Iterative learning method for mechanical arm system with initial state drift
By introducing an event-driven adaptive iterative learning control method in the robotic arm system, the problems of initial state drift and resource waste are solved, the control accuracy and flexibility are improved, and the application scenarios are expanded.
Patent Information
- Application Number
- CN202411154211.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-05-16
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing iterative learning control method has the problem of initial state drift in the robotic arm system, which leads to low control accuracy and resource utilization efficiency, making it difficult to effectively solve the problems of inconsistent initial angles and waste of communication resources.
Adaptive iterative learning control method based on event-driven, and the adaptive iterative learning law is proposed by establishing a robotic arm dynamic model, defining the subsystem coordinate transformation and compensation error, constructing the Liyapunov function, designing the virtual controller and event-driven mechanism. To estimate unknown parameters, an event-driven adaptive iterative learning control structure is formed.
It improves the control accuracy and flexibility of the robotic arm, reduces the waste of communication resources, expands the application scenarios, and realizes high-precision control in initial angle adjustment.
Smart Images

Figure CN118938684A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of robotics technology, and more specifically, relates to an iterative learning method for a robotic arm system with initial state drift. Background Art
[0002] In recent years, with the continuous improvement of artificial intelligence and control methods, robotics technology has been successfully applied to industrial production, daily home, etc. Among them, the research on robotic arms will greatly reduce labor cost losses and improve production efficiency.
[0003] For the research on the control performance of the robotic arm system, the following two issues need to be considered: 1. During repeated operation, the initial angles of the rigid links of the robotic arm are inconsistent. The robotic arm can learn from previous experience and improve the control performance of the system; 2. While maintaining stable operation, reduce the waste of communication resources.
[0004] At present, iterative learning control has been widely used in robotic arm systems. It has the advantages of learning from previous experience to improve performance and high-precision tracking. At the same time, adaptive control is also a key technology that affects the performance of the power system. In order to solve the problem of uncertain parameters in repetitive operation systems, domestic and foreign scholars have conducted many studies on adaptive iterative learning. However, these methods are all based on time-driven control, but the communication resources between the controller and the power system are limited, and the previous control methods that require time-driven control have been challenged. How to design an iterative learning method for a robotic arm system with initial state drift in view of the impact of initial state drift on the control of the robotic arm system is a problem that needs to be solved urgently in this field.
[0005] In order to improve the operating performance of the robotic arm, it is necessary to design a control system based on event-driven control with strong learning ability and angle tracking accuracy to address the power system control problem, so as to ensure the angle adjustment performance of the power system. Summary of the invention
[0006] The technical problem to be solved by the iterative learning method for a robotic arm system with initial state drift disclosed in the present invention is to ensure that the robotic arm can achieve high-precision control based on event-driven while solving the problem of non-repetitive initial angles, so that it has a wider application scenario.
[0007] In order to achieve the above-mentioned invention objectives, the present application adopts the following technical solutions:
[0008] S1: Establish a dynamic model of the robot arm with initial drift;
[0009] S2: define the first-order subsystem coordinate transformation and compensation error;
[0010] S3: construct Lyapunov function and design virtual controller;
[0011] S4: define the second-order subsystem coordinate transformation and compensation error;
[0012] S5: Establish event-driven mechanism and design control input signals;
[0013] S6: Propose an adaptive iterative learning law to estimate unknown parameters.
[0014] In step S1, the establishment of a mechanical arm dynamics model with initial drift specifically includes:
[0015] The robot arm is a nonlinear system with uncertain parameters and multiple variables, and its system dynamics model is:
[0016]
[0017] Among them, q 1,k represents the angle of the rigid link of the robotic arm system at the kth iteration; q 2,k represents the angular velocity of the rigid link of the manipulator system at the kth iteration; represents the angular acceleration of the rigid link of the robotic arm system at the kth iteration; u k represents the input of the rigid link of the robotic system at the kth iteration; y k represents the output of the rigid link of the robotic arm system at the kth iteration; J is the moment of inertia of the servo motor; B is the unknown damping coefficient; M is the total mass of the link; L is the length from the joint axis to the center of gravity; and g is the acceleration due to gravity.
[0018] In step S2, the first-order subsystem coordinate transformation and error compensation are defined, specifically including:
[0019] Considering the initial state drift and combining the error tracking method, the desired trajectory can be defined for the robotic arm system:
[0020]
[0021] Among them, τ is the setup time; q 1,k (0) is the initial angle of the rigid link of the manipulator system; μ(t) is the decay trajectory;
[0022] This trajectory satisfies:
[0023] μ(t) is a monotonically decreasing trajectory in [0,τ];
[0024] μ(0)=1,μ(τ)=0;
[0025] μ (i-1) (τ)=0,i=1,2;
[0026] For the desired trajectory, the following coordinate transformation and compensation error can be defined:
[0027]
[0028] v 1,k =z 1,k -ζ 1,k ;
[0029] in, is the expected trajectory; v 1,k is the compensation error; 1,k is the first-order compensation signal, whose derivative satisfies
[0030] In step S3, constructing a Lyapunov function and designing a virtual controller specifically includes:
[0031] For the defined coordinate transformation and compensation error, the following Lyapunov functions can be chosen:
[0032]
[0033] Among them, v i,k To compensate for the error;
[0034] For the Lyapunov function, taking the derivative we get:
[0035]
[0036] Among them, v 2,k is the second compensation error;
[0037] Combined with the backstepping design process, a virtual controller is designed:
[0038]
[0039] Among them, c1 is a positive weight coefficient; are the desired trajectories of the rigid link angle and angular velocity of the robotic arm system, respectively.
[0040] In step S4, the second-order subsystem coordinate transformation and error compensation are defined, specifically including:
[0041] Consider the following nonlinear tracking differentiator:
[0042]
[0043] Among them, α 1,k is the first virtual controller; 2,k To track the differentiator output, it is used as the estimated value of the virtual controller; ω 2,k is φ 2,kThe derivative of; R, δ1, δ2, l1, l2 are the differentiator design parameters; combined with the above tracking differentiator, the following second-order subsystem state transformation and compensation error are designed:
[0044]
[0045] v 2,k =z 2,k -ζ 2,k ;
[0046] Among them, q 2,k is the angle of the rigid link of the robot arm; is the expected trajectory; φ 2,k is the nonlinear tracking differentiator output; v 2,k is the compensation error; 2,k is the tracking error; 2,k is the second-order compensation signal, whose derivative satisfies
[0047] In step S5, the event-driven mechanism is constructed and the control input signal is designed, which specifically includes:
[0048] Combining the coordinate transformation and compensation error defined above, the following Lyapunov function is selected:
[0049]
[0050] Among them, v 2,k is the second compensation error;
[0051] For the Lyapunov function, taking the derivative we get:
[0052]
[0053] Among them, c2 is a positive design parameter;
[0054] According to the backstepping design process, the following virtual controller is designed:
[0055]
[0056] in, is the k-th estimated value of B;
[0057] In order to achieve effective use of communication resources, the relative threshold event-driven mechanism is combined:
[0058] u k (t) = α 2,k (t h ),[t h ,t h+1 ),h∈Z + ;
[0059] t h+1 = inf{t>t h :|e(t)|≥λ|u k (t)|+a0};
[0060] Among them, α 2,k is the second virtual controller, serving as the input of the driving mechanism; u k is the actual controller, which is the output of the driving mechanism; t h is the hth triggering moment; 0<λ<1 and a0>0 are design parameters.
[0061] In step S6, the step of proposing an adaptive iterative learning law and estimating unknown parameters specifically includes:
[0062] Combined with the iterative learning law design process, the following composite energy function is selected:
[0063]
[0064] Among them, V k =V 1,k +V 2,k ; g0 is a positive design parameter; is the k-th parameter estimation error;
[0065] For the composite energy function, the difference can be obtained:
[0066]
[0067] Where, ΔE k =E k -E k-1 ; is the k-1th parameter estimation error;
[0068] Since V k (0) = 0, we can get:
[0069]
[0070] Combined with the above backstepping design process, the derivative of the Lyapunov function is:
[0071]
[0072] Substituting ΔE k We can get:
[0073]
[0074] Combining the above results, the last term can be transformed into:
[0075]
[0076] Combined with the properties of the projection operator, the following adaptive iterative learning law can be designed:
[0077]
[0078] in, is the estimated value of B in the kth batch, is the estimated value of B in the k-1th batch; is the projection operator; g0 is the adjustment parameter;
[0079] Thus, we can get:
[0080]
[0081] Beneficial effects:
[0082] 1. The present invention discloses an iterative learning method for a robotic arm system with initial state drift, which introduces an event-driven adaptive iterative learning control structure into a robotic arm control system for the first time. Specifically, an event-driven control algorithm is designed in an adaptive iterative learning controller to form an event-driven adaptive iterative learning control structure, thereby improving the control accuracy and flexibility of the robotic arm and having a wider range of application scenarios.
[0083] 2. The present invention discloses an iterative learning method for a robotic arm system with initial state drift. Taking into account the limited communication resources between the control system and the robotic arm system, an event-driven mechanism is added. The control signal is sent to the system only when it meets certain conditions. This ensures the control accuracy of the robotic arm while having good resource conservation, thus meeting the requirements of the robotic arm system control for angle adjustment.
[0084] After reading the specific embodiments of the present invention in conjunction with the accompanying drawings, other features and advantages of the present invention will become more clear. BRIEF DESCRIPTION OF THE DRAWINGS
[0085] Figure 1 A flow chart of an iterative learning method for a robotic arm system with initial state drift proposed by the present invention;
[0086] Figure 2 Output curves of the robot arm system under different batches proposed by the present invention;
[0087] Figure 3 The error curves of the nonlinear tracking differentiator proposed by the present invention and the traditional linear tracking differentiator;
[0088] Figure 4 (a) and Figure 4 (b) are the relative threshold trigger interval curves proposed by the present invention and the traditional fixed threshold trigger interval curves. DETAILED DESCRIPTION
[0089] In order to better illustrate the purpose and advantages of the present invention, the invention is compared with a conventional linear tracking differentiator and a fixed threshold driving control method, and the invention is further described in detail in conjunction with the embodiments and corresponding drawings.
[0090] The present invention aims to solve the problem of network congestion caused by limited communication resources between the control system and the manipulator system, adopts a relative threshold driving mechanism as the manipulator controller update strategy, introduces an adaptive iterative learning control structure into the manipulator system control, and introduces an error tracking method to solve the problem of inconsistent initial angles of rigid links when the manipulator repeatedly runs. The assumption of the same initial conditions is relaxed, and an iterative learning method for a manipulator system with initial state drift is proposed. The specific implementation method of the manipulator system control method is described in detail below.
[0091] See also Figure 1 As shown, the present embodiment discloses an iterative learning method for a robotic arm system with initial state drift, comprising the following steps:
[0092] Step S1: Establish a robot arm dynamics model with initial drift.
[0093] The robot arm is a nonlinear system with uncertain parameters and multiple variables. Considering the influence of initial state drift, the following second-order system is given:
[0094]
[0095] Among them, q 1,k represents the angle of the rigid link of the manipulator drive system at the kth iteration; q 2,k represents the angular velocity of the rigid link of the manipulator drive system at the kth iteration; represents the angular acceleration of the rigid link of the manipulator drive system at the kth iteration; u k represents the input of the rigid link of the manipulator drive system at the kth iteration; y k represents the output of the rigid link of the robot drive system at the kth iteration; J is the moment of inertia of the servo motor; B is the unknown damping coefficient; M is the total mass of the link; L is the length from the joint axis to the center of gravity; g is the value of the gravitational acceleration.
[0096] Step S2: Define the first-order subsystem coordinate transformation and compensation error.
[0097] Specifically:
[0098] Step S21: Considering the initial state drift, designing the first-order state transition and compensation error, and introducing the error tracking method, the following expected trajectory can be defined for the robotic arm system:
[0099]
[0100] Among them, τ is the set time; q 1,k (0) is the initial angle of the rigid link of the manipulator drive system; μ(t) is the attenuation trajectory.
[0101] This trajectory satisfies:
[0102] μ(t) is a monotonically decreasing trajectory in [0,τ]; μ(0)=1,μ(τ=0; μ (i-1) (τ)=0.
[0103] Step S22: Next, for the desired trajectory, the following coordinate transformation may be defined:
[0104]
[0105] Among them, q 1,k is the angle of the rigid link of the robot drive system; is the expected trajectory.
[0106] In order to facilitate subsequent design, the compensation error is defined as:
[0107] v 1,k =z 1,k -ζ 1,k (4)
[0108] Among them, 1,k is the first-order compensation signal, whose derivative satisfies
[0109] Step S3: See Figure 2 As shown in Figure 1, when considering the influence of initial state drift, the Lyapunov function is selected to design the first-order virtual controller.
[0110] Specifically:
[0111] Step S31: For the defined coordinate transformation and compensation error, the following Lyapunov function may be selected:
[0112]
[0113] Among them, v 1,k is the first-order compensation error.
[0114] Step S32: For the Lyapunov function, the following derivative is obtained:
[0115]
[0116] Among them, v 2,k is the second-order compensation error.
[0117] Step S32: thereby obtaining a virtual controller:
[0118]
[0119] Among them, c1 is a positive weight coefficient.
[0120] Step S4: Estimate the virtual controller derivative using the tracking differentiator to define the second-order subsystem state transition and compensation error.
[0121] Specifically:
[0122] Step S41: Use the tracking differentiator to estimate the virtual controller derivative to simplify the controller design process.
[0123] First consider the following linear tracking differentiator:
[0124]
[0125] Nonlinear tracking differentiator:
[0126]
[0127] Among them, α 1,k is the first-order virtual controller; φ 2,k To track the differentiator output, it is used as the estimated value of the virtual controller; ω 2,k is φ 2,k The derivative of ; R, δ1, δ2, l1, l2 are the differentiator design parameters.
[0128] Step S42: In combination with the above tracking differentiator, the following second-order subsystem state transformation is designed:
[0129]
[0130] Among them, q 2,k is the angle of the rigid link of the robot arm; is the expected trajectory; φ 2,k The output of the non-sexual tracking differentiator;
[0131] To facilitate subsequent design, the compensation error is defined:
[0132] v 2,k =z 2,k -ζ 2,k (9)
[0133] ζ 2,k is the second-order compensation signal, whose derivative satisfies
[0134] Step S5: Select the second-order Lyapunov function, design a virtual controller, and combine it with an event-driven mechanism to reduce the number of actual controller updates.
[0135] The details are as follows:
[0136] Step S51: Similar to step 3, in combination with the coordinate transformation and compensation error defined above, the following Lyapunov function is selected:
[0137]
[0138] Among them, v 2,k is the second-order compensation error;
[0139] Step S52: For the Lyapunov function, the following derivative is obtained:
[0140]
[0141] Among them, c2 is a positive design parameter;
[0142] Step S53: According to the back-stepping design process, the following virtual controller can be designed:
[0143]
[0144] in, is the k-th estimate of B.
[0145] Step S54: In order to realize the effective utilization of communication resources, a fixed threshold event-driven mechanism is proposed:
[0146] u k (t) = α 2,k (t h ),[t h ,t h+1 ),h∈Z +
[0147] t h+1 = inf{t>t h :|e(t)|≥a0}
[0148] Design a relative threshold event-driven mechanism:
[0149]
[0150] Among them, α 2,k is the second-order virtual controller, serving as the input of the trigger system; u k is the actual controller, serving as the output of the trigger system; t h is the hth triggering moment; 1>λ>0 and a0>0 are design parameters.
[0151] Step 6: Design an adaptive iterative learning law to estimate the unknown robot arm parameters.
[0152] Step S61: Composite energy function:
[0153] Lyapunov function: V k =V 1,k +V 2,k (15)
[0154] Estimated error: Among them, g0 is a positive design parameter.
[0155] Step S62: For the composite energy function, performing differentiation can obtain:
[0156]
[0157] Where, ΔE k =E k -E k-1 ; is the k-1th parameter estimation error;
[0158] Since V k (0) = 0, we can get:
[0159]
[0160] Combined with the above backstepping design process results, the Lyapunov function satisfies the following conditions:
[0161]
[0162] Substituting ΔE k We can get:
[0163]
[0164] Step S63: Combining the above results, the last item can be converted into:
[0165]
[0166] Combined with the properties of the projection operator, the following adaptive iterative learning law can be designed:
[0167]
[0168] in, is the estimated value of B in the kth batch, is the estimated value of B in the k-1th batch; is the projection operator; g0 is the adjustment parameter;
[0169] Thus, we can get:
[0170]
[0171] The parameter values of the robotic arm system used in the embodiment are set to the moment of inertia J = 1, B = 1, MgL = 10; the controller related parameters are set to: c1 = 0.3, c2 = 0.3, g0 = 0.13; the initial angle of the rigid link is: q 1,k =0.3+0.01rand(k); sampling time is 0.01s.
[0172] Table 1 Number of event-driven times in different batches
[0173] Iterations Fixed threshold Relative Threshold 1 2224 605 10 1608 553 20 1305 501 30 808 436 50 263 251
[0174] For comparison in the robotic arm system, different tracking differentiators ( Figure 3 ) and event-driven mechanisms ( Figure 4 The same robot parameters and control coefficients as in Table 1 are used. The linear tracking differentiator parameters are set to: R = 200; the nonlinear tracking differentiator parameters are set to: R = 200, δ1 = 2, δ2 = 3, δ3 = 4, l1 = l2 = 99 / 100; the fixed threshold parameters are set to: a0 = 0.1; the relative threshold parameters are set to: λ = 0.1; a1 = 0.05.
[0175] Depend on Figure 2 It can be seen that: considering the initial state drift, within 5 seconds, the 1st, 10th, 20th and 30th robot arm rigid link angles under the adaptive iterative learning control method are stable at 0, and as the number of iterations increases, the fluctuation amplitude decreases significantly from 0.125 in the first time to 0.04 in the 30th time; when the 50th iteration, the robot arm rigid link angle is between 0 and 0.2 seconds, and the robot arm rigid link angle is stable at 0; it can be found that the event-driven adaptive iterative learning control method can gradually improve the control performance of the robot arm system by learning previous information.
[0176] Depend on Figure 3 It can be seen that: under the influence of the initial drift, the linear tracking differentiator achieved stable tracking within 0.24 seconds, the error fluctuated significantly three times, and the maximum reached 0.27; the nonlinear tracking differentiator achieved stable tracking within 0.2 seconds, the error fluctuated only twice, and the maximum was only 0.16; comparison found that the nonlinear tracking differentiator proposed in the present invention is 0.04 seconds faster than the linear tracking differentiator in terms of stabilization speed, the oscillation frequency is reduced by 1 time, and the oscillation peak is reduced by 0.11. Compared with the linear tracking differentiator, the proposed nonlinear tracking differentiator has a better tracking effect.
[0177] Depend on Figure 4It can be seen from Table 1 that under the influence of initial drift, the fixed threshold event-driven mechanism stops updating the controller signal after 7.87 seconds, and the first iteration triggers 2201 updates. As the number of iterations increases, the number of triggers decreases, and the 50th iteration is reduced to 263 times; the relative threshold event-driven mechanism stops updating the controller signal after 4.15 seconds, and the first iteration triggers 605 updates. As the number of iterations increases, the number of triggers decreases, and the 50th iteration is reduced to 251 times; comparison shows that the relative threshold driving mechanism proposed in the present invention increases the stabilization speed by 3.72 seconds compared with the fixed threshold, the number of triggers in the first iteration is reduced by 1596 times, and the number of triggers in the 50th iteration is reduced by 12 times, and the number of triggers is significantly reduced; compared with the fixed threshold event-driven mechanism, the proposed relative threshold event-driven mechanism has fewer trigger times, which can effectively save communication resources.
[0178] This embodiment mainly uses the system output angle as a performance parameter. Figure 2 The rigid link angle curves of the robotic arm system using an event-driven adaptive iterative learning control algorithm under different batches are given. Figure 3 The error effect curves of the estimated virtual controller using linear and nonlinear tracking differentiators are given. Figure 4 The trigger interval curves of the fixed and relative threshold strategies are given. It can be seen from the figure that compared with the linear tracking differentiator and the fixed threshold driving strategy, the control method given by the present invention enables the manipulator system to have better control performance under the influence of the initial state drift.
[0179] The specific description above further illustrates the purpose, technical solutions and beneficial effects of the invention in detail. It should be understood that the embodiments are only used to explain the invention and are not used to limit the scope of protection of the invention. Any modifications, equivalent replacements, improvements, etc. under the same principles and concepts as the invention should be included in the scope of protection of the invention.
Claims
1. An iterative learning method for a robotic arm system with initial state drift, characterized in that: The following steps are involved: S1: Establish a dynamic model of the robot arm with initial drift; S2: define the first-order subsystem coordinate transformation and compensation error; S3: construct Lyapunov function and design virtual controller; S4: define the second-order subsystem coordinate transformation and compensation error; S5: Establish event-driven mechanism and design control input signals; S6: Propose an adaptive iterative learning law to estimate unknown parameters.
2. The iterative learning method for a robotic arm system with initial state drift according to claim 1, characterized in that: In S1, a dynamic model of the robot arm with initial drift is established, including: The robot arm is a nonlinear system with uncertain parameters and multiple variables, and its system dynamics model is: Among them, q 1,k represents the angle of the rigid link of the robotic arm system at the kth iteration; q 2,k represents the angular velocity of the rigid link of the manipulator system at the kth iteration; represents the angular acceleration of the rigid link of the robotic arm system at the kth iteration; u k represents the input of the rigid link of the robotic system at the kth iteration; y k represents the output of the rigid link of the robotic arm system at the kth iteration; J is the moment of inertia of the servo motor; B is the unknown damping coefficient; M is the total mass of the link; L is the length from the joint axis to the center of gravity; and g is the acceleration due to gravity.
3. The iterative learning method for a robotic arm system with initial state drift according to claim 1, characterized in that: In S2, the first-order subsystem coordinate transformation and compensation error are defined, including: Considering the initial state drift and combining the error tracking method, the desired trajectory can be defined for the robotic arm system: Among them, τ is the setup time; q 1,k (0) is the initial angle of the rigid link of the manipulator system; μ(t) is the decay trajectory; This trajectory satisfies: μ(t) is a monotonically decreasing trajectory in [0,τ]; μ(0)=1,μ(τ)=0; m (i-1) (τ)=0,i=1,2; For the desired trajectory, the following coordinate transformation and compensation error can be defined: v 1,k =z 1,k -ζ 1,k ; in, is the expected trajectory; v 1,k is the compensation error; 1,k is the first-order compensation signal, whose derivative satisfies 4. The iterative learning method for a robotic arm system with initial state drift according to claim 1, characterized in that: In S3, the Lyapunov function is constructed and the virtual controller is designed, including: For the defined coordinate transformation and compensation error, the following Lyapunov functions can be chosen: Among them, v i,k To compensate for the error; For the Lyapunov function, taking the derivative we get: Among them, v 2,k is the second compensation error; Combined with the backstepping design process, a virtual controller is designed: Among them, c1 is a positive weight coefficient; and are the desired trajectories of the rigid link angle and angular velocity of the robotic arm system, respectively.
5. The iterative learning method for a robotic arm system with initial state drift according to claim 1, characterized in that: In S4, the second-order subsystem coordinate transformation and compensation error are defined, including: Consider the following nonlinear tracking differentiator: Among them, α 1,k is the first virtual controller; 2,k To track the differentiator output, it is used as the estimated value of the virtual controller; ω 2,k is φ 2,k The derivative of; R, δ1, δ2, l1, l2 are the differentiator design parameters; combined with the above tracking differentiator, the following second-order subsystem state transformation and compensation error are designed: v 2,k =z 2,k -ζ 2,k ; Among them, q 2,k is the angle of the rigid link of the robot arm; is the expected trajectory; φ 2,k is the nonlinear tracking differentiator output; v 2,k is the compensation error; 2,k is the tracking error; 2,k is the second-order compensation signal, whose derivative satisfies 6. The iterative learning method for a robotic arm system with initial state drift according to claim 1, characterized in that: In S5, an event-driven mechanism is constructed and control input signals are designed, including: Combining the coordinate transformation and compensation error defined above, the following Lyapunov function is selected: Among them, v 2,k is the second compensation error; For the Lyapunov function, taking the derivative we get: Among them, c2 is a positive design parameter; According to the backstepping design process, the following virtual controller is designed: in, is the k-th estimated value of B; In order to achieve effective use of communication resources, the relative threshold event-driven mechanism is combined: u k (t)=α 2,k (t h ),[t h ,t h+1 ),h∈Z + ; t h+1 =inf{t>t h :|e(t)|≥λ|u k (t)|+a0}; Among them, α 2,k is the second virtual controller, serving as the input of the driving mechanism; u k is the actual controller, which is the output of the driving mechanism; t h is the hth triggering moment; 0<λ<1 and a0>0 are design parameters.
7. The iterative learning method for a robotic arm system with initial state drift according to claim 1, characterized in that: In S6, an adaptive iterative learning law is proposed to estimate unknown parameters, including: Combined with the iterative learning law design process, the following composite energy function is selected: Among them, V k =V 1,k +V 2,k ; g0 is a positive design parameter; is the k-th parameter estimation error; For the composite energy function, the difference can be obtained: Where, ΔE k =E k -E k-1 ; is the k-1th parameter estimation error; Since V k (0) = 0, we can get: Combined with the above backstepping design process, the derivative of the Lyapunov function is: Substituting ΔE k We can get: Combining the above results, the last term can be transformed into: Combined with the properties of the projection operator, the following adaptive iterative learning law can be designed: in, is the estimated value of B in the kth batch, is the estimated value of B in the k-1th batch; is the projection operator; g0 is the adjustment parameter; Thus, we can get:
Citation Information
Cited By
High-precision position tracking AILC control method for industrial mechanical arm under complex working conditions
CN119781291A
Self-adaptive preset time control method and system for flexible mechanical arm
CN121061890A
