Hydraulic arm safety control method and device based on parallel learning and high-order CBF

Abstract

The application discloses a hydraulic arm safety control method and device based on parallel learning and high-order CBF, adopts the Lagrange method to establish a hydraulic mechanical arm dynamics equation, integrates unmodeled dynamics, structural parameter changes and external disturbances into an uncertain term, and adopts a linear parameterization form to describe the generalized uncertainty; in view of the problem of insufficient excitation in a complex environment task, a parallel learning mechanism is introduced, historical and real-time data are combined to realize parameter identification; for high relative order safety constraints (such as obstacle distance constraints and joint limiting constraints) in a task space, a family of obstacle functions is constructed; a dynamic error buffer function is introduced, and a high-order adaptive control obstacle function condition is designed; the above high-order adaptive control obstacle function constraint is embedded into a real-time quadratic programming solving problem, and the obtained input can automatically correct joint instructions to prevent constraint failure while maintaining trajectory accuracy, so that the "minimum intervention" safety control is realized.

Description

Technical Field

[0001] This invention relates to the field of engineering machinery, specifically to a hydraulic arm safety control method and device based on parallel learning and high-order CBF. It is applicable to a general safety control strategy for hydraulic robotic arms with dynamic uncertainties in complex facility maintenance and inspection environments. It can ensure the forward invariance of task space safety constraints and trajectory tracking performance under model uncertainty, external disturbances and finite excitation conditions. Background Technology

[0002] In the maintenance and inspection of large facilities, complex operations such as inspection, cutting, assembly, and grasping are often performed in extreme environments with large temperature differences, high humidity, and confined spaces. To ensure the safety of operators, remote hydraulic robotic arms are often used to replace manual labor in performing critical tasks. These robotic arms operate in confined spaces and complex environments for extended periods, with limited movement, low task frequency, and difficult maintenance, placing higher demands on the safety, stability, and robustness of the control system.

[0003] However, hydraulic robotic arm systems in such environments face technical challenges, including significant model uncertainty, complex external disturbances, insufficient continuous excitation, and stringent safety constraints. Existing safety control methods based on control barrier functions (CBF) can maintain forward invariance of the safety domain under ideal models, but their performance depends on precisely known system parameters; if model uncertainties or external disturbances exist, the barrier conditions may be violated. Traditional high-order control barrier functions (HoCBF) struggle to guarantee the feasibility of safety constraints and the consistency of parameter estimation under low-excitation conditions.

[0004] Therefore, there is an urgent need for a universal method to achieve safe control of robotic arms in complex environments with uncertain models, limited excitations, and strict safety constraints, so as to ensure the reliability and safety of remote operation equipment. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a safety control method and device for hydraulic arms based on parallel learning and a high-order adaptive control barrier function (HoACBF). This method introduces a parallel learning mechanism to achieve adaptive parameter identification under finite excitation conditions; and on this basis, it constructs a high-order adaptive control barrier function (HoACBF) to achieve safety constraint control of uncertain hydraulic robotic arm systems. This method features structural versatility, constraint scalability, and environmental adaptability, making it particularly suitable for remote hydraulic robotic arm systems in complex facility maintenance and inspection tasks.

[0006] The objective of this invention is achieved through the following technical solution: a hydraulic arm safety control method based on parallel learning and high-order CBF, the method comprising the following steps:

[0007] Step 1: Establish the dynamic equations of the hydraulic robotic arm, and integrate the unmodeled dynamics, structural parameter changes and external disturbances into an uncertainty term;

[0008] Step 2: Using a parallel learning mechanism, historical and real-time data are combined to achieve parameter identification under limited excitation conditions, thereby realizing asymptotic estimation of model uncertainties without the need for continuous excitation conditions.

[0009] Step 3: For the high relative order safety constraints in the task space, construct a family of obstacle functions; introduce a dynamic error buffer function and design a high-order adaptive control obstacle function condition. This condition dynamically reduces the error buffer function as the parameters gradually converge, thereby achieving adaptive shrinkage of the safety margin.

[0010] Step 4: Embed the high-order adaptive control obstacle function constraint into the real-time quadratic programming problem to solve for the optimal control input in real time. While maintaining trajectory accuracy, the joint commands are automatically corrected to prevent constraint failure, thus achieving minimal intervention safety control.

[0011] Furthermore, in step 1, the dynamic equations of the hydraulic manipulator are established based on the Lagrange method, and the generalized uncertainty of the uncertainty term is described in a linear parameterized form.

[0012] Furthermore, the generalized coordinate vectors, generalized velocities, and generalized accelerations of the hydraulic robotic arm joints are obtained, and a system dynamics model is established based on the Lagrange equations. Uncertain disturbances are represented by regression matrices and parameter vectors using linear parameterization.

[0013] Furthermore, in step 2, the system state vector is represented by generalized coordinate vector and generalized velocity to obtain the standard affine nonlinear form of the system dynamics; and a parallel learning adaptive law is adopted to construct an integral regression equation through the state and input data within the historical time window, thereby realizing parameter identification under finite excitation conditions; the parallel learning adaptive law adopts the form of joint updating of the current instantaneous regression driving term and the historical integral sample stack.

[0014] Furthermore, in step 3, safety constraints are defined based on a distance function between the end effector of the hydraulic robotic arm and the equipment boundary or obstacle; based on the safety constraints and the extension of the adjusted safety margin... By constructing a family of barrier functions, we can obtain the feasible conditions for higher-order control barrier functions.

[0015] Furthermore, in step 3, when there are unknown parameters in the system, the higher-order adaptive control barrier function contains the disturbance error of the unknown parameters. The disturbance error of the unknown parameters is introduced into a dynamic decay safety margin function that reflects the convergence speed of the parallel learning adaptive law to achieve adaptive correction. As time goes by and the parameter estimation converges, the dynamic decay safety margin gradually decreases, so that the constraint is gradually tightened to the ideal boundary.

[0016] Furthermore, in step 4, the quadratic programming QP constraint optimization problem is solved. While satisfying the safety conditions, the controller generates the minimum intervention correction to the nominal input, achieving an adaptive balance between performance and safety. When the system approaches the safety boundary, the QP solution automatically increases the correction force to avoid going out of bounds. When the system is far from the boundary, the constraints are not activated, and the controller output degenerates to the original nominal law, thereby maintaining control efficiency while ensuring safety.

[0017] Secondly, the present invention also provides a hydraulic arm safety control device based on parallel learning and high-order CBF, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the hydraulic arm safety control method based on parallel learning and high-order CBF.

[0018] Thirdly, the present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the aforementioned hydraulic arm safety control method based on parallel learning and high-order CBF.

[0019] Fourthly, the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the aforementioned hydraulic arm safety control method based on parallel learning and high-order CBF.

[0020] The beneficial effects of this invention are:

[0021] 1. This invention uses parallel learning as the basis for adaptive control. The integral parallel learning mechanism avoids the need for designing a system state observer by integrating historical data over time, thus improving robustness under noise.

[0022] 2. Construct a high-order adaptive control barrier function that combines parallel learning, and achieve convergence of uncertain parameters under finite excitation conditions by processing historical and real-time data in parallel.

[0023] 3. The upper bound of the error obtained by parallel learning adaptive law is embedded into the feasibility condition of the high-order control barrier function in the form of a buffer term, realizing the robust and safe execution of the remote hydraulic robotic arm system for high relative order and time-varying output constraints in complex environment facility maintenance and inspection tasks. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a block diagram of a hydraulic closed-chain robotic arm control system.

[0026] Figure 2 This is a comparison of the safety constraint experiment results under a straight trajectory and a schematic diagram of the controller output.

[0027] Figure 3 This is a comparison of the safety constraint experiment results under elliptical trajectory and a schematic diagram of the controller output.

[0028] Figure 4 This is a comparison of the safety constraint experiment results under a rectangular trajectory and a schematic diagram of the controller output.

[0029] Figure 5 This is a structural diagram of a hydraulic arm safety control device based on parallel learning and high-order CBF according to the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are merely illustrative and not intended to limit the invention.

[0031] like Figure 1 As shown, this invention provides a hydraulic arm safety control method based on parallel learning and high-order CBF, applicable to the safety control of hydraulic robotic arms with dynamic uncertainties in complex facility maintenance and inspection environments. The method includes the following steps:

[0032] Step 1: Dynamic modeling and uncertainty description of the hydraulic robotic arm.

[0033] This step establishes a dynamic model that can be used for controller and barrier function design, and clarifies the definition and structure of uncertainty terms, laying a mathematical foundation for the subsequent design of parallel learning adaptive laws and high-order adaptive control barrier functions. Taking tasks such as the maintenance and repair of nuclear power plant facilities and waste disposal in spent fuel pools as examples, hydraulic robotic arms in such complex environments need to perform precision operations under harsh environmental and space-constrained conditions. Their systems are exposed to high temperatures, high humidity, and even nuclear radiation for extended periods, exhibiting strong nonlinearity, time-varying parameters, and complex external disturbances. Traditional deterministic models cannot fully reflect these characteristics, necessitating the introduction of generalized uncertainty modeling methods.

[0034] Step 1-1, Dynamics Modeling of the Hydraulic Robotic Arm. Assume the hydraulic robotic arm has... Each joint has a generalized coordinate vector as follows: Generalized speed is Generalized acceleration is Based on the Lagrange equations, a dynamic model of the system can be established:

[0035]

[0036] in, The inertia matrix satisfies symmetric positive definiteness; Indicate the Coriolis and centrifugal terms; Represents the gravity vector; For control input; This is the equivalent uncertain disturbance term. For engineering applicability and ease of online estimation, this invention employs a linear parameterization form to represent... This can be represented as the product of the regression matrix and the parameter vector:

[0037] ,

[0038] in, The regression matrix (which can be obtained from approximate model terms, basis functions, or empirical construction); A vector of unknown constant parameters, or a vector of unknown but bounded parameters, reflects the uncertainties caused by the system's structure and environment. This expression has wide applicability: when the uncertainty is a time-varying parameter or can be approximated as a constant vector, it can be used. This indicates that for more complex perturbations, expansion can also be achieved by selecting enriched basis functions. The column space is used to approximate the perturbation.

[0039] Steps 1-2: Constructing the equivalent affine transformation and integral regression of the uncertain system. To implement the parallel learning algorithm design, the above system is transformed into a standard affine nonlinear form. The system state vector is defined as... Then the system dynamics can be expressed as:

[0040] ,

[0041] in, , ,

[0042] One of the significant characteristics of operations in complex environments is the lack of excitation signals. For example, robotic arm operations inside nuclear reactor compartments are mostly slow, repetitive, or posture-maintaining tasks, and traditional adaptive laws that rely on continuous excitation conditions cannot ensure parameter convergence. To address this issue, this invention introduces a parallel learning mechanism, constructing an integral regression equation using state and input data within a historical time window, thereby achieving parameter identification under limited excitation conditions.

[0043] Set time window Integrating over the system, using elementary integral identities, we can obtain the linear integral relationship between the state difference and the parameters:

[0044]

[0045] in For the reason The known terms obtained by integrating with the past state inputs The time integral regression matrix, This represents the control input trajectory within the interval. This integral relationship provides historical information for parallel learning, enabling the acquisition of identification information through the accumulation of historical data even when instantaneous excitation is insufficient. The above steps establish a learnable linear parameterized model and a historical regression data structure, laying the foundation for the design of the parallel learning adaptive law in step 2.

[0046] Step 2: Design and stability analysis of the parallel learning adaptive law.

[0047] After establishing and parameterizing the dynamic model of the hydraulic robotic arm, this step designs an adaptive law based on parallel learning to achieve online estimation of unknown parameters and compensation for system dynamic uncertainties. This adaptive mechanism can utilize the state and input information within the current time window and historical time windows to achieve exponential convergence of parameter errors under finite excitation conditions, thus breaking through the strict dependence of traditional adaptive laws on continuous excitation conditions. This characteristic is of great significance for low-speed, repetitive, and low-dynamic operation tasks in complex environments, ensuring that the system retains parameter identification and safety constraint maintenance capabilities during long-term stable operation.

[0048] Step 2-1, Constructing the Adaptive Law for Parallel Learning. Let the desired trajectory be... Define tracking error To achieve trajectory tracking and parameter identification, a nominal control law (such as PD or VDC based on model compensation) is used to provide nominal input. (Desired input u, u_desire), and based on this, online adaptive correction is performed. The parallel learning adaptive law is adopted in the form of joint updating of the current instantaneous regression driving term and the historical integral sample stack:

[0049]

[0050] in This refers to quantities constructed based on instantaneous errors (e.g., linear combinations of velocity / acceleration errors). This is the adaptive gain matrix. To learn weights in parallel, The historical sample stack is obtained from integral equations at multiple different points in the past, u i The first term is the input at time i, and N is the length of the history stack, i.e., the number of historical samples. The second term is the parallel learning term, which enhances parameter convergence through historical data.

[0051] Step 2-2, Proof of the stability of the parallel learning adaptive law. To verify that the proposed adaptive law can guarantee system stability and parameter error convergence under finite excitation conditions, the following Lyapunov candidate function is constructed:

[0052] ,

[0053] The first term reflects the system's kinetic energy error, and the second term reflects the parameter estimation error energy. This refers to the estimation error. Based on the symmetric positive definiteness of the inertia matrix, For all and It is a positive definite function. Taking its derivative, we get:

[0054] .

[0055] The error dynamic equation can be obtained from the robot arm's dynamic equation:

[0056] ,

[0057] in Given a symmetric positive definite matrix, substituting the derivative of the Lyapunov function and utilizing its properties... Since it is a skew-symmetric matrix, we can obtain:

[0058] ,

[0059] Will Substitute and apply the proposed parallel learning law:

[0060] ,

[0061] have to:

[0062] .

[0063] Since the historical sample stack satisfies finite excitation, i.e., there exists a constant... Make: We can obtain:

[0064] .

[0065] Assuming external disturbance Bounded, its upper bound is Then, by the standard inequality, we can obtain:

[0066] ,

[0067] in, ,in and Let represent the minimum and maximum eigenvalues ​​of the matrix, respectively. According to Lyapunov stability theory, the state error and parameter estimation error of the closed-loop system are uniformly bounded, and the system achieves asymptotic stability when the external disturbance approaches zero, i.e.:

[0068] .

[0069] The above analysis provides a theoretical basis for constructing the adaptive safety margin function in step 3. It should be noted that the above conclusions are based on common parallel learning theories and Lyapunov techniques, and the parameter convergence rate is related to the quality of historical samples and the selection of gain. In engineering implementation, appropriate sampling and storage strategies should be selected to ensure that the FE condition is approximately true.

[0070] Step 3: Design and safety proof of the high-order adaptive control barrier function.

[0071] Based on the parameter identification results in step 2, this step proposes a high-order adaptive control barrier function (HoACBF) design method for hydraulic robotic arm systems in complex environments with dynamic uncertainties and external disturbances. This method introduces a parameter estimation error compensation term and a time-decay safety margin to achieve adaptive feasible control of safety constraints. This mechanism can maintain the system state within the safety set even when the parameters have not fully converged, and gradually restore the original constraint boundary after estimation convergence, ensuring long-term safety in complex environment operations.

[0072] Step 3-1, Definition of Higher-Order Control Barrier Functions and System Constraint Modeling. Let the system's safety constraints be defined by the function... Defined, its physical meaning is a distance function between the end effector of the hydraulic robotic arm and the boundary or obstacle of the maintenance equipment. Constraints This indicates that the system is in a safe zone. Indicates that it is located at the safety boundary. This indicates entering an unsafe region; the safe set is defined as follows. .

[0073] In high relative order systems, constraint functions The time derivative needs to be continuously differentiated up to a certain order. Only then can control inputs be explicitly included. For a hydraulic robotic arm system, its dynamic equations show that the relative order is typically... That is, the control input affects the end position through the acceleration term. The family of obstacle functions is defined as follows:

[0074] ,

[0075] in To expand Class functions are used to adjust the safety margin. The feasibility conditions for higher-order control barrier functions are:

[0076] ,

[0077] in , It is a function Along the vector field and The Lie derivative. The geometric meaning of the above condition is: if The system trajectory is at the control input. Under the influence of this, it will remain within the safe set, thereby achieving the forward invariance of the constraint.

[0078] However, when the system has unknown parameters At that time, Director Li and All are affected by it. The expression for the disturbance term can be obtained from step 2:

[0079] ,

[0080] Therefore, the derivative of the barrier function of a real system should be expressed as:

[0081] ,

[0082] If not error term Making corrections would compromise the feasibility of safety constraints. Therefore, this invention introduces an adaptive compensation term into the traditional high-order control barrier function structure to dynamically correct for the impact of parameter estimation errors.

[0083] Step 3-2: Construction and modification of the adaptive higher-order control barrier function. To ensure the effectiveness of the safety constraints even when parameter estimation errors exist, a dynamically decaying safety margin function is introduced:

[0084]

[0085] in This is the initial safety margin; The attenuation coefficient reflects the convergence speed of the parallel learning law; The value is a small positive number, ensuring that the margin always has a non-zero lower bound. This function provides a large safety buffer in the initial stage, and as time progresses and parameter estimates converge, Gradually reduce the constraint to tighten it towards the ideal boundary.

[0086] Based on this definition, a higher-order adaptive control barrier function is constructed:

[0087] .

[0088] To ensure the forward invariance of the safety set, the constraints are taken as follows: This condition has an additional adaptive correction term compared to the traditional HoCBF: Used to compensate for parameter errors The effects of introduced non-ideal disturbances. In the early stages of identification, when parameter errors are large, When the value is high, the barrier function constraint is conservative; as the parallel learning adaptive law is executed, Asymptotic decay allows the controller to gradually relax the constraint conservatism, ultimately achieving precise and safe boundary control.

[0089] Step 3-3, Forward Invariance Proof (based on the Invariance Principle). The goal of the control barrier function is to guarantee the forward invariance of the safe set. According to the extended Nagumo condition, if there exists any (Boundary), existence control This makes the Lie derivative of the actual system satisfy:

[0090]

[0091] but It is a forward-invariant set. Because... If you choose Satisfy adaptive conditions Furthermore, parallel learning guarantees that the estimation error of the pair is bounded and decays over time (the conclusion obtained in step 2), thus there exists an adaptive correction term that guarantees the true Lie derivative is not less than 1 / 2. This satisfies the Nagumo condition, thus guaranteeing The forward invariance of the estimation is demonstrated by matching the upper bound of the estimation error with the invariance principle, emphasizing the synergistic effect of "the estimation is bounded and can be gradually reduced" and "the margin of time decay".

[0092] Steps 3-4 involve designing a safety-optimized control law based on a high-order adaptive control barrier function. To maintain good trajectory tracking performance while satisfying safety constraints, this invention embeds the high-order adaptive control barrier function constraints into a quadratic programming (QP) framework to solve for the optimal control input in real time. The nominal control law is defined. For trajectory tracking control output, the actual control input is determined by the following formula:

[0093] , ,

[0094] This optimization problem is in convex QP form, with low computational complexity, making it suitable for real-time execution. By solving this constrained optimization problem, the controller can generate minimal intervention corrections to the nominal input while satisfying safety conditions, achieving an adaptive balance between performance and safety. When the system approaches the safety boundary, the QP solution automatically increases the correction force to avoid exceeding the boundary; when the system is far from the boundary, the constraints are not activated, and the controller output degenerates to the original nominal law, thus maintaining high control efficiency while ensuring safety.

[0095] From the convexity of the QP solution and the continuity of the constraints, we can see that the control law Continuously differentiable, it does not introduce control jumps. Combining this with the aforementioned Lyapunov analysis, the closed-loop system, under the constraint of a high-order adaptive control barrier function, satisfies:

[0096] ,

[0097] like Figures 2-4 As shown, the system trajectory asymptotically converges to the safety set while maintaining the validity of the task space constraints throughout the process. This step achieves safe control under uncertain dynamics and finite excitation conditions by combining a high-order adaptive control barrier function with quadratic programming constraint solving. This mechanism can ensure the operational safety and control robustness of the hydraulic manipulator in high-risk tasks in complex environments, and provides a general implementation framework for adaptive safety control in multi-task coordination and complex environments.

[0098] Corresponding to the aforementioned embodiment of a hydraulic arm safety control method based on parallel learning and high-order CBF, the present invention also provides an embodiment of a hydraulic arm safety control device based on parallel learning and high-order CBF.

[0099] See Figure 5 The present invention provides a hydraulic arm safety control device based on parallel learning and high-order CBF, comprising a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement a hydraulic arm safety control method based on parallel learning and high-order CBF in the above embodiment.

[0100] The embodiment of the hydraulic arm safety control device based on parallel learning and high-order CBF provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 5 The diagram shown is a hardware structure diagram of any device with data processing capabilities, which includes a hydraulic arm safety control device based on parallel learning and high-order CBF provided by the present invention. (Except for...) Figure 5 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0101] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0102] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0103] This invention also provides a computer-readable storage medium storing a program that, when executed by a processor, implements a hydraulic arm safety control method based on parallel learning and high-order CBF as described in the above embodiments.

[0104] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0105] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned hydraulic arm safety control method based on parallel learning and high-order CBF.

[0106] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A hydraulic arm safety control method based on parallel learning and high-order CBF, characterized in that, The method includes the following steps: Step 1: Establish the dynamic equations of the hydraulic robotic arm, and integrate the unmodeled dynamics, structural parameter changes and external disturbances into an uncertainty term; Step 2: Parallel learning adaptive law is adopted to construct an integral regression equation through the state and input data within the historical time window, thereby achieving parameter identification under limited excitation conditions and realizing asymptotic estimation of the uncertain terms of the model without continuous excitation conditions; the parallel learning adaptive law adopts the form of joint updating of the current instantaneous regression driving term and the historical integral sample stack. Step 3: For the high relative order safety constraints in the task space, construct a family of obstacle functions; introduce a dynamic error buffer function and design conditions for high-order adaptive control obstacle functions. When there are unknown parameters in the system, the high-order adaptive control obstacle function contains the disturbance error of the unknown parameters. For the disturbance error of the unknown parameters, introduce a dynamic decay safety margin function that reflects the convergence speed of the parallel learning adaptive law to achieve adaptive correction; as time goes by and the parameter estimation gradually converges, dynamically reduce the error buffer function, and the dynamic decay safety margin gradually decreases, so that the constraints are gradually tightened to the ideal boundary, realizing the adaptive contraction of the safety margin. Step 4: Embed the high-order adaptive control obstacle function constraint into the real-time quadratic programming problem to solve for the optimal control input in real time. While maintaining trajectory accuracy, the joint commands are automatically corrected to prevent constraint failure, thus achieving minimal intervention safety control.

2. The hydraulic arm safety control method based on parallel learning and high-order CBF according to claim 1, characterized in that, In step 1, the dynamic equations of the hydraulic manipulator are established based on the Lagrange method, and the generalized uncertainty of the uncertainty term is described in a linear parameterized form.

3. The hydraulic arm safety control method based on parallel learning and high-order CBF according to claim 2, characterized in that, The generalized coordinate vectors, generalized velocities, and generalized accelerations of the hydraulic robotic arm joints are obtained, and a system dynamics model is established based on the Lagrange equations. Uncertain disturbances are represented by regression matrices and parameter vectors using linear parameterization.

4. The hydraulic arm safety control method based on parallel learning and high-order CBF according to claim 3, characterized in that, In step 2, the system state vector is represented by generalized coordinate vector and generalized velocity, thus obtaining the standard affine nonlinear form of the system dynamics.

5. The hydraulic arm safety control method based on parallel learning and high-order CBF according to claim 4, characterized in that, In step 3, safety constraints are defined based on the distance function between the end effector of the hydraulic robotic arm and the equipment boundary or obstacle; based on the safety constraints and the extension of the safety margin adjustment... By constructing a family of barrier functions, we can obtain the feasible conditions for higher-order control barrier functions.

6. The hydraulic arm safety control method based on parallel learning and high-order CBF according to claim 1, characterized in that, In step 4, the quadratic programming QP constraint optimization problem is solved. While satisfying the safety conditions, the controller generates the minimum intervention correction to the nominal input, achieving an adaptive balance between performance and safety. When the system approaches the safety boundary, the QP solution automatically increases the correction force to avoid going out of bounds; when the system is far from the boundary, the constraints are not activated, and the controller output degenerates to the original nominal law, thereby maintaining control efficiency while ensuring safety.

7. A hydraulic arm safety control device based on parallel learning and high-order CBF, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements a hydraulic arm safety control method based on parallel learning and high-order CBF as described in any one of claims 1-6.

8. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a hydraulic arm safety control method based on parallel learning and high-order CBF as described in any one of claims 1-6.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a hydraulic arm safety control method based on parallel learning and high-order CBF as described in any one of claims 1-6.

Images