A fault-tolerant control method and system for redundant robot joint deadlock failure

Abstract

This invention discloses a fault-tolerant control method and system for redundant robot joint deadlock faults, relating to the field of robot control technology. First, a nonlinear optimization objective function incorporating hard constraints on the faulty joint is constructed. The search space of the Jacobian matrix is ​​reconstructed in real time on a reduced-dimensional manifold defined by the joint deadlock. Redundant degrees of freedom are used to compensate for the failed joint, driving the remaining healthy joints to avoid singular configurations and confined regions by maximizing the operability index. Subsequently, a two-layer control architecture of "inverse dynamics feedforward perturbation observer compensation" is adopted, using a torque residual filtering algorithm to offset the static friction and residual torque of the deadlocked joint in real time. Finally, composite control commands are generated concurrently. This invention effectively suppresses dynamic abrupt changes caused by joint faults, significantly improving the operational accuracy and operational stability of redundant robots under extreme conditions.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to a fault-tolerant control method and system for redundant robot joint deadlock faults. Background Technology

[0002] With the rapid development of human-machine collaboration technology and intelligent manufacturing capabilities, redundant robotic arms with seven degrees of freedom or higher are increasingly widely used in high-value technology fields. Seven-DOF redundant robotic arms, with their extra degrees of freedom exceeding the dimension of the task space, can flexibly adjust their attitude to achieve functions such as local obstacle avoidance, attitude optimization, and force or visual sensor attitude alignment while ensuring the execution of the end-effector task. Therefore, they have significant application value in scenarios such as spacecraft on-orbit servicing, minimally invasive surgery, precision assembly, and collaborative robots. Redundancy not only improves the feasibility and spatial accessibility of trajectory planning, but also enhances operability and reduces singularity risks through online redundancy allocation, thereby strengthening the system's adaptability to complex environments and task constraints.

[0003] Despite the flexibility and robustness advantages brought by redundant structures, joint-level hardware failures still frequently threaten the continuous operation of systems in actual engineering operations. Typical failure modes include brake malfunction, drive motor stalling, reducer jamming, or encoder step loss. These failures can cause a single joint to "delock"—that is, the joint cannot generate the expected relative motion and loses its degrees of freedom at a fixed angle. Joint deadlock causes abrupt changes in the system's degrees of freedom, discontinuous changes in the kinematic constraint set, and subsequent degradation and rank deficiency problems in the Jacobian matrix column. At the same time, the friction, jamming, and unmodeled static or dynamic torques caused by deadlocked joints introduce significant disturbances and uncertainties at the dynamic level.

[0004] Existing fault-tolerant control methods for redundant robot joint deadlock, such as the patent application titled "A Fault-Tolerant Motion Planning Method for a Redundant Robotic Arm" (publication number CN102126219B), mostly employ a pure kinematic fault-tolerant analytical scheme based on quadratic programming. Because they do not consider the abrupt change in dynamic state during deadlock and do not model frictional torques, they are prone to dynamic shocks during joint deadlock. To address this, the patent application titled "A Multi-Fault Fault-Tolerant Control Method for a Modular Robotic Arm Based on Dynamic Programming" (publication number CN110170992B) introduces a dynamic compensation mechanism. It estimates and compensates for system disturbances online using an adaptive fault observer and a dynamic programming algorithm. However, this scheme fails to coordinate the deadlock hardware constraints with the reduced-dimensional kinematic manifold space. When eliminating disturbances, it cannot utilize redundancy to guide healthy joints to actively avoid singular configurations and confined regions, resulting in low operational accuracy under confined manifolds and a tendency to get trapped in singularities, leading to control failure.

[0005] Therefore, there is an urgent need for a fault-tolerant control scheme that can formally model deadlock constraints at the kinematic level and reconstruct the Jacobian mapping to retain the value of the remaining degrees of freedom, and can also estimate and compensate for unmodeled disturbances caused by deadlock in real time at the dynamic level. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, the present invention aims to provide a fault-tolerant control method and system for redundant robot joint deadlock faults. Through a multi-level control strategy of spatial dimensionality reduction and reconstruction, redundancy collaborative optimization, and dynamic residual compensation, the method improves the stability of robot task execution and end-effector tracking accuracy by eliminating nonlinear interference of deadlocked joints, while ensuring the continuity of motion after the degradation of the mechanism's degrees of freedom. This meets the requirements of aerospace, medical and high-precision industries for continuity, safety and accuracy.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A fault-tolerant control method for redundant robot joint deadlock faults is proposed. The method reconstructs the kinematic mapping of the robot's Jacobian matrix on the dimension-reduced constrained manifold defined by the deadlocked joints, and redundancy is allocated to the remaining healthy joints based on the reconstructed Jacobian matrix and the operability index. A control architecture including inverse dynamics feedforward and disturbance compensation is adopted to generate and issue composite control commands to offset the dynamic mutations caused by the deadlocked joints, so that the robot can maintain the end-effector task execution capability under deadlock faults.

[0008] A fault-tolerant control method for redundant robot joint deadlock faults includes the following steps: 1) Redundant robotic arm kinematic model and fault constraint definition: For a redundant robotic arm with A redundant robotic arm with one degree of freedom, whose end effector is positioned and oriented in the workspace. With joint spatial position vector Satisfying nonlinear mapping relationship Ask about time The derivative of the equation yields the differential equation. ,in The Jacobian matrix of the system; When the first When a deadlock occurs in a joint, the first joint will... Joint variables of each joint Fixed at deadlock angle Define its constraints as follows: Based on the original Jacobian matrix Remove the first Column-wise parameterization yields the reconstructed Jacobian matrix. ; 2) Fault-tolerant inverse kinematics solution based on operability optimization: Construct a multi-objective nonlinear optimization function as shown in formula (2). : in, For the target location, is a constant and represents the weighting coefficients of each sub-item in the objective function. This is the end-point positioning error term, used to ensure that the end effector can accurately track the target trajectory; The joint limitation penalty item is defined as follows: , For the first The midpoint of each joint and This is a joint limiting constant; As an operability indicator, it is defined as follows: , used to characterize the movement flexibility of a robotic arm in a fault state; During the solution process, the search domain is... Limited to (in This enables adaptive fault reconstruction at the kinematic level. 3) Dynamics Modeling and Inverse Dynamics Control Law: The total dynamics equations of the robotic arm considering fault disturbances are as follows: Design the inverse dynamics feedforward control law : in, The inertia matrix, The matrix represents the centripetal force and the Coriolis force. The vector of the gravity term. Output torque command to the controller. For the unmodeled disturbance torque generated by the deadlock joint, Let the joint tracking error vector be... and These are the positive definite proportional and differential gain matrices, respectively; 4) Derivation of Perturbation Observer (DOB) Residual Compensation: Introducing a perturbation observer based on residual filtering, defining... Torque residual signal at time 1 : in, This refers to the composite torque command actually issued at the previous sampling time. The inverse dynamic torque calculated using the nominal model at the current moment; The residual is filtered by a first-order low-pass digital filter to obtain the disturbance estimate: Among them, the filter coefficients With system cutoff frequency and sampling period The relationship satisfies: Ultimately, the composite control law command that integrates disturbance compensation is... Represented as: in To compensate for the constant gain matrix, the control law applies the observed disturbance... Feedback to the input terminal enables closed-loop suppression of nonlinear errors caused by deadlock faults.

[0009] A fault-tolerant control system for redundant robot joint deadlock faults, implementing the aforementioned fault-tolerant control method for redundant robot joint deadlock faults.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constrains deadlock and reconstructs the Jacobian at the kinematic level, and combines redundancy optimization based on operability to maintain obstacle avoidance and end-point localization performance; at the dynamic level, it adopts a residual filter-type disturbance observer and uses the estimated disturbance as compensation, which significantly improves the tracking accuracy and robustness under deadlock faults, while taking into account real-time performance and ease of implementation. Attached Figure Description

[0011] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of robot kinematic reconstruction under fault conditions according to an embodiment of the present invention.

[0013] Figure 3 This is a block diagram illustrating the residual filtering principle based on the Disturbance Observer (DOB) in an embodiment of the present invention.

[0014] Figure 4 This is a comparison diagram of 3D trajectory tracking of the end effector in an embodiment of the present invention.

[0015] Figure 5 This is a curve showing the joint position tracking error in an embodiment of the present invention. Detailed Implementation

[0016] The method of the present invention will be further described below with reference to the embodiments and accompanying drawings.

[0017] A fault-tolerant control method for redundant robot joint deadlock faults, when the first joint deadlock fault occurs... The joint is locked at In this process, the original kinematic model is dimensionality reduced, the Jacobian matrix is ​​reconstructed, and end-effector velocity mapping is achieved on the constrained manifold. A multi-objective optimization function is constructed, which includes end-effector error, joint limit penalty, and operability index. The optimal joint parameters are solved within the neighborhood of the deadlocked joint, and fault-tolerant inverse kinematics calculation is achieved. Based on the complete dynamic model, an inverse dynamic feedforward control law is designed, and a disturbance observer is constructed through the torque residual. The unmodeled disturbance torque is estimated by low-pass filtering and proportionally compensated to the control input to form a composite control law, which improves the stability and closed-loop control performance under joint deadlock conditions.

[0018] Reference Figure 1 The aforementioned fault-tolerant control method for redundant robot joint deadlock faults includes the following steps: 1) Redundant robotic arm kinematic model and fault constraint definition: For a redundant robotic arm with A redundant robotic arm with one degree of freedom, whose end effector is positioned and oriented in the workspace. With joint spatial position vector Satisfying nonlinear mapping relationship Ask about time The derivative of the equation yields the differential equation. ,in The Jacobian matrix of the system; When the first When a deadlock occurs in a joint, the first joint will... Joint variables of each joint Fixed at deadlock angle Since this joint cannot produce relative motion, its constraint condition is defined as follows: Based on the original Jacobian matrix Remove the first Column-wise parameterization yields the reconstructed Jacobian matrix. ; 2) Fault-tolerant inverse kinematics solution based on operability optimization: In order to find the optimal obstacle avoidance and flexible posture by utilizing redundancy characteristics while reducing one degree of freedom, a multi-objective nonlinear optimization function as shown in formula (2) is constructed. : in, For the target location, is a constant and represents the weighting coefficients of each sub-item in the objective function. This is the end-point positioning error term, used to ensure that the end effector can accurately track the target trajectory; The joint limitation penalty item is defined as follows: , For the first The midpoint of each joint and This is a joint limiting constant; As an operability indicator, it is defined as follows: , used to characterize the movement flexibility of a robotic arm in a fault state; During the solution process, the search domain is... Limited to (in This enables adaptive fault reconstruction at the kinematic level, with the reconstructed kinematic architecture as follows: Figure 2 As shown; 3) Dynamic Modeling and Inverse Dynamic Control Law: Based on the Full Dynamic Equations of the Robotic Arm Considering Fault Disturbances Design the inverse dynamics feedforward control law : in, The inertia matrix, The matrix represents the centripetal force and the Coriolis force. The vector of the gravity term. Output torque command to the controller. For the unmodeled disturbance torque generated by the deadlock joint, Let the joint tracking error vector be... and These are the positive definite proportional and differential gain matrices, respectively; 4) Derivation of Disturbance Observer (DOB) Residual Compensation: Due to the significant static friction and unmodeled dynamic characteristics of the deadlocked joint, a disturbance observer based on residual filtering is introduced, defined as follows: Torque residual signal at time 1 : in, This refers to the composite torque command actually issued at the previous sampling time. The inverse dynamic torque calculated using the nominal model at the current moment; The residual is filtered by a first-order low-pass digital filter to obtain the disturbance estimate: Among them, the filter coefficients With system cutoff frequency and sampling period The relationship satisfies: Ultimately, the composite control law command that integrates disturbance compensation is... Represented as: in To compensate for the constant gain matrix, the control law applies the observed disturbance... Feedback to the input terminal achieves closed-loop suppression of nonlinear errors caused by deadlock faults. The residual filtering principle is as follows: Figure 3 As shown.

[0019] After obtaining the reconstructed kinematic commands and dynamic compensation torques, the robotic arm can be driven to complete the end-effector task. The 3D trajectory tracking comparison curves of the end-effector calculated based on the composite control commands are shown below. Figure 4 The results show that in three-dimensional spatial operations, the actual trajectory (solid line) and the desired trajectory (dashed line) highly overlap. This result further verifies that even under constrained manifolds caused by the degradation of degrees of freedom, the present invention still possesses extremely high spatial positioning and operational accuracy, without exhibiting trajectory deviation or mission interruption phenomena common in traditional methods.

[0020] Furthermore, by combining the dynamic residual compensation strategy with kinematic planning, this embodiment extracts and summarizes the position tracking error curves of the remaining six healthy joints within the entire continuous working cycle (0-10s) after the deadlock failure. Figure 5 As shown in the figure, the tracking error of each healthy joint exhibits a smooth sinusoidal convergence pattern. The errors of all healthy joints during operation are strictly limited to an extremely small range (the maximum error peak is effectively controlled within a certain range). to (between). The various error indicators of the scheme in this embodiment converge significantly, enabling the robotic arm to operate with high precision even in the event of a self-locking failure.

Claims

1. A fault-tolerant control method for redundant robot joint deadlock faults, characterized in that: The kinematic mapping of the robot's Jacobian matrix is ​​reconstructed on the dimension-reduced constrained manifold defined by the deadlocked joints, and the remaining healthy joints are redundantly allocated based on the reconstructed Jacobian matrix and the operability index. A control architecture including inverse dynamics feedforward and disturbance compensation is adopted to generate and issue composite control commands to counteract the dynamic abrupt changes caused by the deadlocked joints, so that the robot can maintain the end-effector task execution capability under deadlock failure.

2. The fault-tolerant control method for redundant robot joint deadlock faults according to claim 1, characterized in that, Includes the following steps: 1) Redundant robotic arm kinematic model and fault constraint definition: For a redundant robotic arm with A redundant robotic arm with one degree of freedom, whose end effector is positioned and oriented in the workspace. With joint spatial position vector Satisfying nonlinear mapping relationship Ask about time The derivative of the equation yields the differential equation. ,in The Jacobian matrix of the system; When the first When a deadlock occurs in a joint, the first joint will... Joint variables of each joint Fixed at deadlock angle Define its constraints as follows: Based on the original Jacobian matrix Remove the first Column-wise parameterization yields the reconstructed Jacobian matrix. ; 2) Fault-tolerant inverse kinematics solution based on operability optimization: Construct a multi-objective nonlinear optimization function as shown in formula (2). : in, For the target location, is a constant and represents the weighting coefficients of each sub-item in the objective function. This is the end-point positioning error term, used to ensure that the end effector can accurately track the target trajectory; The joint limitation penalty item is defined as follows: , For the first The midpoint of each joint and This is a joint limiting constant; As an operability indicator, it is defined as follows: , used to characterize the movement flexibility of a robotic arm in a fault state; During the solution process, the search domain is... Limited to (in This enables adaptive fault reconstruction at the kinematic level. 3) Dynamics Modeling and Inverse Dynamics Control Law: The total dynamics equations of the robotic arm considering fault disturbances are as follows: Design the inverse dynamics feedforward control law : in, The inertia matrix, The matrix represents the centripetal force and the Coriolis force. The vector of the gravity term. Output torque command to the controller. For the unmodeled disturbance torque generated by the deadlock joint, Let the joint tracking error vector be... and These are the positive definite proportional and differential gain matrices, respectively; 4) Derivation of Perturbation Observer (DOB) Residual Compensation: Introducing a perturbation observer based on residual filtering, defining... Torque residual signal at time 1 : in, This refers to the composite torque command actually issued at the previous sampling time. The inverse dynamic torque calculated using the nominal model at the current moment; The residual is filtered by a first-order low-pass digital filter to obtain the disturbance estimate: Among them, the filter coefficients With system cutoff frequency and sampling period The relationship satisfies: Ultimately, the composite control law command that integrates disturbance compensation is... Represented as: in To compensate for the constant gain matrix, the control law applies the observed disturbance... Feedback to the input terminal enables closed-loop suppression of nonlinear errors caused by deadlock faults.

3. A fault-tolerant control system for redundant robot joint deadlock faults, characterized in that: Implement the fault-tolerant control method for redundant robot joint deadlock faults as described in any one of claims 1-2.

Images