A method for obstacle avoidance trajectory correction and control of a pneumatic muscle multi-axis robotic arm.

Abstract

This invention relates to the field of humanoid soft robot control technology, and provides a method and control method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis manipulator. The method includes: acquiring the target trajectory of the pneumatic muscle multi-axis manipulator; transforming the target trajectory into a two-dimensional planar trajectory using the Mercator projection method; acquiring the relative position and shape of obstacles, calculating external forces, introducing the external forces into the trajectory generation equation of the dynamic motion primitive algorithm, and generating the obstacle avoidance trajectory of the pneumatic muscle multi-axis manipulator based on the two-dimensional planar trajectory using the dynamic motion primitive algorithm, and performing an inverse transformation using the Mercator projection method. This solves the problems of lag and insufficient accuracy in traditional trajectory planning under complex environments.

Description

Technical Field

[0001] This invention belongs to the field of humanoid soft robot control technology, and particularly relates to a method and control method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis manipulator. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Pneumatic Artificial Muscles (PAMs) possess advantages such as lightweight, high power output, high power-to-weight ratio, no need to consider mechanical friction / wear, and high reliability, exhibiting excellent biomimetic characteristics and environmental compatibility. Compared to rigid actuators, PAMs have significant advantages in human-machine interaction, enabling them to more realistically cooperate with / replace human labor in completing precision tasks, and are characterized by high compliance, controllability, and safety.

[0004] Currently, pneumatic muscle soft robots face the following critical problems that urgently need to be solved:

[0005] (1) Traditional trajectory planning cannot avoid obstacles in a timely manner in unstructured scenarios (such as multiple obstacles and irregular obstacles), and has problems such as adjustment lag and insufficient accuracy.

[0006] (2) Vibration and overshoot problems of flexible structures. Although the flexible driving characteristics of pneumatic muscles give them a biomimetic compliance advantage, the low stiffness structure is susceptible to changes in air pressure step, multi-degree-of-freedom coupling interference, and external impacts, which can lead to low-frequency mechanical resonance and trajectory tracking overshoot. Traditional control methods (such as linear feedback) cannot dynamically adjust the air pressure response rate, resulting in phase lag and high-frequency chatter in the actuator output, which exacerbates the vibration amplitude of the flexible structure and may even cause mechanical fatigue damage. In addition, the delay characteristics and nonlinear coupling of pneumatic muscles further amplify the overshoot of trajectory tracking, which can easily cause target deviation or safety hazards in precision operations (such as grasping and interaction). Therefore, a control mechanism that balances dynamic suppression and tracking accuracy is urgently needed.

[0007] (3) Difficulty in modeling nonlinear dynamic characteristics. Due to the highly nonlinear and anisotropic physical characteristics of pneumatic muscles, traditional Lagrangian models cannot meet the accuracy requirements of modeling. For example, when describing the instantaneous pressure-deformation characteristics of pneumatic muscles, the Lagrangian model cannot accurately capture their instantaneous stress-strain response due to their hysteresis. At the same time, the significant differences in different directions caused by load and disturbances make the Lagrangian model highly asymmetric, making it impossible to control them through traditional modeling methods. In addition, the equivalent stiffness, damping and other parameters of pneumatic muscles drift in real time with air pressure fluctuations, material aging, and changes in external loads, resulting in decreased stability and deteriorated tracking accuracy of control strategies based on fixed parameter models (such as PID and sliding mode control) in practical applications. Existing methods rely on high-precision modeling or complex parameter identification, which is difficult to adapt to dynamic working conditions (such as sudden load changes and interactive force disturbances), seriously restricting the reliability of pneumatic muscles in high-precision bionic actuators (such as bionic arms and rehabilitation robots).

[0008] (4) Insufficient motion protection mechanisms. Pneumatic muscles inherently possess unidirectional constraints. Excessive contraction may cause excessive tension on the antagonist muscle, leading to damage or even instability of the mechanical system. In practical applications, excessive contraction can cause excessive gas pressure inside the muscle, resulting in material fatigue or even rupture, while excessive relaxation may reduce the muscle's elastic recovery performance, weakening motion control precision. For example, when bionic robotic arms are used for precise operational tasks (such as industrial assembly or medical surgery), excessive tension in the antagonist muscle may not only cause the actuator to exceed its control range but may also damage the operational target or threaten patient safety due to uncontrolled end-effector forces. To prevent such problems, effective motion protection mechanisms need to be designed, such as angle constraints and velocity constraints. Summary of the Invention

[0009] To address the technical problems mentioned above, this invention provides a method for correcting and controlling the obstacle avoidance trajectory of a pneumatic muscle multi-axis manipulator. Based on the fusion design of the Mercator projection algorithm and the DMP algorithm, an external force is introduced into the motion generation equation of the DMP algorithm to adjust the trajectory to meet obstacle avoidance requirements. This achieves efficient trajectory mapping and real-time obstacle avoidance adjustment of the bionic hand from joint space to a two-dimensional plane, solving the problems of lag and insufficient accuracy in traditional trajectory planning under complex environments.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] The first aspect of the present invention provides a method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis manipulator, comprising:

[0012] Acquire the target trajectory of a pneumatic muscle multi-axis manipulator;

[0013] For the target trajectory, the Mercator projection method is used to transform the target trajectory into a two-dimensional plane trajectory;

[0014] The relative position and shape of the obstacle are obtained, the external force is calculated, and the external force is introduced into the trajectory generation equation of the dynamic motion primitive algorithm. Based on the two-dimensional plane trajectory, the obstacle avoidance trajectory of the pneumatic muscle multi-axis manipulator is generated by the dynamic motion primitive algorithm, and the inverse transformation is performed by the Mercator projection method.

[0015] Furthermore, the trajectory generation equation after introducing external forces is:

[0016]

[0017]

[0018] in, x This is the current position of the pneumatic muscle multi-axis robotic arm. v It is the speed of a pneumatic muscle multi-axis robotic arm. It is the time scaling factor. K and D It refers to the elasticity and damping coefficients of a pneumatic muscle multi-axis robotic arm. g It is the target location. x 0 is the initial position. It is a nonlinear function. F obstacle It is an external force.

[0019] Furthermore, the external force is calculated using a repulsive force model.

[0020] Furthermore, the external force is expressed as:

[0021]

[0022] in, It is a constant, representing the intensity of the force; This is the current position of the pneumatic muscle multi-axis robotic arm; x nearest It is the point on the obstacle surface that is closest to the system; It is the decay index.

[0023] A second aspect of the present invention provides a control method for a pneumatic muscle multi-axis manipulator, comprising:

[0024] For the obstacle avoidance trajectory obtained by the obstacle avoidance trajectory correction method of the pneumatic muscle multi-axis manipulator as described in the first aspect, after trajectory discretization and inverse kinematics calculation, the target angle values ​​of each joint are obtained.

[0025] Based on the target angle values ​​of each joint, an adaptive controller is used to obtain control signals to control the movement of each joint.

[0026] Furthermore, the adaptive controller is constructed by combining a deep neural network and a backstepping control method.

[0027] A third aspect of the present invention provides a pneumatic muscle multi-axis robotic arm obstacle avoidance trajectory correction system, comprising:

[0028] The target acquisition module is configured to acquire the target trajectory of the pneumatic muscle multi-axis manipulator.

[0029] The projection module is configured to use the Mercator projection method to transform the target trajectory into a two-dimensional plane trajectory.

[0030] The obstacle avoidance module is configured to: acquire the relative position and shape of the obstacle, calculate the external force, introduce the external force into the trajectory generation equation of the dynamic motion primitive algorithm, generate the obstacle avoidance trajectory of the pneumatic muscle multi-axis manipulator based on the two-dimensional plane trajectory through the dynamic motion primitive algorithm, and perform inverse transformation through the Mercator projection method.

[0031] A fourth aspect of the present invention provides a control system for a pneumatic muscle multi-axis manipulator, comprising:

[0032] The trajectory discretization module is configured to: for the obstacle avoidance trajectory obtained by the obstacle avoidance trajectory correction system of the pneumatic muscle multi-axis manipulator described in the third aspect, perform trajectory discretization and inverse kinematics calculation to obtain the target angle values ​​of each joint;

[0033] The control module is configured to obtain control signals based on the target angle values ​​of each joint through an adaptive controller, so as to control the movement of each joint.

[0034] A fifth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis manipulator or a method for controlling a pneumatic muscle multi-axis manipulator.

[0035] A sixth aspect of the present invention provides a computer device including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method for correcting the obstacle avoidance trajectory of a pneumatic muscle multi-axis manipulator or a method for controlling a pneumatic muscle multi-axis manipulator.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] This invention is based on the fusion design of Mercator projection algorithm and DMP (Dynamic Motion Primitive) algorithm, and introduces external force into the motion generation equation of DMP algorithm to adjust the trajectory to meet obstacle avoidance requirements. It realizes efficient trajectory mapping and real-time obstacle avoidance adjustment of bionic hand from joint space to two-dimensional plane, solves the problem of lag and insufficient accuracy of traditional trajectory planning in complex environments. Moreover, through the dynamic trajectory planning mechanism, while ensuring the accuracy of target tracking error, it can complete obstacle avoidance response at different set speeds, significantly improving the operation efficiency and safety of bionic robotic arm in unstructured scenarios (such as multiple obstacles and irregular obstacles), and laying a solid foundation for the flexible application of robotic arm in complex environments.

[0038] This invention, through the deep integration of pneumatic muscle-based intelligent neural network control and dynamic obstacle avoidance algorithm, successfully constructs an autonomous environmental perception-decision-execution closed loop for a bionic robotic arm in single obstacle avoidance tasks and multi-obstacle avoidance task interaction scenarios. Based on obstacle information fed back by sensors in real time during the operation, it can dynamically plan the obstacle avoidance path, adjust the optimal controller parameters in real time under state constraints, and track and monitor errors. This not only improves the environmental adaptability of the bionic robotic arm, but also provides core technical support for its multi-objective operation needs in fields such as medical rehabilitation and hazardous operations, and significantly enhances the task collaborative control capability of the robotic arm.

[0039] The multi-dimensional trajectory planning and dynamic obstacle avoidance technology proposed in this invention, along with the significant improvement in multi-modal environment adaptation and task collaborative control capabilities, greatly enrich the application scope of bionic robotic arms and significantly enhance their operational flexibility. In fields such as medical rehabilitation where operational precision is extremely high, and in dangerous operational scenarios facing complex and ever-changing environmental challenges, this invention can provide solid technical support for bionic robotic arms, ensuring that they can efficiently and safely complete various tasks. This innovation not only significantly enhances the practical value and market competitiveness of robotic arms, but also injects strong new momentum into promoting the intelligent and automated development of related fields.

[0040] This invention proposes a dynamic damping injection technique, which introduces a virtual damping term based on real-time state into the backstepping framework to dynamically adjust the air pressure response rate of the pneumatic muscle and suppress low-frequency mechanical resonance. Nonlinear dynamic surface control (DSC) replaces the step-by-step differentiation operation of the traditional backstepping method with a low-pass filter, smoothing the virtual control signal, avoiding high-frequency noise amplification, and reducing phase lag. Neural network feedforward compensation uses a neural network to predict the delayed response of the pneumatic muscle, generating feedforward air pressure commands to compensate for overshoot caused by nonlinear coupling. This achieves the technical effects of suppressing resonance, dynamic adaptation, and controlling overshoot.

[0041] This invention proposes a neural network perturbation observer that treats the unmodeled dynamics (hysteresis, asymmetry) of pneumatic muscles as a lumped perturbation, which is approximated and compensated online by a neural network without relying on a precise mathematical model. An adaptive backstepping framework, through recursive design of adaptive laws, updates the neural network weights and parameter estimates in real time, tracking and compensating for time-varying factors such as air pressure fluctuations and material aging. This achieves the technical effects of model independence and parameter adaptive capability.

[0042] This invention proposes an adaptive adjustment of the air pressure safety threshold. Based on neural network prediction of muscle contraction limits, it dynamically adjusts the upper limit of air pressure commands to avoid material rupture or elastic failure. Fault detection and fault-tolerant control are also included: an integrated state observer monitors muscle pressure and deformation in real time, and an integrated fault observer ensures the triggering of emergency decompression or braking mechanisms, guaranteeing a safe shutdown before instability. This achieves the technical effects of safety protection and precise dynamic constraint. Attached Figure Description

[0043] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0044] Figure 1 This is a flowchart of an obstacle avoidance trajectory correction method for a pneumatic muscle multi-axis manipulator according to Embodiment 1 of the present invention;

[0045] Figure 2 This is a schematic diagram of the target trajectory conversion in Embodiment 1 of the present invention;

[0046] Figure 3 This is a schematic diagram of obstacle avoidance trajectory planning in a two-dimensional plane according to Embodiment 1 of the present invention;

[0047] Figure 4 This is a schematic diagram of a three-dimensional spherical obstacle avoidance trajectory planning according to Embodiment 1 of the present invention;

[0048] Figure 5 This is a control flowchart of Embodiment 2 of the present invention;

[0049] Figure 6 This is a schematic diagram of the training and simulation results of sinusoidal tracking according to Embodiment 2 of the present invention;

[0050] Figure 7 This is a simulation tracking result diagram of the obstacle avoidance trajectory in Embodiment 2 of the present invention;

[0051] Figure 8 This is an error variation curve diagram of Embodiment 2 of the present invention;

[0052] Figure 9 This is a schematic diagram of the structure of a computer device according to Embodiment Six of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0054] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0055] Example 1

[0056] This embodiment provides a method for correcting the obstacle avoidance trajectory of a pneumatic muscle multi-axis robotic arm.

[0057] In terms of trajectory planning, the goal of this embodiment is to design a high-quality path planning scheme suitable for pneumatic artificial muscle arms, while also being able to reflect obstacle avoidance planning for different types of obstacles. To achieve this research goal, this invention has explored and designed a restricted motion obstacle avoidance algorithm based on Dynamic Movement Primitives (DMP).

[0058] The theoretical basis of the obstacle avoidance trajectory correction method for a pneumatic muscle multi-axis manipulator provided in this embodiment is: DMP motion obstacle avoidance algorithm; Mercator projection method.

[0059] This embodiment provides a method for correcting the obstacle avoidance trajectory of a pneumatic muscle multi-axis manipulator. Based on the Mercator projection method, the target trajectory on the sphere in the joint space is transformed into a two-dimensional plane. Then, the obstacle avoidance algorithm of DMP is used to plan the trajectory in the plane, and then the transformation is reversed to output the joint angle, thereby realizing precise control of the bionic wrist joint and accurate obstacle avoidance of the fingertips.

[0060] This embodiment provides a method for correcting the obstacle avoidance trajectory of a pneumatic muscle multi-axis robotic arm, such as... Figure 1 As shown, it includes the following steps:

[0061] Step 1: Initial setup.

[0062] In order to solve the motion planning and obstacle avoidance problems of robotic arms using mathematical methods, this invention chooses joint space for modeling. Currently, the wrist of the bionic robotic arm is fixed, and only the up-down and left-right rotation of the wrist joint is considered, that is, two angular variables are considered.

[0063] Step 2: Select the target trajectory.

[0064] To provide sufficient complexity to verify the obstacle avoidance algorithm and trajectory learning capability of the bionic hand, this invention calculates the target trajectory (example trajectory or original trajectory) using mathematical functions, for example, generating a curved trajectory on a three-dimensional spherical shell using trigonometric functions.

[0065] The mathematical function is expressed as:

[0066]

[0067] in, q d (t) yes t The target joint angle vector (in radians) at any given moment; θ 1 (t),θ 2 (t) It refers to the left-right and up-down rotation angles of the wrist joint; A 1 ,A 2 represents the amplitude (typical value 0.5-1.5 rad); ω 1 ,ω 2 is the angular frequency (rad / s), which determines the speed of motion; and It is the phase shift (initial phase angle); θ 10 ,θ 20 It is the center position of the joint (default zero position).

[0068] Step 3: Project the target trajectory.

[0069] Inspired by methods for creating Earth maps, this invention employs the Mercator projection method to transform the trajectory of a target on a sphere into a two-dimensional planar trajectory, such as... Figure 2 As shown.

[0070] Mercator projection is a normal-axis conformal cylindrical projection method that can preserve angular relationships and is suitable for trajectory tracking and obstacle avoidance planning of three-dimensional spherical trajectories on a two-dimensional plane.

[0071] This projection simplifies the original spherical trajectory into a path in a two-dimensional plane, making trajectory planning and calculation much easier.

[0072] Step 4: Trajectory planning based on the DMP algorithm.

[0073] DMP is a trajectory generation and control method based on nonlinear dynamic systems, which can generate a smooth path between a given initial position and a target position.

[0074] In this embodiment, the DMP algorithm is used to process the planar trajectory after Mercator projection, while introducing external forces into the basic equations of motion to adjust the trajectory to meet obstacle avoidance requirements. This algorithm uses the basic framework of an elastic-damped system, combined with a learned nonlinear trajectory generation function and dynamic obstacle avoidance external forces, thereby achieving obstacle avoidance and flexible motion planning.

[0075] The motion generation equations of DMP mainly include the following forms: ;

[0076] ;

[0077] in, x It is the current position of the nonlinear dynamic system (i.e., the pneumatic muscle multi-axis manipulator). v It is the velocity of a nonlinear dynamic system. It is the time scaling factor. K and D These are the elasticity and damping coefficients of a nonlinear dynamic system. g It is the target location. x 0 is the initial position. It is a nonlinear function responsible for introducing the nonlinear behavior of the learned function, and it depends on the phase variable. s .

[0078] (2) Phase variables and nonlinear terms.

[0079] The phase variable s controls the progress of motion and is usually defined as follows: ;in, α s It is a positive time constant that determines the rate at which the phase variable s decays from its initial value (usually 1) to 0.

[0080] nonlinear functions It is usually obtained by a linear combination of Gaussian functions to approximate the shape of the training trajectory, i.e.:

[0081]

[0082] in, These are basis functions. It's weight. N The Gaussian function represents the function used to approximate the shape of the training trajectory. The total number.

[0083] (3) The introduction of obstacle avoidance mechanism.

[0084] In obstacle avoidance scenarios, external forces F obstacle The acceleration equations of the DMP are introduced to correct the original trajectory, allowing it to avoid obstacles, i.e.:

[0085]

[0086] (4) Calculation of external forces.

[0087] external force F obstacle Calculations are typically based on the relative position and shape of obstacles to ensure the trajectory deviates from the danger zone. For example, when an obstacle is near a predetermined trajectory, the external force calculated using a repulsion model is as follows:

[0088]

[0089] in, It is a constant, representing the intensity of the force; This is the system's current location; x nearest It is the point on the obstacle surface that is closest to the system; x obstacle It is the location of the obstacle; It is the decay exponent; Shape is the shape of the obstacle, and d(x,Shape) represents the shortest distance from the current position x of the system to the surface of the obstacle.

[0090] Based on the above calculations, taking an elliptical obstacle as an example, a reasonable obstacle avoidance trajectory is planned as follows: Figure 3 As shown.

[0091] Step 5: Inverse transformation of the coordinates after trajectory planning, mapping them back to the three-dimensional joint space to generate the target joint angle of the bionic robotic arm. Specifically, the two-dimensional trajectory optimized for obstacle avoidance is restored to the target trajectory on the sphere through the inverse transformation of Mercator projection, generating the motion path of the wrist joint; the target joint angle value is generated based on the projected trajectory to guide the movement of the pneumatic muscles.

[0092] like Figure 4 As shown, on the three-dimensional sphere, Trajectory1 is the target trajectory used for learning, and Trajectory2 is the actual trajectory after obstacle avoidance is planned.

[0093] To verify the effectiveness of the algorithm, preliminary simulations were conducted in this embodiment. By adjusting the dynamic parameters of the DMP, the motion effect of the bionic arm under obstacles of different shapes was observed. Preliminary results show that the Mercator projection method combined with the DMP obstacle avoidance algorithm can effectively guide the bionic arm to avoid obstacles, and the path generation is smooth and has high controllability.

[0094] This embodiment provides a method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis manipulator. Based on the fusion design of Mercator projection algorithm and DMP algorithm, external force is introduced into the motion generation equation of DMP algorithm to adjust the trajectory to meet obstacle avoidance requirements. This achieves efficient trajectory mapping and real-time obstacle avoidance adjustment of the bionic hand from joint space to two-dimensional plane, breaking through the problems of adjustment lag and insufficient accuracy of traditional trajectory planning in complex environments. Through dynamic trajectory planning mechanism, while ensuring accurate target tracking error, obstacle avoidance response can be completed at different set speeds, significantly improving the operation efficiency and safety of the bionic manipulator in unstructured scenarios (such as multiple obstacles and irregular obstacles).

[0095] Example 2

[0096] This embodiment provides a control method for a pneumatic muscle multi-axis robotic arm.

[0097] In order to build a stable controller for a typical nonlinear system such as a pneumatic artificial muscle arm, this embodiment proposes to combine a deep neural network (DNN) and a backstepping control method to build an intelligent controller, so as to achieve the purpose of tracking and controlling the trajectory of the Lagrange system.

[0098] Its core idea lies in using deep neural networks (DNNs) to approximate the Lagrangian function. Furthermore, different convolutional neural networks and feedforward convex neural networks (FCNNs) are used to design virtual control laws in the controller to compensate for time delay errors and external disturbances in the system, simulate the damping term of the controller, and combine them to obtain the control law. In this way, through real-time feedback data from the platform, the controller autonomously adjusts the parameters of the virtual control law and damping term during motion to achieve adaptive control, improve the robustness and stability of the nonlinear system, and effectively address the challenges posed by characteristics such as hysteresis and creep in pneumatic muscles to precise control.

[0099] This embodiment provides a control method for a pneumatic muscle multi-axis manipulator, based on the following theoretical foundations: Lagrange equations; fault observer; deep neural network, Lagrange neural network, and convex neural network.

[0100] This embodiment provides a control method for a pneumatic muscle multi-axis robotic arm, the design concept of which is as follows:

[0101] (1) Control system modeling and fault detection.

[0102] First, based on the physical structure, a dynamic model of the PAM-driven robot hand was established, taking into account the relationship between system failures and time. The model is as follows:

[0103]

[0104] in, It is the mass matrix of the fitted Lagrange model, which describes the inertial characteristics of the system; It is the Coriolis matrix, which describes the velocity-like properties generated by the system's motion; D is the gravity vector, describing the generalized force caused by potential energy; D is the damping force vector, including dissipative effects such as friction. It is a joint angle vector. It is the control input, and u is the system state, i.e., the control input. It is a parameter matrix. This indicates the relationship between system failure and time. It is a real-time fault function.

[0105] To facilitate stability analysis, the above model is rewritten as follows:

[0106] ;

[0107] .

[0108] in, This is the transformed fault function. Considering that the physical properties of pneumatic muscles are difficult to obtain due to their complex structure, a deep neural network is used to fit the Lagrange model. By acquiring the system state and velocity under natural conditions and under different degrees of freedom and different input air pressures, the data is processed for noise reduction and normalization to form a dataset for training, resulting in an offline controlled object model to solve the physical modeling difficulties of Challenge 3.

[0109] (2) Introduce a trajectory corrector based on Dynamic Motion Primitives (DMP).

[0110] In the controller front-end design, this invention introduces a trajectory corrector based on Dynamic Motion Primitives (DMP). Its core idea is to model the trajectory generation problem as a spring-damped system perturbed by external correction terms. DMP dynamically adjusts the endpoints and shape of the trajectory through nonlinear functions, achieving accurate correction of the reference trajectory. The specific form is as follows:

[0111]

[0112]

[0113] in, x It is the current position of the nonlinear dynamic system (i.e., the pneumatic muscle multi-axis manipulator). v It is the velocity of a nonlinear dynamic system. It is the time scaling factor. K and D These are the elasticity and damping coefficients of a nonlinear dynamic system. g It is the target location. x0 is the initial position. It is a nonlinear function responsible for introducing the nonlinear behavior of the learned function, and it depends on the phase variable. s .

[0114] (3) Design a fault observer.

[0115] This embodiment designs a fault observer to detect faults in the system in real time.

[0116] The dynamics of the fault observer are as follows:

[0117]

[0118] Where x is the actual state vector of the system (which can be directly measured). It controls the amount of input. It is the gain of the observer to be designed. It is a parameter matrix. yes The estimated value, It is a nonlinear state function that describes the dynamic model of the system (based on estimated state). and control input u).

[0119] (4) Development of an adaptive controller based on deep neural networks.

[0120] Therefore, combining the physical modeling, trajectory corrector, and fault observer previously developed based on Lagrange's theorem, this invention proposes an adaptive tracking controller based on a deep neural network (DNN) to simultaneously compensate for unknown dynamics / disturbances and adjust the optimal control gain. The adaptive controller design based on the backstepping method is as follows:

[0121]

[0122] in, It is the mass matrix of the fitted Lagrange model, which describes the inertial characteristics of the system; It is the Coriolis matrix, which describes the velocity-like properties generated by the system's motion; It is a gravity vector, describing the generalized force caused by potential energy; It is tracking error. It is a speed-related error signal.

[0123] In the design of the adaptive tracking controller, a Lagrange neural network is introduced to represent the unknown Lagrange dynamics model of the system, and is used to estimate the kinetic and potential energy of the entire system. That is, the fully connected layer Lagrange neural network is as follows:

[0124]

[0125] here, and These are the estimated values ​​of kinetic energy and potential energy, respectively. and Considering the high-dimensional and nonlinear nature of the model, this invention selects the Kriging model and the Surrogate-Based Optimization (SBO) algorithm to obtain the optimal parameters. and ,get

[0126]

[0127] Furthermore, to compensate for disturbances, the controller approximates the damping term, dynamically adjusts the optimal control gain, and introduces auxiliary functions composed of different FICNN networks. and two virtual control rates , This ensures both the convexity of the neural network terms and the effect of adaptive control.

[0128] For the design of neural networks in the controller, multiple fully connected layers such as LNN and FICNN and deep convolutional neural networks were constructed, along with an approximate Lagrange model, virtual control law, and auxiliary functions.

[0129] This invention constructs a neural network training dataset by collecting pneumatic muscle arm states under different inputs. A fully connected layer neural network is built to model the LNN model, and multiple deep convolutional neural networks are introduced to fit the virtual control law and damping matrix. The parameters are adaptively adjusted online, and the sinusoidal target trajectory is tracked during simulation. The results are as follows: Figure 6 As shown.

[0130] Combining the dynamic obstacle avoidance function in the input correction section, obstacle avoidance simulation is performed in a single obstacle environment in spherical coordinates. The planned trajectory and the actual tracking trajectory are as follows: Figure 7 As shown, the tracking error is recorded as follows: Figure 8 As shown.

[0131] In summary, this embodiment provides a control method for a pneumatic muscle multi-axis robotic arm, such as... Figure 5 As shown, the specific steps include:

[0132] Step 1, Trajectory Discretization: For the obstacle avoidance trajectory obtained by the obstacle avoidance trajectory correction method of the pneumatic muscle multi-axis manipulator provided in Example 1, the continuous trajectory is discretized into a series of discrete path points in the time domain, and the parameters such as the angle, position or speed of each joint of the manipulator at each path point are obtained. These parameters are used as the target values ​​for subsequent control.

[0133] Step 2, Inverse Kinematics Calculation: Based on the discretized path point coordinates, the desired end effector position and orientation are converted into target angle values ​​for each joint using the inverse kinematics algorithm of the robotic arm.

[0134] Step 3: Based on the target angle values ​​(obstacle avoidance trajectory) of each joint and the actual state of the system, obtain the error signal, input it into the fault observer, and output the fault quantification signal through analysis and calculation of the residual. This signal is used to monitor sensor deviation and actuator saturation, so as to view and quantify the fault of the system in real time.

[0135] Step 4: The fault information output by the fault observer, combined with the feedback signal and the filtering state output by the closed-loop filter, is recursively identified and subjected to Lyapunov stability analysis to obtain the output of the time-varying parameter estimation stage, which is the estimated dynamic parameter (time-varying parameter). The fault observer and the time-varying parameter in steps 3 and 4 work together to ensure the normal operation of the motion protection mechanism by monitoring the faults and parameter changes of the system during operation.

[0136] Step 5: Design and training of the estimation network for the controlled object: In order to obtain a suitable backstepping control law, a fully connected layer neural network is introduced to train and approximate the unknown Lagrangian dynamic model, which is used to estimate the kinetic and potential energy of the entire system; at the same time, an auxiliary function that is essentially a convolutional neural network is designed to be used as an equivalent observer of the external disturbance torque to estimate the damping caused by the external disturbance.

[0137] The Lagrange model estimation network is pre-trained offline using natural data from a pneumatic muscle multi-axis manipulator platform to obtain an offline fully connected layer Lagrange neural network (LNN). In practical applications, the LNN input is the feedback signal obtained through a closed-loop filter, i.e., the state and velocity signals, to obtain physical parameters such as the inertia matrix, Coriolis force matrix, and gravity matrix of the Lagrange estimation model.

[0138] The optimization and training of the external disturbance damping estimation network are synchronized with the actual movement of the pneumatic arm. The parameter initialization of this part of the network is completed offline, while the parameters of each level of the auxiliary function in the subsequent network are propagated forward as the system runs, continuously updated and optimized. The auxiliary function in the external disturbance damping estimation network, together with the virtual control law 1, constitutes the external disturbance damping compensation term in the subsequent controller input. In practical applications, the external disturbance damping estimation network is fed with the input state and velocity signals, and the parameters of the auxiliary function are updated. These parameters are then combined with the virtual control law 1 to calculate the external disturbance damping compensation term.

[0139] Step 6: Controller Design and Implementation of Intelligent Control Law: Based on the physical parameters such as the inertia matrix, Coriolis force matrix, and gravity matrix obtained from the Lagrange neural network in Step 3, the external disturbance damping compensation term obtained from the auxiliary function and virtual control law 1, the time delay compensation term obtained from the output signal of the closed-loop filter and virtual control law 2, the smooth feedback signal output by the closed-loop filter, and the real-time state of the controlled object obtained from the sensor, the control signal u is calculated through an adaptive controller based on the backstepping method, which serves as the input to the controlled object. The fault quantization signal obtained by the fault observer and the output of the time-varying parameter estimation serve as a constraint mechanism to cut off the operation of the control system when necessary.

[0140] Step 7: Convert the control signal u output by the controller into an appropriate drive signal to drive the movement of each joint of the pneumatic muscle multi-axis robotic arm (multi-muscle release soft robotic arm).

[0141] Step 8, Feedback and Correction: During the movement of the robotic arm, sensors (such as encoders, force sensors, and vision sensors) are used to acquire the actual position, attitude, and force information of the robotic arm in real time, which serves as feedback signals. These are compared with the desired trajectory (obstacle avoidance trajectory), and an error signal is obtained through the feedback path. The error signal is tracked, and the error signal and feedback signal are input into a closed-loop filter to obtain a smoothed and denoised feedback signal. This feedback signal is then fed back into the Lagrange model estimation network and the external disturbance damping estimation network to continuously update and adjust the accurate state signals of the actuators in the system, as well as the virtual control law 1 and virtual control law 2 in the adaptive controller, to adjust the control signals in the next cycle and correct the movement of the robotic arm in real time, so that the steady-state error eventually converges.

[0142] Example 3

[0143] This embodiment provides a pneumatic muscle multi-axis robotic arm obstacle avoidance trajectory correction system, which specifically includes:

[0144] The target acquisition module is configured to acquire the target trajectory of the pneumatic muscle multi-axis manipulator.

[0145] The projection module is configured to use the Mercator projection method to transform the target trajectory into a two-dimensional plane trajectory.

[0146] The obstacle avoidance module is configured to: acquire the relative position and shape of the obstacle, calculate the external force, introduce the external force into the trajectory generation equation of the dynamic motion primitive algorithm, generate the obstacle avoidance trajectory of the pneumatic muscle multi-axis manipulator based on the two-dimensional plane trajectory through the dynamic motion primitive algorithm, and perform inverse transformation through the Mercator projection method.

[0147] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0148] Example 4

[0149] This embodiment provides a pneumatic muscle multi-axis manipulator control system, which specifically includes:

[0150] The trajectory discretization module is configured to: for the obstacle avoidance trajectory obtained by the obstacle avoidance trajectory correction system of the pneumatic muscle multi-axis manipulator described in Embodiment 3, perform trajectory discretization and inverse kinematics calculation to obtain the target angle values ​​of each joint;

[0151] The control module is configured to obtain control signals based on the target angle values ​​of each joint through an adaptive controller, so as to control the movement of each joint.

[0152] It should be noted that each module in this embodiment corresponds one-to-one with each step in embodiment two, and their specific implementation process is the same, so it will not be repeated here.

[0153] Example 5

[0154] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the obstacle avoidance trajectory correction method for a pneumatic muscle multi-axis manipulator as described in Embodiment 1 or the control method for a pneumatic muscle multi-axis manipulator as described in Embodiment 2.

[0155] Example 6

[0156] This embodiment provides a computer device, such as... Figure 9 As shown, the device includes a display device, an input device, a computer-readable storage medium (volatile memory and non-volatile storage medium), a processor, a communication interface (i.e., a network interface), and a computer program stored on the computer-readable storage medium and executable on the processor. The processor, communication interface, and computer-readable storage medium can be connected via a bus or other means. The communication interface is used to receive and send data. When the processor executes the program, it implements the steps in the obstacle avoidance trajectory correction method for a pneumatic muscle multi-axis manipulator as described in Embodiment 1 or the control method for a pneumatic muscle multi-axis manipulator as described in Embodiment 2.

[0157] Any references to memory, storage, database, or other media used in this application and embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0158] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0159] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0161] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for correcting the obstacle avoidance trajectory of a pneumatic muscle multi-axis robotic arm, characterized in that, include: Acquire the target trajectory of a pneumatic muscle multi-axis manipulator; For the target trajectory, the Mercator projection method is used to transform the target trajectory into a two-dimensional plane trajectory; Obtain the relative position and shape of the obstacle, calculate the external force, and represent the external force as follows: in, It is a constant, representing the intensity of the force; This is the current position of the pneumatic muscle multi-axis robotic arm; x nearest It is the point on the obstacle surface that is closest to the system; It is the decay index; After introducing external force into the trajectory generation equation of the dynamic motion primitive algorithm, the obstacle avoidance trajectory of the pneumatic muscle multi-axis manipulator is generated based on the two-dimensional plane trajectory through the dynamic motion primitive algorithm, and then inversely transformed by the Mercator projection method. After discretizing the obtained obstacle avoidance trajectory and performing inverse kinematics calculation, the target angle values ​​of each joint are obtained. Based on the target angle values ​​of each joint, an adaptive controller is used to obtain control signals to control the movement of each joint. The adaptive controller is constructed by combining a deep neural network and a backstepping control method.

2. The method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis robotic arm as described in claim 1, characterized in that, The trajectory generation equation after introducing external forces is: in, x This is the current position of the pneumatic muscle multi-axis robotic arm. v It is the speed of a pneumatic muscle multi-axis robotic arm. It is the time scaling factor. K and D It refers to the elasticity and damping coefficients of a pneumatic muscle multi-axis robotic arm. g It is the target location. x 0 is the initial position. It is a nonlinear function. F obstacle It is an external force.

3. The method for obstacle avoidance trajectory correction of a pneumatic muscle multi-axis robotic arm as described in claim 1, characterized in that, The external force was calculated using a repulsion model.

4. A pneumatic muscle multi-axis robotic arm obstacle avoidance trajectory correction system, characterized in that, include: The target acquisition module is configured to acquire the target trajectory of the pneumatic muscle multi-axis manipulator. The projection module is configured to use the Mercator projection method to transform the target trajectory into a two-dimensional plane trajectory. The obstacle avoidance module is configured to: acquire the relative position and shape of obstacles, calculate the external force, and represent the external force as follows: in, It is a constant, representing the intensity of the force; This is the current position of the pneumatic muscle multi-axis robotic arm; x nearest It is the point on the obstacle surface that is closest to the system; It is the decay index; After introducing external force into the trajectory generation equation of the dynamic motion primitive algorithm, the obstacle avoidance trajectory of the pneumatic muscle multi-axis manipulator is generated based on the two-dimensional plane trajectory through the dynamic motion primitive algorithm, and then inversely transformed by the Mercator projection method. The trajectory discretization module is configured to: discretize the obtained obstacle avoidance trajectory and perform inverse kinematics calculation to obtain the target angle values ​​of each joint; the control module is configured to: obtain control signals based on the target angle values ​​of each joint through an adaptive controller to control the movement of each joint; wherein, the adaptive controller is built by combining a deep neural network and a backstepping control method.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the obstacle avoidance trajectory correction method for a pneumatic muscle multi-axis manipulator as described in any one of claims 1-3.

6. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the obstacle avoidance trajectory correction method for a pneumatic muscle multi-axis manipulator as described in any one of claims 1-3.

Images