Carrying robot control method based on artificial intelligence

Abstract

The invention relates to the technical field of robot control, and discloses an artificial intelligence-based transfer robot control method, which comprises the following steps of: 1, defining an actual clamping force, a target clamping force and a clamping force error, and constructing a state dynamic relationship; 2, according to the clamping force error defined in the step 1, it is determined that control input is obtained by multiplying the self-adaptive gain by the clamping force error, and a self-adaptive gain online updating scheme is formulated; and step 3, constructing a candidate function as a weighted sum of error square and adaptive gain deviation square by adopting a high-order Lyapunov method according to the control input determined in the step 2, and verifying closed-loop stability. The intelligent grabbing optimization technical scheme is adopted, the optimal clamping force adjusting strategy is predicted in real time through the deep learning model, the technical effect of improving the grabbing success rate and the system robustness is achieved, and compared with a control scheme depending on fixed parameters in the prior art, the problem that grabbing is unstable in the complex environment is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot control, and particularly to a control method for a handling robot based on artificial intelligence. Background Art

[0002] In modern industrial automation, handling robots play a crucial role. Due to the complexity of the production environment, robots need to perform precise grasping tasks under variable conditions. However, traditional control methods rely on the setting of fixed parameters and lack flexibility and adaptability. When facing different object characteristics and environmental changes, it is easy to cause unstable grasping and affect production efficiency.

[0003] Traditional control systems usually rely on preset fixed parameters. They may be effective in static environments, but in dynamic and complex environments, fixed parameters cannot adapt to rapidly changing conditions. As a result, robots are prone to errors during the grasping process and the success rate decreases.

[0004] In the prior art, sensor noise is a common problem. High-precision force sensors can provide detailed data, but the presence of noise will interfere with the system's judgment and lead to a decrease in control accuracy. Traditional methods lack an effective noise filtering mechanism and cannot ensure the precise control of the clamping force.

[0005] Many existing control methods lack the ability to adapt and cannot adjust the control strategy according to real-time data. Control methods lacking flexibility show obvious deficiencies when facing a changing environment. Robots cannot self-adjust according to environmental changes, resulting in inflexible operation and affecting the overall performance.

[0006] Therefore, the present invention proposes a control method for a handling robot based on artificial intelligence to solve the above-mentioned problems. Summary of the Invention

[0007] Aiming at the deficiencies of the prior art, the present invention provides a control method for a handling robot based on artificial intelligence to solve the problems raised in the above background art.

[0008] To achieve the above object, the present invention is realized through the following technical solutions: A control method for a handling robot based on artificial intelligence, comprising:

[0009] Step 1, define the actual clamping force, the target clamping force, and the clamping force error, and construct a state dynamic relationship;

[0010] Step 2, determine that the control input is the adaptive gain multiplied by the clamping force error according to the clamping force error defined in Step 1, and formulate an online update scheme for the adaptive gain;

[0011] Step 3: Using the high-order Lyapunov method, construct a candidate function as the weighted sum of the square of the error and the square of the deviation of the adaptive gain based on the control input determined in Step 2, and verify the closed-loop stability;

[0012] Step 4: Define the performance index as the weighted integral of the square of the error and the square of the control energy consumption according to the closed-loop stability obtained in Step 3, and establish a Bellman dynamic programming model to derive the optimal control law;

[0013] Step 6: Integrate a high-precision force sensor and a data acquisition module according to the performance index proposed in Step 4 to realize real-time acquisition of state data and parameter estimation;

[0014] Step 6: Evaluate the control performance using the Monte Carlo method based on the real-time data obtained in Step 5, and complete the integration and experimental verification of the online control system.

[0015] Preferably, in Step 1, further defining the state dynamic relationship by defining the actual clamping force, the target clamping force, and the clamping force error includes:

[0016] Sub-step 1.1: Let F(t) be the actual clamping force, F d be the target clamping force, and e(t) = F d - F(t) be the clamping force error;

[0017] Sub-step 1.2: Establish a state dynamic formula to express the change rate of the clamping force error as:

[0018]

[0019] where is the rate of change of the clamping force error with time, θ(t) is the dynamic attenuation factor, e(t) is the clamping force error, is the input amplification factor, u(t) is the control input, and d(t) is the external disturbance;

[0020] Sub-step 1.3: Combine the actual clamping force, the target clamping force, and the clamping force error defined in Sub-step 1.1 with the state dynamic formula established in Sub-step 1.2 to form a complete system state modeling, providing a data basis for subsequent controller design.

[0021] Preferably, in Step 2, further generating the control input and formulating the adaptive gain update scheme according to the clamping force error constructed in Step 1 includes:

[0022] Sub-step 2.1: Set the control input as:

[0023] u(t) = k(t) × e(t),

[0024] where \(u(t)\) is the control input, \(k(t)\) is the adaptive gain, and \(e(t)\) is the clamping force error;

[0025] Sub-step 2.2, formulating the adaptive gain update law as:

[0026]

[0027] where is the change rate of the adaptive gain, \(\gamma\) is the adaptive adjustment rate coefficient, \(\lambda\) is the forgetting factor, and \(k_0\) is the initial gain estimate;

[0028] Sub-step 2.3, connecting the control input determined in Sub-step 2.1 with the adaptive gain update law formulated in Sub-step 2.2 to form a complete controller design.

[0029] Preferably, in the said Step 3, further including verifying the closed-loop stability by using the high-order Lyapunov method:

[0030] Sub-step 3.1, constructing a candidate function:

[0031]

[0032] where \(V(e,k)\) is the candidate Lyapunov function, \(e(t)\) is the clamping force error, \(k(t)\) is the adaptive gain, \(k\) * is the ideal gain, and \(\gamma\) is the adaptive rate coefficient;

[0033] Sub-step 3.2, taking the time derivative of the candidate function to obtain:

[0034]

[0035] where is the time derivative of the candidate function, is the rate of change of the clamping force error with respect to time, is the change rate of the adaptive gain;

[0036] Sub-step 3.3, connecting the candidate function constructed in Sub-step 3.1 with the derivative expression obtained in Sub-step 3.2 to form a complete closed-loop energy analysis.

[0037] Preferably, in the said Step 4, defining the performance index as the weighted integral of the square of the error and the square of the control energy consumption according to the closed-loop stability obtained in Step 3, and further including establishing a Bellman dynamic programming model to derive the optimal control law:

[0038] Sub-step 4.1, defining the performance index function \(J\) as the weighted integral of the square of the error and the square of the control energy consumption, and the formula is:

[0039]

[0040] Among them, J is the performance index function, e(t) is the clamping force error, u(t) is the control input, and μ is the weighting coefficient;

[0041] In sub-step 4.2, according to the performance index defined in sub-step 4.1, construct the Bellman dynamic programming model. Let the optimal value function V * (e) satisfy the Bellman equation:

[0042]

[0043] Among them, V * (e) is the optimal value function in state e, is the rate of change of the error, and Δt is the time increment;

[0044] In sub-step 4.3, according to the Bellman equation constructed in sub-step 4.2, deduce the optimal control law u * (t), and the formula is:

[0045]

[0046] Among them, u * (t) is the optimal control input, is the partial derivative of the optimal value function with respect to the error;

[0047] Through solving, obtain the optimal control strategy under the given state.

[0048] Preferably, in step 5, integrating the high-precision force sensor and the data acquisition module according to the performance index proposed in step 4 to realize real-time acquisition of state data and parameter estimation further includes:

[0049] In sub-step 5.1, integrate the high-precision force sensor on the gripper of the handling robot. The force sensor is used to measure the clamping force F(t) in real time, and transmit the measurement data to the control system through the data acquisition module;

[0050] In sub-step 5.2, use the real-time clamping force data obtained in sub-step 5.1 to calculate the clamping force error e(t): e(t) = F d - F(t),

[0051] Among them, F d is the target clamping force, F(t) is the actually measured clamping force, and the error calculation is used for subsequent adjustment of the control input;

[0052] In sub-step 5.3, according to the clamping force error calculated in sub-step 5.2, perform parameter estimation and state update, and use the Kalman filter to filter and estimate the clamping force error to update the system state variables.

[0053] Preferably, in step 6, the Monte Carlo method is used to evaluate the control performance based on the real-time data obtained in step 5. The completion of the online control system integration and experimental verification further includes:

[0054] Sub-step 6.1: Using the clamping force data and error information collected in real time in step 5, a Monte Carlo simulation model is constructed to simulate various possible environmental disturbances and system parameter changes, and evaluate the performance of the control system under different scenarios;

[0055] Sub-step 6.2: According to the Monte Carlo simulation model constructed in sub-step 6.1, calculate the performance index R of the control system. The formula is:

[0056]

[0057] where R is the performance index, N is the number of simulation times, e i is the clamping force error in the i-th simulation, u i is the control input in the i-th simulation, and μ is the weighting coefficient;

[0058] Sub-step 6.3: According to the performance index calculated in sub-step 6.2, adjust the control system parameters to complete the online control system integration, and verify the stability and robustness of the system in actual operation through experiments, ensuring accurate control and real-time adjustment of the clamping force in a dynamic environment.

[0059] Preferably, the experimental verification step includes online system testing, offline data analysis, comparison of optimal control strategies, and result feedback. Each verification step provides a basis for subsequent optimization.

[0060] A terminal device includes a flexible gripper, a high-precision force sensor, a data acquisition module, and an embedded controller. The embedded controller implements a control method for a handling robot based on artificial intelligence according to the handling robot control method.

[0061] A storage medium stores a software program for implementing a control method for a handling robot based on artificial intelligence. The software program executes the control method for a handling robot based on artificial intelligence on the terminal device to implement state modeling, controller design, stability analysis, optimal control association, and engineering implementation.

[0062] The present invention provides a control method for a handling robot based on artificial intelligence. It has the following beneficial effects:

[0063] 1. The present invention adopts an intelligent grasping optimization technical solution, and through a deep learning model, it can predict the optimal clamping force adjustment strategy in real time, achieving the technical effects of improving the grasping success rate and system robustness. Compared with the control schemes that rely on fixed parameters in the prior art, it solves the problem of unstable grasping in complex environments.

[0064] 2. The present invention utilizes a high-precision force sensor and a real-time data acquisition module, integrates a Kalman filter for state estimation, and realizes precise control of the clamping force. Compared with the traditional method, it solves the problem of insufficient control accuracy caused by sensor noise.

[0065] 3. The present invention optimizes the control strategy through Bellman dynamic programming and Monte Carlo simulation to ensure performance under different operating conditions. Compared with the control methods lacking adaptability in the prior art, it solves the deficiency of inflexible control strategies in a changing environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] To enable those skilled in the art to understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0068] The present invention will be described in detail below in conjunction with the drawings:

[0069] Embodiment:

[0070] Please refer to the attached Figure 1 , the embodiment of the present invention provides a control method for a handling robot based on artificial intelligence, including:

[0071] Step 1, define the actual clamping force, the target clamping force, and the clamping force error, and construct a state dynamic relationship;

[0072] Sub-step 1.1, let F(t) be the actual clamping force, F d be the target clamping force, and e(t) = F d -F(t) be the clamping force error;

[0073] Sub-step 1.2, establish a state dynamic formula to express the change rate of the clamping force error as:

[0074]

[0075] where is the rate of change of the clamping force error with time, θ(t) is the dynamic decay factor, e(t) is the clamping force error, is the input amplification factor, u(t) is the control input, and d(t) is the external disturbance;

[0076] Sub-step 1.3: Relate the actual clamping force, target clamping force, and clamping force error defined in Sub-step 1.1 to the state dynamic formula established in Sub-step 1.2 to form a complete system state modeling, providing a data basis for subsequent controller design;

[0077] Step 2: Determine that the control input is the adaptive gain multiplied by the clamping force error according to the clamping force error defined in Step 1, and formulate an online update scheme for the adaptive gain;

[0078] Sub-step 2.1: Set the control input as:

[0079] u(t) = k(t) × e(t),

[0080] where u(t) is the control input, k(t) is the adaptive gain, and e(t) is the clamping force error;

[0081] Sub-step 2.2: Formulate the adaptive gain update law as:

[0082]

[0083] where, is the change rate of the adaptive gain, γ is the adaptive adjustment rate coefficient, λ is the forgetting factor, and k0 is the initial gain estimate;

[0084] Sub-step 2.3: Relate the control input determined in Sub-step 2.1 to the adaptive gain update law formulated in Sub-step 2.2 to form a complete controller design;

[0085] Step 3: Use the high-order Lyapunov method to construct a candidate function as the weighted sum of the square of the error and the square of the deviation of the adaptive gain based on the control input determined in Step 2, and verify the closed-loop stability;

[0086] Sub-step 3.1: Construct the candidate function:

[0087]

[0088] where V(e,k) is the candidate Lyapunov function, e(t) is the clamping force error, k(t) is the adaptive gain, k * is the ideal gain, and γ is the adaptive rate coefficient;

[0089] Sub-step 3.2: Take the time derivative of the candidate function to obtain:

[0090]

[0091] where, is the time derivative of the candidate function, is the rate of change of the clamping force error with respect to time, is the change rate of the adaptive gain;

[0092] Sub-step 3.3: Connect the candidate function constructed in sub-step 3.1 with the derivative expression obtained in sub-step 3.2 to form a complete closed-loop energy analysis;

[0093] Step 4: Define the performance index according to the closed-loop stability obtained in step 3 as the weighted integral of the square of the error and the square of the control energy consumption, and establish a Bellman dynamic programming model to derive the optimal control law;

[0094] Sub-step 4.1: Define the performance index function J as the weighted integral of the square of the error and the square of the control energy consumption, and the formula is:

[0095]

[0096] where J is the performance index function, e(t) is the clamping force error, u(t) is the control input, and μ is the weighting coefficient;

[0097] Sub-step 4.2: Construct a Bellman dynamic programming model according to the performance index defined in sub-step 4.1, and assume that the optimal value function V * (e) satisfies the Bellman equation:

[0098]

[0099] where V * (e) is the optimal value function in state e, is the rate of change of the error, and Δt is the time increment;

[0100] Sub-step 4.3: Derive the optimal control law u * (t) according to the Bellman equation constructed in sub-step 4.2, and the formula is:

[0101]

[0102] where u * (t) is the optimal control input, is the partial derivative of the optimal value function with respect to the error;

[0103] By solving, obtain the optimal control strategy under the given state;

[0104] Step 5: Integrate a high-precision force sensor and a data acquisition module according to the performance index proposed in step 4 to realize real-time acquisition of state data and parameter estimation;

[0105] Sub-step 5.1: Integrate a high-precision force sensor on the gripper of the handling robot. The force sensor is used to measure the clamping force F(t) in real time, and transmit the measurement data to the control system through the data acquisition module;

[0106] Sub-step 5.2: Using the real-time clamping force data obtained in sub-step 5.1, calculate the clamping force error e(t): e(t) = F d - F(t),

[0107] where F d is the target clamping force, F(t) is the actually measured clamping force, and the error calculation is used for the adjustment of subsequent control inputs;

[0108] Sub-step 5.3: Based on the clamping force error calculated in sub-step 5.2, perform parameter estimation and state update. Use the Kalman filter to filter and estimate the clamping force error, and update the system state variables;

[0109] Step 6: Evaluate the control performance using the Monte Carlo method based on the real-time data obtained in Step 5, and complete the online control system integration and experimental verification;

[0110] Sub-step 6.1: Using the clamping force data and error information collected in real-time in Step 5, construct a Monte Carlo simulation model to simulate various possible environmental disturbances and system parameter changes, and evaluate the performance of the control system under different scenarios;

[0111] Sub-step 6.2: Based on the Monte Carlo simulation model constructed in sub-step 6.1, calculate the performance index R of the control system. The formula is:

[0112]

[0113] where R is the performance index, N is the number of simulation times, e i is the clamping force error in the i-th simulation, u i is the control input in the i-th simulation, and μ is the weighting coefficient;

[0114] Sub-step 6.3: Based on the performance index calculated in sub-step 6.2, adjust the control system parameters, complete the online control system integration, and verify the stability and robustness of the system in actual operation through experiments to ensure accurate control and real-time adjustment of the clamping force in a dynamic environment.

[0115] The benefits of Step 1: By defining the actual clamping force, target clamping force, and clamping force error, and constructing the state dynamic relationship, the system can accurately identify and quantify the errors in the clamping process. This provides a data basis for subsequent controller design, ensures that the control strategy can be optimized for different operating conditions, and improves the stability and accuracy of grasping;

[0116] The benefits of Step 2 are that by setting the control input as the adaptive gain multiplied by the gripping force error and formulating an online update scheme for the adaptive gain, the system can adjust the control strategy in real time to cope with environmental changes. The adaptability enables the robot to maintain efficient performance under different operating conditions, reduce grasping failures caused by fixed parameters, and improve the overall operating efficiency and success rate;

[0117] The benefits of Step 3 are that the high-order Lyapunov method is used to verify the closed-loop stability to ensure the stability of the system in a dynamic environment. By constructing a candidate function and performing closed-loop energy analysis, the system can identify and correct potential unstable factors to improve the robustness of the system and provide theoretical support for the further optimization of the control strategy, ensuring the stable operation of the robot in a complex environment;

[0118] The benefits of Step 4 are that by defining the performance index as the weighted integral of the square of the error and the square of the control energy consumption and establishing a Bellman dynamic programming model, the system can derive the optimal control law to ensure the performance under different operating conditions, optimize the balance between energy consumption and accuracy. At the same time, the system can reduce energy consumption while ensuring high precision, improving the overall economy and sustainability;

[0119] The benefits of Step 5 are that by integrating a high-precision force sensor and a data acquisition module to achieve real-time acquisition of state data and parameter estimation, the system can continuously monitor and adjust the gripping force during the operation. By processing the data with a Kalman filter, the system can effectively filter noise and improve the accuracy of the data;

[0120] The benefits of Step 6 are that by using the Monte Carlo method to evaluate the control performance, the system can perform simulations and verifications under various possible environmental disturbances and system parameter changes to verify the performance of the system in different scenarios and provide data support for the further optimization of the control strategy. Through experimental verification, the system can ensure the stability and robustness in actual operation, improving the overall reliability and user satisfaction.

[0121] In summary, the present invention significantly improves the operation efficiency and stability of the handling robot through a series of innovative control methods and optimization strategies. Each step supports the overall performance of the system, forming a complete closed-loop control system from data acquisition to control strategy optimization and then to performance verification.

[0122] The experimental verification steps include online system testing, offline data analysis, comparison of optimal control strategies, and result feedback. Each verification step provides a basis for subsequent tuning.

[0123] A terminal device includes a flexible gripper, a high-precision force sensor, a data acquisition module, and an embedded controller. The embedded controller implements an artificial intelligence-based handling robot control method according to the handling robot control method.

[0124] A storage medium stores a software program for implementing a control method for an artificial intelligence-based handling robot. The software program executes the control method for the artificial intelligence-based handling robot on a terminal device to achieve state modeling, controller design, stability analysis, optimal control correlation, and engineering implementation.

[0125] Through online system testing, offline data analysis, comparison of optimal control strategies, and result feedback, the experimental verification steps provide a basis for the continuous optimization of the system. Online system testing can monitor and adjust system performance in real time to ensure stability in actual operation. Offline data analysis identifies potential performance bottlenecks and improvement areas through in-depth data mining. Comparison of optimal control strategies helps identify the gap between the current strategy and the theoretical optimal strategy to guide further optimization. Result feedback integrates all analysis results to provide a clear direction for subsequent system tuning.

[0126] The terminal device integrates a flexible gripper, a high-precision force sensor, a data acquisition module, and an embedded controller to form a complete intelligent control system. The flexible gripper provides the ability to adapt to objects of different shapes and materials. The high-precision force sensor ensures accurate measurement of the clamping force. The data acquisition module collects operation data in real time to provide a basis for adjusting control strategies. The embedded controller executes control strategies in real time according to the control method for the handling robot to ensure the efficient operation of the system. The integrated design improves the flexibility and adaptability of the system to meet diverse industrial requirements.

[0127] The software program stored in the storage medium implements a control method for an artificial intelligence-based handling robot. By executing the program on the terminal device, the system can achieve full-process automation from data acquisition to control strategy optimization. The flexibility and scalability of the software program enable the system to quickly adapt to new application scenarios and requirements, improving the overall intelligent level and operation efficiency.

[0128] Although embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A control method for a handling robot based on artificial intelligence, characterized in that, Including: Step 1: Define the actual clamping force, target clamping force, and clamping force error, and construct the state dynamic relationship; Step 2: Determine that the control input is the adaptive gain multiplied by the clamping force error according to the clamping force error defined in Step 1, and formulate an online update scheme for the adaptive gain; Step 3: Use the high-order Lyapunov method to construct a candidate function as the weighted sum of the square of the error and the square of the deviation of the adaptive gain based on the control input determined in Step 2, and verify the closed-loop stability; Step 4: Define the performance index as the weighted integral of the square of the error and the square of the control energy consumption according to the closed-loop stability obtained in Step 3, establish a Bellman dynamic programming model, and derive the optimal control law; Step 5: Integrate a high-precision force sensor and a data acquisition module according to the performance index proposed in Step 4 to realize real-time acquisition of state data and parameter estimation; Step 6: Evaluate the control performance using the Monte Carlo method based on the real-time data obtained in Step 5, and complete the integration and experimental verification of the online control system.

2. The control method of a handling robot based on artificial intelligence according to claim 1, wherein, In the said Step 1, further including: defining the actual clamping force, target clamping force, and clamping force error to construct the state dynamic relationship; Sub-step 1.1, let F(t) be the actual clamping force, and F d be the target clamping force, e(t) = F d - F(t) is the clamping force error; Sub-step 1.2: Establish a state dynamic formula to express the change rate of the clamping force error as: Among them, is the rate of change of the clamping force error with respect to time, θ(t) is the dynamic decay factor, and e(t) is the clamping force error, is the input amplification factor, u(t) is the control input, and d(t) is the external disturbance; Sub-step 1.3: Connect the actual clamping force, target clamping force, and clamping force error defined in Sub-step 1.1 with the state dynamic formula established in Sub-step 1.2 to form a complete system state modeling, providing a data basis for subsequent controller design.

3. The control method of a handling robot based on artificial intelligence according to claim 1, characterized in that, In the said Step 2, further including: generating the control input and formulating the adaptive gain update scheme according to the clamping force error constructed in Step 1; Sub-step 2.1: Set the control input as: u(t) = k(t)×e(t), where, u(t) is the control input, k(t) is the adaptive gain, and e(t) is the clamping force error; Sub-step 2.2: Formulate the adaptive gain update law as: wherein, is the change rate of the adaptive gain, γ is the adaptive adjustment rate coefficient, λ is the forgetting factor, and k0 is the initial gain estimate; Sub-step 2.3: Connect the control input determined in Sub-step 2.1 with the adaptive gain update law formulated in Sub-step 2.2 to form a complete controller design.

4. A control method for a handling robot based on artificial intelligence according to claim 1, characterized in that, In the said Step 3, further including: using the high-order Lyapunov method to verify the closed-loop stability; Sub-step 3.1: Construct a candidate function: Among them, V(e, k) is the candidate Lyapunov function, e(t) is the clamping force error, k(t) is the adaptive gain, and k * is the ideal gain, and γ is the adaptive rate coefficient; Sub-step 3.2: Take the time derivative of the candidate function to obtain: Among them, is the time derivative of the candidate function, is the rate of change of the clamping force error with respect to time, is the rate of change of the adaptive gain; Sub-step 3.3: Connect the candidate function constructed in Sub-step 3.1 with the derivative expression obtained in Sub-step 3.2 to form a complete closed-loop energy analysis.

5. A control method for a handling robot based on artificial intelligence according to claim 1, characterized in that In the said Step 4, further including: defining the performance index as the weighted integral of the square of the error and the square of the control energy consumption according to the closed-loop stability obtained in Step 3, establishing a Bellman dynamic programming model, and deriving the optimal control law; Sub-step 4.1: Define the performance index function J as the weighted integral of the square of the error and the square of the control energy consumption, and the formula is: where, J is the performance index function, e(t) is the clamping force error, u(t) is the control input, and μ is the weighting coefficient; Sub-step 4.2: According to the performance metrics defined in sub-step 4.1, construct a Bellman dynamic programming model, and let the optimal value function V * (e) satisfy the Bellman equation: where, V * (e) is the optimal value function in state e, is the change rate of the error, and Δt is the time increment; Sub-step 4.3: Derive the optimal control law u * (t) according to the Bellman equation constructed in sub-step 4.2, with the formula: where, u * (t) is the optimal control input, is the partial derivative of the optimal value function with respect to the error; Through solution, obtain the optimal control strategy under the given state.

6. The control method of a handling robot based on artificial intelligence according to claim 1, characterized in that, In the said Step 5, further including: integrating a high-precision force sensor and a data acquisition module according to the performance index proposed in Step 4 to realize real-time acquisition of state data and parameter estimation; Sub-step 5.1: Integrate a high-precision force sensor onto the gripper of the handling robot. The force sensor is used to measure the clamping force F(t) in real time, and the measurement data is transmitted to the control system through a data acquisition module. Sub-step 5.2, using the real-time clamping force data obtained in sub-step 5.1, calculate the clamping force error e(t): e(t) = F d - F(t), Among them, F d is the target clamping force, F(t) is the actually measured clamping force, and the error calculation is used for the adjustment of subsequent control inputs; Sub-step 5.3: Based on the clamping force error calculated in sub-step 5.2, perform parameter estimation and state update. Use a Kalman filter to filter and estimate the clamping force error, and update the system state variables.

7. A control method for a handling robot based on artificial intelligence according to claim 1, characterized in that In step 6, evaluate the control performance using the Monte Carlo method based on the real-time data obtained in step 5. Completing the online control system integration and experimental verification further includes: Sub-step 6.1: Utilize the clamping force data and error information collected in real time in step 5 to construct a Monte Carlo simulation model, simulate various possible environmental disturbances and system parameter changes, and evaluate the performance of the control system under different scenarios. Sub-step 6.2: Based on the Monte Carlo simulation model constructed in sub-step 6.1, calculate the performance index R of the control system. The formula is: where R is the performance index, N is the number of simulations, and e i is the clamping force error in the i-th simulation, and u i is the control input in the i-th simulation, and μ is the weighting coefficient; Sub-step 6.3: Based on the performance index calculated in sub-step 6.2, adjust the control system parameters, complete the online control system integration, and verify the stability and robustness of the system in actual operation through experiments to ensure precise control and real-time adjustment of the clamping force in a dynamic environment.

8. A control method for a handling robot based on artificial intelligence according to claim 1, characterized in that, The experimental verification steps include online system testing, offline data analysis, comparison of optimal control strategies, and result feedback. Each verification step provides a basis for subsequent optimization.

9. A terminal device, characterized in that, It includes a flexible gripper, a high-precision force sensor, a data acquisition module, and an embedded controller. The embedded controller implements the control method of the handling robot based on artificial intelligence according to the handling robot control method.

10. A storage medium, characterized in that, A software program for implementing the control method of the handling robot based on artificial intelligence is stored. The software program executes the control method of the handling robot based on artificial intelligence on the terminal device to achieve state modeling, controller design, stability analysis, optimal control association, and engineering implementation.

Images