j (k) represents the approximation error of the network, W cj (k),j=1,2 represents the weight update law for evaluating network weights, S cj(·) is an activation function, Z cj (k t ) is the neural network input vector.

[0052] The event trigger mechanism in step S3 is introduced into the network channel between the adaptive neural network controller and the actuator, and the evaluation network weight update law is selected as:

[0053] W cj (k+1) = W cj (k) - τ j (k) γ cj λ j γ j S cj (k) [γ j Q j (k) - Q j (k-1) + q j (k)]

[0054] wherein z1(k) = z2(k), z2(k) = z3(k), z3(k) = z4(k), z4(k) = e j (k) is the input signal of the high-order neural network, and the input quantity is Z cj (k t ) = [z2(k t ), z3(k t ), z4(k t ), e j (k t )] T The high-order neural network basis function vector is selected as S(Z) = [s c1 (Z), s c2 (Z)] T ; 0 < γ cj , 0 < γ j < 1, which are design parameters to be determined.

[0055] Further, step S6 specifically comprises: designing an ideal control rate by using the backstepping method, and defining a system error equation:

[0056] e 1，1 (k) = x 1,1 (k) - y d1 (k)

[0057] e 1,2 (k) = x 1,2 (k) - α1(k)

[0058] e 2,1 (k) = x 2,1 (k) - y d2 (k)

[0059] e 2,2 (k) = x 2,2 (k) - a2(k)

[0060] wherein a1(k) and a2(k) are virtual control quantities;

[0061] According to Lyapunov stability analysis, the virtual control quantities are designed as follows:

[0062]

[0063]

[0064] The adaptive neural network controller based on event-triggered mechanism is designed as follows:

[0065]

[0066]

[0067] wherein, is a compensation term for actuator fault, W1(k), W2(k) are ideal network weights which will be designed in step S7, Z1(k t ) = [x T (k t ), y d1 (k+1)] T , Z2(k t ) = [x T (k t ), y d1 (k+1), y d2 (k+2)] T are input signals of high-order neural network, and the high-order neural network basis function vector is selected as S(Z) = [S1(Z), S2(Z)] T .

[0068] Further, step S7 specifically comprises: designing the following event-triggered mechanism-based execution network weight update law as:

[0069] W j (k+1) = W j (k) - τ j (k)Γ j S j (Z j (k))[Q j (k) + W j T (k)S j (Z j (k)]

[0070] wherein, Γ jQ is an adjustable parameter for implementing the network weight update law. j (k) is an evaluation network long-term performance index function defined in step S4, j = 1, 2.

[0071] Further, step S8 specifically includes: designing an actuator fault compensation term:

[0072]

[0073] wherein σ j is an adjustable parameter for the actuator fault compensation term, τ j (k) is an event-triggered state indicator function, e j,1 (k) is an error function.

[0074] The beneficial effects of the present application are:

[0075] 1) Adaptive performance improvement: The present application adopts an adaptive control method based on neural networks, which can learn and approximate the nonlinear mapping relationship of the double-link robot system, thereby realizing adaptive adjustment of the system dynamic characteristics. Compared with traditional precise modeling methods, the present application can better adapt to the uncertainty and changes of system parameters, improving the robustness and adaptability of the controller.

[0076] 2) Resource utilization efficiency improvement: By introducing an event-triggering mechanism, the present application reduces the communication overhead between the controller and the actuator. Only when a specific event occurs, communication and control instruction transmission are carried out, effectively saving the use of network bandwidth resources. This optimization of resource utilization can improve the real-time performance and network transmission efficiency of the system.

[0077] 3) Control performance improvement: The evaluation network and the execution network of the present application are combined, using a smooth utility function and an optimal tracking controller, which can improve the control performance of the double-link robot system. By reducing the execution network jump problem caused by the non-smooth utility function, more stable and accurate control is realized.

[0078] 4) Fault tolerance enhancement: The present application designs an actuator fault compensation term, which can detect and compensate the influence of actuator faults on system performance. By introducing a dynamic compensation term, the fault tolerance of the system to actuator faults is improved, enhancing the stability and reliability of the system.

[0079] Other advantages, objects, and features of the present application will be apparent to those skilled in the art from the following specification, which is to be taken in conjunction with the accompanying drawings, wherein: BRIEF DESCRIPTION OF DRAWINGS

[0080] In order to make the objects, technical solutions and advantages of the present application clearer, the preferred embodiments of the present application will be described in detail below with reference to the drawings, in which:

[0081] Figure 1 A schematic diagram of a double-link mechanical arm system;

[0082] Figure 2 A whole control block diagram of the double-link mechanical arm system of the embodiment of the present application;

[0083] Figure 3 A diagram of the angle position and the expected position tracking trajectory of joint 1 of the embodiment of the present application;

[0084] Figure 4 A diagram of the angle position and the expected position tracking trajectory of joint 2 of the embodiment of the present application;

[0085] Figure 5 An event trigger interval diagram of the embodiment of the present application;

[0086] Figure 6 A long-term performance index function diagram of the embodiment of the present application;

[0087] Figure 7 An execution network weight norm diagram of the embodiment of the present application;

[0088] Figure 8 An evaluation network weight norm diagram of the embodiment of the present application. DETAILED DESCRIPTION

[0089] The embodiments of the present application will be described in detail below with reference to the specific examples, and other advantages and effects of the present application can be easily understood by those skilled in the art from the content disclosed in the specification. The present application can also be implemented or applied in other different specific embodiments, and each detail in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0090] The drawings are only used for exemplary illustration, and the representation is only a schematic diagram, not a physical diagram, and should not be understood as a limitation on the present application; in order to better illustrate the embodiments of the present application, some components in the drawings can be omitted, enlarged or reduced, and do not represent the size of the actual product; it can be understood by those skilled in the art that some well-known structures and their descriptions in the drawings can be omitted.

[0091] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it is understood that if the orientations or positional relationships indicated by the terms "upper", "lower", "left", "right", "front", "back", etc. are based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the terms describing the positional relationships in the drawings are only used for exemplary illustration and cannot be understood as a limitation on the present application, for those skilled in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0092] Please refer to Figures 1-8 , the embodiment of the present application provides a kind of adaptive evaluation control method of double connecting rod mechanical arm based on event triggering, double connecting rod mechanical arm system schematic diagram as shown in Figure 1 , overall control block diagram as shown in Figure 2 , its detailed implementation process includes:

[0093] Step 1, the dynamics model of double connecting rod mechanical arm is established:

[0094]

[0095]

[0096]

[0097]

[0098] Wherein, q=[q1,q2] T And Angle position and angular velocity of mechanical arm joint are respectively represented, M (q) and Rotational inertia matrix and Coriolis moment matrix are respectively represented, G (q) ∈ R 2 Gravity acceleration vector is represented, τ is input moment, τ d External disturbance is; a2=2l1l2, a4=(m1+m2)l1g, a5=m2l2g, m1, m2 represent the mass of connecting rod, l1 and l2 represent the length of connecting rod, and g represents gravity acceleration.

[0099] The expected tracking trajectory of double connecting rod mechanical arm is established:

[0100] y d1 =F1 (q1)

[0101] y d2 =F2 (q2)

[0102] Wherein, yd1 y d2 are the desired tracking trajectories of q1, q2, respectively, and f1(q1), f2(q2) are known functions.

[0103] The relevant parameters of the double-link mechanical arm system selected in this embodiment are designed as follows:

[0104] m1 = 1 kg, m2 = 1 kg, l1 = 1 m, l2 = 1 m, g = 9.81 m / s.

[0105] The desired trajectory given in this embodiment is y d1 = 0.5 sin (πkT / 25), y d2 = 0.5 cos (πkT / 25).

[0106] Step 2, discretize the dynamics model of the double-link mechanical arm as follows:

[0107] Let x 1，1 = q1, x 2，1 = q2, x = [x j,1 , x j,2 ] T , using a period of T, the dynamics model of the mechanical arm is discretized using the Euler method to obtain:

[0108]

[0109] where j = 1, 2, g(x j (k)) = TM -1 (x j (k)), u(k) = τ(k), d j (k) = g(x j (k))τ d (k), f(x(k)) = -TM -1 (x j,1 (k))[C(x j,1 (k), x j,2 (k))x j,2 (k) + G(x j,1 (k))].

[0110] The sampling period parameter in this embodiment is designed as T = 0.1 s, the external disturbance is set as d1(k) = 0.1 sin(x 1,1 (k)), d2(k) = 0.1 cos(x 2,1 (k)), and the initial condition is set as X(0) = 0.

[0111] Step 3, design the event-triggered mechanism of the network channel between the actuator and the sensor and the controller:

[0112] Define the signal transmission error of the network channel between the sensor and the controller as:

[0113]

[0114] where k t represents the last event-triggering time, x(k) is the current state of the system, x(k t ) is the system state transmitted last time.

[0115] Define τ j (k) as the event-triggering state indicator function of the jth subsystem:

[0116]

[0117] Design the triggering condition of the signal transmitted by the network channel:

[0118] k t+1 ={k∈N|ET1∨ET2∨…ET N}

[0119] where ET j is designed as:

[0120]

[0121] where A j > 0, C j > 0 are the parameters of the triggering condition to be determined, e j,1 (k) is the system error, defined in step 4, W j is the execution network weight update law, defined in step 7, j = 1, 2, k t represents the last event-triggering time. k t is the last event-triggering time, k t+1 is the next event-triggering time. In the time interval (k t , k t+1 ), when the triggering condition is met, x(k t ) is updated to the value of the triggering time x(k t+1 ), otherwise x(k t ) will always remain the value of the last triggering time under the action of the zero-order holder.

[0122] In this embodiment, the parameter design of the event-triggering mechanism is as follows: A1 = 0.1, B1 = 1, C1 = 1, A2 = 1, B2 = 1, C2 = 1.

[0123] Step 4, design the evaluation network performance index as:

[0124] Define system error:

[0125] e 1，1 (k) = x 1,1 (k) - y d1 (k)

[0126] e 2,1 (k) = x 2,1 (k) - y d2 (k)

[0127] Define utility function q j (k) and long-term performance index function Q j (k) respectively:

[0128]

[0129] Q j (k-1) = ζ j Q j (k) + q j (k)

[0130] where 0 < η j , 0 < ζ j < 1 are adjustable parameters,

[0131] In this embodiment, the parameters of the utility function are designed as: ζ1 = 0.6, ζ2 = 0.6, η1 = 0.01, and η2 = 0.01.

[0132] Step 5, design an evaluation network based on event triggering:

[0133] Use the evaluation network to approximate the long-term performance index function Q j (k) in step 4:

[0134] Q j (k) = W cj T (k) S cj (Z cj (k t )) + ε j (k)

[0135] where ε j (k) is the approximation error of the network, W cj (k), j = 1, 2 is the weight update law of the evaluation network weight,

[0136] Introduce the event triggering mechanism in step 3 into the network channel between the adaptive neural network controller and the actuator, then select the evaluation network weight update law as:

[0137] W cj (k+1) = W cj(k)-τ j (k)γ cj λ j γ j S cj (k)[γ j Q j (k)-Q j (k-1)+q j (k)]

[0138] where, z1(k) = z2(k), z2(k) = z3(k), z3(k) = z4(k), z4(k) = e j (k) is the input signal of the high-order neural network, and the input quantity is Z cj = [z2, z3, z4, e j ] T The high-order neural network base function vector is selected as S(Z) = [s c1 (Z), s c2 (Z)] T .

[0139] In this embodiment, the parameters of the network weight update law are respectively set as: λ1 = 0.3, λ2 = 0.3, γ1 = 0.3, and γ2 = 0.5.

[0140] Step 6, design an adaptive neural network controller based on event triggering as:

[0141] An ideal control rate is designed by using backstepping method, and the system error is defined as:

[0142] e 1，1 (k) = x 1,1 (k) - y d1 (k)

[0143] e 1,2 (k) = x 1,2 (k) - α1(k)

[0144] e 2,1 (k) = x 2,1 (k) - y d2 (k)

[0145] e 2,2 (k) = x 2,2 (k) - α2(k)

[0146] Wherein, α1(k) and α2(k) are virtual control quantities;

[0147] According to Lyapunov stability analysis, the virtual control quantity is designed as:

[0148]

[0149]

[0150] The design of the adaptive neural network controller based on event-triggered mechanism is as follows:

[0151]

[0152]

[0153] wherein, is the compensation term for actuator fault, W1(k), W2(k) are ideal weights of the execution network which will be designed in step 7, Z1(k t ) = [x T (k t ), y d1 (k+1)] T , Z2(k t ) = [x T (k t ), y d1 (k+1), y d2 (k+2)] T are input signals of the high-order neural network, and the high-order neural network basis function vector is selected as S(Z) = [S1(Z), S2(Z)] T .

[0154] Step 7, the design of the execution network based on event triggering is as follows:

[0155] The execution network weight update law based on the event-triggered mechanism is designed as follows:

[0156] W j (k+1) = W j (k) - τ j (k) Γ j S j (Z j (k))[Q j (k) + W j T (k) S j (Z j (k)]

[0157] wherein, Γ j is an adjustable parameter of the execution network weight update law, and Q j (k) is the long-term performance index function of the evaluation network defined in step 4.

[0158] In this example, the adjustable parameter of the execution network weight update law is designed as follows: Γ1 = 0.3, Γ2 = 0.3

[0159] Step 8, the design of the actuator fault compensation term is as follows:

[0160]

[0161] wherein σ j is an adjustable parameter of the actuator fault compensation term, τ j (k) is an event-triggered state indicator function, e j,1 (k) is an error function.

[0162] In the embodiment, the adjustable parameters of the actuator fault compensation term are designed as σ1=0.1 and σ2=0.1, respectively.

[0163] In the embodiment, Figure 3 is a graph of the angular position of joint 1 of the embodiment of the application versus a desired position tracking trajectory, Figure 4 is a graph of the angular position of joint 2 of the embodiment of the application versus a desired position tracking trajectory. From Figure 3 and Figure 4 it can be seen that the tracking performance of the angular position of the joints of the double-link robot arm system is good, and the tracking error converges to a small region around zero. Figure 5 is a graph of the event-triggering interval of the embodiment of the application. From Figure 5 it can be seen that the time-triggered control method needs to be triggered 2000 times, while the event-triggered control method only needs to be triggered 1153 times, which is about 43% less than the network occupancy of the time-triggered control scheme, so the event-triggered control method can effectively reduce the number of triggers on the basis of ensuring the position tracking performance, thereby saving the network bandwidth resources of the system. Figure 6 is a graph of the long-term performance index function of the embodiment of the application. Figure 6 it can be seen that the output of the evaluation network gradually stabilizes around zero, further indicating that the system achieves satisfactory control performance. Figure 7 is a graph of the norm of the weights of the execution network of the embodiment of the application, Figure 8 is a graph of the norm of the weights of the evaluation network of the embodiment of the application. Figure 7 and Figure 8 indicate the boundedness of the weights of the execution network and the evaluation network.

[0164] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the application and are not limiting. Although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions, and all should be covered in the scope of the claims of the application.

Claims

1. An adaptive evaluation control method for an event-triggered dual-link robot arm, characterized by, The method specifically comprises the following steps: S1: establish a dynamics model of the double-link mechanical arm and a desired tracking trajectory, specifically including: establishing a dynamics model of the double-link mechanical arm with joint angle positions and angular velocities , torques , link masses and lengths as state variables, and establishing a desired tracking trajectory of joint angle positions and of the double-link mechanical arm; S2: discretize the dynamic model of the double-link mechanical arm by using Euler method; S3: design an event-triggered mechanism for the network channel between the sensor and the controller, specifically comprising: wherein, denotes the last event trigger time, is the next event trigger time, is the current state of the system, is the system state of the last transmission; Definitions is the event-triggered state indication function for the th subsystem: define the signal transmission error of the network channel between the sensor and the controller as: wherein are designed to: wherein, , and are pending trigger condition parameters, is a system error, is a network weight update law, ; is a design parameter, is a lipschitz constant, is an upper bound. In the time interval, when the trigger condition is met, the value will be updated to the trigger time , otherwise it will always remain the value of the last event trigger time under the action of the zero-order holder. design the trigger condition of the network channel signal transmission: S4: design the evaluation network performance index, specifically comprising: wherein , are the desired tracking trajectories of the joint angle positions and of the dual-link robot arm, respectively, , ; Defining the utility function and the long-term performance indicator function are respectively: wherein , , is an adjustable parameter; S5: design an event-triggered evaluation network, specifically including: using the evaluation network to approximate the long-term performance index function in step S4 approximation: wherein, is an approximation error of the network, is a weight update law for evaluating the network weights, is an activation function, is a neural network input vector; define the system error equation: wherein, is the input signal of the high-order neural network, and the input quantity is The high-order neural network base function vector is selected as ; , is the design parameter to be determined; introduce the event-triggered mechanism in the network channel between the adaptive neural network controller and the actuator in step S3, and then the evaluation network weight update law is selected as: S6: design the event-triggered adaptive neural network controller, specifically comprising: wherein and are virtual control quantities; introduce the event-triggered mechanism in the network channel between the controller and the actuator, and construct the event-triggered adaptive neural network controller by using the backstepping design method and Lyapunov stability theory; design the ideal control rate by using the backstepping method, and define the system error equation as: wherein, , is a compensator for actuator faults, , is an ideal network weight, , is an input signal of a high-order neural network, and a high-order neural network base function vector is selected as ; design the virtual control quantity according to the Lyapunov stability analysis as: wherein, is an adjustable parameter for performing the network weight update law, is an evaluation function for the long-term performance index of the network, ; the adaptive neural network controller based on the event-triggered mechanism is:

2. The adaptive evaluation control method according to claim 1, characterized by, S7: design the event-triggered execution network, specifically comprising: wherein, and denote the angular position and angular velocity of the double-link robot arm joint, respectively, and denote the moment of inertia matrix and the Coriolis matrix, respectively, denotes the gravity acceleration vector, is the input torque, is the external disturbance; , , , , , , denotes the mass of the link, and denote the length of the link, denotes the gravity acceleration.

3. The adaptive evaluation control method according to claim 2, characterized by, In step S1, the joint angle positions of the dual-link robot arm are established and The desired tracking trajectory is: wherein , are respectively , the desired tracking trajectory, is a known function.

4. The adaptive evaluation control method according to claim 3, characterized by, Step S2 specifically includes: letting , the dynamics model of the mechanical arm is discretized by Euler method with period T: wherein , , , , , denotes the adoption of a periodicity, denotes the moment of inertia matrix.

5. The adaptive critiquing control method of claim 1, wherein, design the execution network weight update law based on the event-triggered mechanism as: S8: design the actuator fault compensation term. In step S1, the dynamic model of the double-link mechanical arm is established as: step S8 specifically comprises: design the actuator fault compensation term: wherein is an adjustable parameter for the actuator fault compensation term, is an event-triggered state indicator function, is an error function.

Images