Event-Sampling Neurodynamic Fault-Tolerant Design Method for Distributed Multi-Robot Synchronous Motion
By constructing an event sampling neurodynamic fault-tolerant design method, the problem of synchronization and computational load caused by joint performance degradation in distributed multi-manipulator systems is solved. This method achieves high-precision collaborative synchronization and reduces redundant computation, making it suitable for industrial collaboration and intelligent robot operation.
Patent Information
- Application Number
- CN202611170287.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-04
- Publication Date
- 2026-11-10
AI Technical Summary
In existing distributed multi-manipulator systems, the problems of synchronous cooperation fault tolerance and controller computational load caused by joint performance degradation have not been effectively solved. In particular, when joints wear, deform, and output capacity decrease, it is easy to cause end-effector trajectory errors and clamping failure or collision risks.
An event-sampling neurodynamic fault-tolerant design method is adopted. By constructing a kinematic model of the robotic arm, an undirected communication topology, and a Laplace matrix, combined with a joint fault model and physical constraints, the optimal joint velocity of the multi-robot system is solved online using a recurrent neural network to achieve synchronized cooperative motion. An event sampling mechanism is also introduced to reduce redundant computation.
Under joint failure, it achieves high-precision collaborative synchronization and trajectory tracking of multi-robot systems, reduces the computational overhead of local controllers, and improves fault tolerance, autonomy, and hardware security. It is suitable for industrial collaborative assembly, intelligent robot operation, and cluster experimental platforms.
Smart Images

Figure CN123067341A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot control technology, specifically to an event sampling neurodynamics fault-tolerant design method for distributed multi-manipulator synchronous motion. Background Technology
[0002] As the scale and precision of tasks increase, a single robotic arm can hardly complete high-load, highly collaborative tasks independently. Distributed multi-robotic arm systems have become the core technology route of intelligent manufacturing. In collaborative tasks, multiple robotic arms are required to maintain their end-effector positions and synchronously track trajectories under communication topology constraints.
[0003] Robotic arm joints are prone to wear, deformation, and reduced output capacity under long-term cyclic operation or complex loads. If joint performance degrades and is still controlled according to an ideal model, it will not only cause end-effector trajectory and synchronization errors but may also lead to gripping failure or collision risks. Existing fault-tolerant methods are mostly geared towards single-machine or complete lock-up failures, and do not adequately consider the performance degradation caused by varying degrees of speed loss in joints in multi-machine systems. Therefore, solving optimization problems with multi-dimensional physical constraints online under joint degradation conditions to achieve reliable and safe distributed cooperative motion generation is a key challenge in the field of robotics.
[0004] Recurrent Neural Networks (RNNs), as an efficient neurodynamic model, have become the mainstream solution for solving the aforementioned constraint optimizations in real time online due to their parallel computing characteristics. However, existing RNNs typically employ continuous or fixed high-frequency time-triggered mechanisms, which result in dense, repeated iterations even when the system tends to stabilize or the state changes are minimal, leading to significant redundant computational overhead for the robot's local controller. Summary of the Invention
[0005] To overcome the shortcomings of existing technologies in addressing joint performance degradation, fault tolerance in synchronous collaboration, and insufficient consideration of controller computational load, this invention provides an event-sampling neurodynamic fault-tolerant design method for distributed multi-manipulator synchronous motion. This invention unifies joint fault modeling, leader-follower collaboration constraints, physical limitations, and quadratic programming optimization. It solves the problem online through a distributed event-triggered recurrent neural network, effectively reducing the redundant computational overhead of the system while ensuring high-precision multi-manipulator collaboration under joint degradation or lockup conditions.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] This invention provides an event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion, comprising the following steps:
[0008] Construct the kinematic model of the robotic arm, and construct the undirected communication topology and Laplace matrix between the robotic arms;
[0009] A joint performance loss fault model is constructed to characterize normal joint output, multiplicative performance degradation, additive faults, and joint locking faults.
[0010] In a multi-robot collaborative task, end-effector trajectory tracking constraints are constructed based on the desired end-effector trajectory of the master robot arm, and position coupling constraints and velocity coupling constraints between different robot arm end-effectors are constructed based on the communication topology.
[0011] Based on the upper limit of the joint angle, the lower limit of the joint angle, the upper limit of the joint speed, and the lower limit of the joint speed of the robotic arm, construct a speed-level physical constraint inequality.
[0012] Taking the minimum joint velocity norm of the robotic arm system as the optimization objective, the end-effector trajectory tracking constraint, position coupling constraint, velocity coupling constraint and velocity level physical constraint are unified into a synchronous cooperative motion generation problem in the form of quadratic programming.
[0013] Based on nonlinear complementary functions, the problem of synchronous cooperative motion generation in the form of quadratic programming is transformed into a problem of finding the zero of error function. An event sampling distributed neurodynamic fault-tolerant model is constructed and the optimal joint velocity of the multi-manipulator system is solved.
[0014] The optimal joint speed is sent to the joint execution units of each robotic arm, enabling the multi-robotic arm system with joint performance degradation or joint lockup to perform synchronized cooperative motion.
[0015] As a preferred technical solution, a kinematic model of the robotic arm is constructed, represented as follows:
[0016] ;
[0017] in, Indicates the first The joint state variables of the robotic arm, Let f(x) represent a smooth mapping function from joint space to Cartesian space. Indicates the first The motion trajectory of the end effector of the robotic arm in the world coordinate system. This indicates the number of robotic arms.
[0018] As a preferred technical solution, an undirected communication topology and Laplace matrix between robotic arms are constructed, specifically including:
[0019] Undirected communication topology is represented as:
[0020] ;
[0021] in, This represents a set of nodes, where each node corresponds to a robotic arm. This represents a set of edges used to indicate the communication relationships between robotic arms. Represents the adjacency matrix;
[0022] The Laplace matrix is expressed as:
[0023] ;
[0024] in, Representation degree matrix.
[0025] As a preferred technical solution, a joint performance loss fault model is constructed, and the actual joint angle output is represented as follows:
[0026] ;
[0027] in, For the first Taiwan robotic arm The actual joint angle output of each joint. To achieve the desired joint angle output, This represents the failover function. For joint effective output rate, For additive faults occurring in the joint;
[0028] Output of actual joint angles The corresponding fault joint velocity is obtained by taking the derivative.
[0029] As a preferred technical solution, the fault switching function is expressed as follows:
[0030] ;
[0031] in, The moment when the joint failure occurs. The time when the task execution terminates, before the failure occurs. The actual output equals the expected output after a fault occurs. The joint output is affected by the effective output rate and fault items;
[0032] when and At this time, the joint operates normally;
[0033] when and At this time, the joint's multiplicative properties deteriorate;
[0034] when and At that time, the joint experiences additive failure;
[0035] when and ,or and At that time, the joint locked up.
[0036] As a preferred technical solution, in a multi-robot collaborative task, an end-effector trajectory tracking constraint is constructed based on the desired end-effector trajectory of the main robot arm, expressed as:
[0037] ;
[0038] in, For the desired terminal trajectory, Indicates the first The robotic arm is the main robotic arm. Indicates the first The robotic arm is a follower robotic arm. Indicates the first The joint state variables of the robotic arm, Represents a smooth mapping function from joint space to Cartesian space;
[0039] Set end position error The error convergence relationship is set as follows: , As the convergence rate adjustment parameter, combined with the robotic arm's Jacobian matrix, the velocity-level trajectory tracking constraint can be obtained as follows:
[0040] ;
[0041] in, For the first Jacobian matrix of the robotic arm For the first The joint speed of the robotic arm, Indicates the first The desired end effector speed of the robotic arm.
[0042] As a preferred technical solution, the position coupling constraint is expressed as follows:
[0043] ;
[0044] in, For the first Taiwan robotic arm and the first The expected relative relationship between the end effectors of the robotic arm. Represents the elements of the adjacency matrix. , They represent the first Taiwan robotic arm, the first The desired end effector trajectory of the robotic arm;
[0045] The velocity coupling constraint is expressed as:
[0046] ;
[0047] in, In order to be with the first A set of robotic arms that have a communication relationship.
[0048] As a preferred technical solution, a velocity-level physical constraint inequality is constructed based on the upper limit of the joint angle, the lower limit of the joint angle, the upper limit of the joint speed, and the lower limit of the joint speed of the robotic arm, specifically expressed as follows:
[0049] ;
[0050] ;
[0051] ;
[0052] ;
[0053] ;
[0054] in, and These are the lower and upper limits of the joint angle, respectively. and These are the lower and upper limits of joint velocity, respectively. and These are the lower limit and upper limit of the speed level, respectively.
[0055] As a preferred technical solution, the synchronous cooperative motion generation problem in the form of quadratic programming is expressed as follows:
[0056] ;
[0057] ;
[0058] ;
[0059] ;
[0060] in, This indicates the combined joint velocities of a multi-robot robotic arm system. For convergence parameters, It is the identity matrix. The matrix form representing the physical limits of the velocity level. Represents a relative position matrix. Represents the Jacobian matrix. Indicates the combined joint positions of a multi-robot robotic arm system. Represents a smooth mapping function.
[0061] , , , The constraints consist of end-point trajectory tracking, position coupling, and velocity coupling, combined with graph theory. Defined as:
[0062] ;
[0063] ;
[0064] ;
[0065] ;
[0066] in, This represents the selection matrix for the navigating robotic arm. and They are respectively and Dimensional identity matrix For the desired terminal trajectory, Indicates the first The desired end effector speed of the robotic arm.
[0067] As a preferred technical solution, the problem of synchronous cooperative motion generation in the form of quadratic programming is transformed into a problem of finding the zero of the error function based on nonlinear complementary functions. An event-sampling distributed neurodynamic fault-tolerant model is constructed, and the optimal joint velocity of the multi-robotic arm system is solved. Specifically, this includes:
[0068] Based on the optimality condition, a nonlinear complementarity relation is constructed, and the nonlinear complementarity function is expressed as follows:
[0069] ;
[0070] in, and They are complementary variables;
[0071] The optimality condition of the quadratic programming problem is transformed into an error function. ,in, It is a combined variable consisting of combined joint velocities, equality constraint multipliers, and inequality constraint multipliers;
[0072] The error function for solving time-varying neural dynamics is constructed and expressed as follows:
[0073] ;
[0074] in, For convergence gain, It is a monotonically increasing odd activation function;
[0075] Construct an event-sampling distributed neurodynamic fault-tolerant model, represented as follows:
[0076] ;
[0077] in, , , Represents matrix parameters, The moment when the neural network update is triggered. To account for measurement error, and To adjust the sensitivity parameters, by... By taking the derivative and combining it with sampling and updating iterations, the optimal combination of joint velocities can be obtained.
[0078] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0079] (1) This invention unifies joint fault modeling, leader-follower cooperation constraints, physical constraints and quadratic programming optimization. Specifically, it comprehensively describes various joint fault forms from slight performance degradation (speed loss) to extreme lock-up through joint fault modeling. It integrates multi-robot end-effector trajectory tracking, position / velocity synchronization coupling and joint physical safety boundaries into a unified quadratic programming framework, ensuring that the motion generation results simultaneously meet the requirements of task accuracy and hardware safety. Each robotic arm only relies on its own state and neighbor communication information to participate in distributed solution, which greatly reduces the dependence on the central controller and improves the fault tolerance and autonomy under local faults and communication disturbances.
[0080] (2) This invention introduces event sampling into the recurrent neural network, transforming the high-frequency continuous iteration of the prior art into on-demand triggering updates. By dynamically constraining the activation and state update of the recurrent neural network through the event sampling mechanism, it enables the network to perform online calculations only when the system deviation reaches the triggering condition. This significantly reduces the redundant computational overhead of the local controller while ensuring high-precision formation synchronization and trajectory tracking of multiple terminal units under joint failure.
[0081] (3) The present invention realizes dynamic fault-tolerant redistribution. When some joints degenerate or are partially locked, the remaining intact joints can be dynamically scheduled through the speed redistribution mechanism to perform performance compensation, maintain high-precision alignment synchronization at the end, and is widely applicable to industrial collaborative assembly, embodied intelligent robots, dual-arm humanoid operation and cluster experimental platforms. Attached Figure Description
[0082] Figure 1 This is a flowchart illustrating the event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to the present invention.
[0083] Figure 2 This is a schematic diagram of the communication topology of the multi-robot manipulator system of the present invention;
[0084] Figure 3 This is a schematic diagram of the solution structure for the event sampling distributed neurodynamic fault-tolerant model of the present invention. Detailed Implementation
[0085] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0086] Example
[0087] like Figure 1 As shown, this embodiment provides an event-sampling neurodynamics fault-tolerant design method for distributed multi-manipulator synchronous motion. It dynamically constrains the activation and state update of the recurrent neural network through an event sampling mechanism, ensuring that online calculations are performed only when system deviations reach the triggering condition. This method is applicable to isomorphic or near-isomorphic multi-manipulator systems. While ensuring high-precision formation synchronization and trajectory tracking of multiple end-effectors under joint failures, it significantly reduces the redundant computational overhead of the local controller. The method includes the following steps:
[0088] S1: Constructed by A multi-robotic arm system consisting of multiple robotic arms was established. A kinematic model of the robotic arm from joint space to Cartesian space is established, and the undirected communication topology and Laplace matrix of the multi-robot robotic arm system are constructed. The undirected communication topology is used to describe the interaction relationship between the robotic arms.
[0089] In this embodiment, the first The kinematic model of the robotic arm is represented as follows:
[0090] ;
[0091] in, For the first The joint state variables of the robotic arm, Let be a smooth mapping function from joint space to Cartesian space. For the first The motion trajectory of the end effector of the robotic arm in the world coordinate system. The number of joints in the robotic arm. For the end-task space dimension;
[0092] like Figure 2 As shown, to describe the information interaction relationships between multiple robotic arms, an undirected communication topology is established, which is represented as follows: ,in, It is a set of nodes, with each node corresponding to a robotic arm. Let the set of edges represent the communication relationships between robotic arms. Let adjacency matrix and degree matrix be used. Laplace matrix ;
[0093] S2: For the first Taiwan robotic arm A joint performance loss fault model is constructed for each joint. The joint performance loss fault model is used to characterize the joint's normal output, multiplicative performance degradation, additive fault, and joint lock-up fault.
[0094] In this embodiment, a joint performance loss fault model is established using the joint's actual output, expected output, effective output rate, fault switching function, and additive fault term to identify joint normality, multiplicative performance degradation, additive faults, and lock-up faults.
[0095] In multi-robot collaborative operation, some joints of each robotic arm may experience varying degrees of performance degradation due to long-term use, load impact, reducer wear, encoder malfunction, or external collisions. If the... Taiwan robotic arm If a joint experiences multiplicative performance degradation, the actual joint angle output is expressed as:
[0096] ;
[0097] in, To achieve the desired joint angle output, For the effective output rate of the joint, when The closer to 1, the closer the joint output capacity is to normal; when The closer it is to 0, the weaker the joint output capability;
[0098] If the first Taiwan robotic arm If an additive fault occurs in one joint, the actual joint angle output is expressed as:
[0099] ;
[0100] in, Indicates an additive fault term;
[0101] Combining multiplicative performance degradation and additive faults, a joint performance loss fault model can be obtained, and the actual joint angle output is expressed as:
[0102] ;
[0103] in, For the first Taiwan robotic arm The actual joint angle output of each joint;
[0104] Taking the derivative of the above model, the corresponding faulty joint velocity is expressed as:
[0105] ;
[0106] in, To achieve the desired joint angle output, For joint effective output rate, This is the failover function. This is an additive fault term;
[0107] In this embodiment, the fault switching function Used to describe the abrupt change in the joint state before and after a failure, specifically represented as:
[0108] ;
[0109] in, The moment when the joint failure occurs. The time when the task execution terminates, before the failure occurs. The actual output equals the expected output after a fault occurs. The joint output is affected by the effective output rate and fault items.
[0110] when and At this time, the joint operates normally;
[0111] when and At this time, the joint's multiplicative properties deteriorate;
[0112] when and At that time, the joint experiences additive failure;
[0113] when and ,or and At that time, the joint locked up.
[0114] Therefore, this embodiment can cover a variety of actual failure modes of the robotic arm, ranging from slight degradation to severe lock-up;
[0115] S3: Under the master-slave collaboration framework, construct end-effector trajectory tracking constraints based on the expected end-effector trajectory of the master robotic arm, and construct position coupling constraints and velocity coupling constraints between different robotic arm end-effectors based on the communication topology.
[0116] In this embodiment, in a multi-robot collaborative task, one or more robotic arms are designated as the lead robotic arm, whose end effector tracks the desired trajectory. Other robotic arms are follower robotic arms, and their end effector motion is determined by the communication topology and the state of neighboring robotic arms.
[0117] The end effector trajectory tracking constraint of the guiding robotic arm is expressed as:
[0118] ;
[0119] in, For the desired terminal trajectory, Indicates the first The robotic arm in question is a navigation robotic arm. Indicates the first The robotic arm is a follower robotic arm;
[0120] Set end position error To ensure that the terminal position error converges to zero, the error convergence relation is defined according to the neurodynamic design principle as follows: , As the convergence rate adjustment parameter, combined with the robotic arm's Jacobian matrix, the velocity-level trajectory tracking constraint can be obtained as follows:
[0121] ;
[0122] in, For the first Jacobian matrix of the robotic arm For the first The joint speed of the robotic arm, For the desired end-effector velocity, this constraint allows the master robotic arm to compensate for end-effector trajectory errors by optimizing joint velocities even when joint performance degradation exists.
[0123] To maintain the cooperative formation and relative positional relationship among multiple robotic arm end effectors, let... In order to be with the first A set of robotic arms that have communication relationships. For the first Taiwan robotic arm and the first The desired relative relationship between the end effectors of the robotic arm, expressed by position coupling constraints, is as follows:
[0124] ;
[0125] because To maintain a constant velocity coupling constraint in collaborative tasks, the following expression is used:
[0126] ;
[0127] With the above constraints, the following robotic arm does not need to directly obtain global task information, but can maintain cooperative synchronization with the system as a whole only based on the status of neighboring robotic arms;
[0128] S4: Construct speed-level physical constraint inequality based on the upper limit of the joint angle, the lower limit of the joint angle, the upper limit of the joint speed, and the lower limit of the joint speed of the robotic arm.
[0129] To prevent joint overtravel, excessive speed, or unsafe operation of the robotic arm due to fault compensation, this embodiment introduces physical constraints on joint angles and joint velocities. Furthermore, the physical boundaries of joint angles and velocities are uniformly converted into velocity-level physical constraint inequalities, expressed as:
[0130] ;
[0131] ;
[0132] in, and These are the lower and upper limits of the joint angle, respectively. and These are the lower and upper limits of joint velocity, respectively.
[0133] By unifying the angle and velocity limits to the velocity layer, we obtain:
[0134] ;
[0135] ;
[0136] ;
[0137] in, and The lower and upper limits of the velocity level are obtained by uniformly converting the joint angle constraints, joint velocity constraints, and scaling thresholds.
[0138] The above processing method ensures that the multi-robot robotic arm system always meets the hardware safety boundaries during the online optimization process;
[0139] S5: Taking the minimum joint velocity norm of the robotic arm system as the optimization objective, the end-point trajectory tracking constraint, position coupling constraint, velocity coupling constraint and velocity level physical constraint are unified into a reliable synchronous cooperative motion generation problem in the form of quadratic programming.
[0140] In this embodiment, the joint positions and velocities of all robotic arms are combined as follows:
[0141] ;
[0142] ;
[0143] Combining graph theory and the first The constraints of a single robot, and the constraint matrix form for the end effector trajectory tracking of the master robotic arm of a multi-robot, are as follows:
[0144] ;
[0145] in, This is represented as the selection matrix for the navigation robotic arm. and They are respectively and Dimensional identity matrix It is a smooth mapping function;
[0146] Furthermore, the speed constraint is as follows:
[0147] ;
[0148] in, It is represented as a Jacobian matrix for multiple robots.
[0149] Similar to the derivation process above, the position and velocity coupling matrices between multiple robots are as follows:
[0150] ;
[0151] ;
[0152] in, It is represented as a relative position matrix of multiple robots;
[0153] Meanwhile, the physical constraints of multiple robots are represented as follows:
[0154] ;
[0155] in Represented as an identity matrix, Represented in physical constraint matrix form and and ;
[0156] Using the minimum velocity norm as the optimization objective, the problem of generating reliable synchronous cooperative motion is constructed as a quadratic programming problem with equality and inequality constraints, expressed as:
[0157] ;
[0158] ;
[0159] ;
[0160] ;
[0161] in, For the combined joint speeds of a multi-robot robotic arm system, For convergence parameters, and Composed of velocity-level physical constraint inequalities, the solution to this quadratic programming problem is the optimal combination of joint velocities under the conditions of cooperative synchronization and physical constraints.
[0162] In this embodiment, , , , The constraints consist of end-point trajectory tracking, position coupling, and velocity coupling, combined with graph theory. The definition, specifically, is as follows:
[0163] ;
[0164] ;
[0165] ;
[0166] ;
[0167] S6: Based on nonlinear complementary functions, the problem of synchronous cooperative motion generation in the form of quadratic programming is transformed into the problem of finding the zero of error function. An event sampling distributed neurodynamic fault-tolerant model is constructed and the optimal joint velocity of the multi-manipulator system is solved online.
[0168] In this embodiment, Lagrange multipliers are introduced to solve the quadratic programming problem online. and The corresponding Lagrange function is expressed as:
[0169] ;
[0170] Using the Karush-Kuhn-Tucker (KKT) conditions, we can obtain:
[0171] ;
[0172] Based on the optimality condition, a nonlinear complementary relationship is constructed, and the nonlinear complementary function is expressed as:
[0173] ;
[0174] in, and They are complementary variables;
[0175] Based on this nonlinear complementary function, the KKT conditions can be rewritten as follows:
[0176] ;
[0177] in, , For scalar values, For the Hadamard product, take the error function as... , , and ;
[0178] Subsequently, the optimality condition of the quadratic programming problem is transformed into an error function. ,in, It is a combined variable consisting of combined joint velocities, equality constraint multipliers, and inequality constraint multipliers;
[0179] To make the error function converge to zero, a time-varying neural dynamics is constructed to solve for the error function, expressed as:
[0180] ;
[0181] in, For convergence gain, It is a monotonically increasing odd activation function;
[0182] like Figure 3 As shown, to reduce resource consumption, an event-sampling distributed neurodynamic fault-tolerant model is constructed, represented as follows:
[0183] ;
[0184] in, The moment when the neural network update is triggered. To account for measurement error, and To adjust the sensitivity parameters, by... By taking the derivative and combining it with sampling updates, the online iterative equation can be obtained efficiently, and then the optimal combination of joint velocities can be found. ;
[0185] Specifically, the matrix parameter is defined as follows:
[0186] ;
[0187] ;
[0188] ;
[0189] Through the By taking the derivative and combining it with sampling updates, the online iterative equation can be obtained efficiently, thereby obtaining the optimal combination of joint velocities in real time. ;
[0190] This embodiment introduces an event sampling mechanism to dynamically constrain the activation timing of the network, and solves for the optimal joint velocity online only when the deviation meets the triggering condition, so as to eliminate redundant calculations;
[0191] S7: Sends the optimal joint speed to the joint execution units of each robotic arm, enabling a multi-robotic arm system with degraded joint performance or joint lockup to perform reliable synchronous cooperative motion.
[0192] In this embodiment, the optimal combination of joint velocities is used. Decomposed into the optimal joint speed for each robotic arm The data is then sent to the actuators of each robotic arm joint. When some joints experience performance degradation, the event-triggered recurrent neural network fault-tolerant model dynamically constrains the timing of network state updates. Online computation updates are only activated when system collaboration deviations or changes in neurodynamic states exceed a preset trigger threshold. Compensation is achieved by adjusting the speeds of other healthy joints, ensuring the end effector maintains the desired collaborative trajectory and the relative positional relationship between robots. In extreme failures such as joint locking, this event-triggered recurrent neural network system can constrain the speed of the locked joint to zero and continue feasible collaborative movements using the remaining movable joints. Simultaneously, at the algorithmic level, it effectively avoids meaningless high-frequency continuous iterative calculations during periods of system stabilization or failure-induced stagnation.
[0193] In this embodiment, the multi-robot manipulator system may include four six-degree-of-freedom manipulators, one of which is the master manipulator and the rest are follower manipulators. The communication topology can be set as a distributed structure in which the lead manipulator communicates with some of the follower manipulators, and the follower manipulators communicate locally with each other. Even if some joints of the follower manipulators experience a 50%, 60%, or 100% degradation in output capability, the distributed event-triggered recurrent neural network solver integrated locally by each manipulator can still perform highly timely, non-continuously triggered joint velocity redistribution based on the constraint optimization results, completing collaborative tracking tasks such as flower-shaped trajectories, pentagram trajectories, or Chinese character stroke trajectories.
[0194] This embodiment can be used to verify the collaborative safety of multiple robotic arms under conditions of joint degradation, external disturbances, and local joint lockup, as well as the superior performance of event-triggered recurrent neural networks in reducing the computational load on the local controller and redundant control calculations. It can be connected to a robotic arm controller, a real-time status acquisition unit, a joint health status estimation unit, an event-triggered detection unit, and a host computer task planning unit to perform multi-arm robot assembly, collaborative handling, complex trajectory demonstrations, embodied intelligent operation benchmark testing, and fault-tolerant control algorithm evaluation.
[0195] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.
Claims
1. A fault-tolerant design method for event sampling neurodynamics of distributed multi-robotic arm synchronous motion, characterized in that, Includes the following steps: Construct the kinematic model of the robotic arm, and construct the undirected communication topology and Laplace matrix between the robotic arms; A joint performance loss fault model is constructed to characterize normal joint output, multiplicative performance degradation, additive faults, and joint locking faults. In a multi-robot collaborative task, end-effector trajectory tracking constraints are constructed based on the desired end-effector trajectory of the master robot arm, and position coupling constraints and velocity coupling constraints between different robot arm end-effectors are constructed based on the communication topology. Based on the upper limit of the joint angle, the lower limit of the joint angle, the upper limit of the joint speed, and the lower limit of the joint speed of the robotic arm, construct a speed-level physical constraint inequality. Taking the minimum joint velocity norm of the robotic arm system as the optimization objective, the end-effector trajectory tracking constraint, position coupling constraint, velocity coupling constraint and velocity level physical constraint are unified into a synchronous cooperative motion generation problem in the form of quadratic programming. Based on nonlinear complementary functions, the problem of synchronous cooperative motion generation in the form of quadratic programming is transformed into a problem of finding the zero of error function. An event sampling distributed neurodynamic fault-tolerant model is constructed and the optimal joint velocity of the multi-manipulator system is solved. The optimal joint speed is sent to the joint execution units of each robotic arm, enabling the multi-robotic arm system with joint performance degradation or joint lockup to perform synchronized cooperative motion.
2. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, The kinematic model of the robotic arm is constructed as follows: ; in, Indicates the first The joint state variables of the robotic arm, Let f(x) represent a smooth mapping function from joint space to Cartesian space. Indicates the first The motion trajectory of the end effector of the robotic arm in the world coordinate system. This indicates the number of robotic arms.
3. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, Constructing the undirected communication topology and Laplace matrix between robotic arms, specifically including: Undirected communication topology is represented as: ; in, This represents a set of nodes, where each node corresponds to a robotic arm. This represents a set of edges used to indicate the communication relationships between robotic arms. Represents the adjacency matrix; The Laplace matrix is expressed as: ; in, Representation degree matrix.
4. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, Construct a joint performance loss fault model, and represent the actual joint angle output as follows: ; in, For the first Taiwan robotic arm The actual joint angle output of each joint. To achieve the desired joint angle output, This represents the failover function. For joint effective output rate, For additive faults occurring in the joint; Output of actual joint angles The corresponding fault joint velocity is obtained by taking the derivative.
5. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 4, characterized in that, The fault-to-connect function is expressed as follows: ; in, The moment when the joint failure occurs. The time when the task execution terminates, before the failure occurs. The actual output equals the expected output after a fault occurs. The joint output is affected by the effective output rate and fault items; when and At this time, the joint operates normally; when and At this time, the joint's multiplicative properties deteriorate; when and At that time, the joint experiences additive failure; when and ,or and At that time, the joint locked up.
6. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, In a multi-robot collaborative task, an end-effector trajectory tracking constraint is constructed based on the desired end-effector trajectory of the master robot arm, expressed as: ; in, For the desired terminal trajectory, Indicates the first The robotic arm is the main robotic arm. Indicates the first The robotic arm is a follower robotic arm. Indicates the first The joint state variables of the robotic arm, Represents a smooth mapping function from joint space to Cartesian space; Set end position error The error convergence relationship is set as follows: , As the convergence rate adjustment parameter, combined with the robotic arm's Jacobian matrix, the velocity-level trajectory tracking constraint can be obtained as follows: ; in, For the first Jacobian matrix of the robotic arm For the first The joint speed of the robotic arm, Indicates the first The desired end effector speed of the robotic arm.
7. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, Position coupling constraints are expressed as: ; in, For the first Taiwan robotic arm and the first The expected relative relationship between the end effectors of the robotic arm. Represents the elements of the adjacency matrix. , They represent the first Taiwan robotic arm, the first The desired end effector trajectory of the robotic arm; The velocity coupling constraint is expressed as: ; in, In order to be with the first A set of robotic arms that have a communication relationship.
8. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, Based on the upper limit of the joint angle, the lower limit of the joint angle, the upper limit of the joint velocity, and the lower limit of the joint velocity, a velocity-level physical constraint inequality is constructed, specifically expressed as follows: ; ; ; ; ; in, and These are the lower and upper limits of the joint angle, respectively. and These are the lower and upper limits of joint velocity, respectively. and These are the lower limit and upper limit of the speed level, respectively.
9. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, The problem of generating synchronous cooperative motion in the form of quadratic programming is represented as: ; ; ; ; in, This indicates the combined joint velocities of a multi-robot robotic arm system. For convergence parameters, It is the identity matrix. The matrix form representing the physical limits of the velocity level. Represents a relative position matrix. Represents the Jacobian matrix. Indicates the combined joint positions of a multi-robot robotic arm system. Represents a smooth mapping function. , , , The constraints consist of end-point trajectory tracking, position coupling, and velocity coupling, combined with graph theory. Defined as: ; ; ; ; in, This represents the selection matrix for the navigating robotic arm. and They are respectively and Dimensional identity matrix For the desired terminal trajectory, Indicates the first The desired end effector speed of the robotic arm.
10. The event sampling neurodynamics fault-tolerant design method for distributed multi-robotic arm synchronous motion according to claim 1, characterized in that, Based on nonlinear complementary functions, the problem of synchronous cooperative motion generation in the form of quadratic programming is transformed into a problem of finding the zero of the error function. An event-sampling distributed neurodynamic fault-tolerant model is constructed to solve the optimal joint velocities of a multi-robotic arm system. Specifically, this includes: Based on the optimality condition, a nonlinear complementarity relation is constructed, and the nonlinear complementarity function is expressed as follows: ; in, and They are complementary variables; The optimality condition of the quadratic programming problem is transformed into an error function. ,in, The combined variable consists of the combined joint velocity, equality constraint multiplier, and inequality constraint multiplier. The error function for solving time-varying neural dynamics is constructed and expressed as follows: ; in, For convergence gain, It is a monotonically increasing odd activation function; Construct an event-sampling distributed neurodynamic fault-tolerant model, represented as follows: ; in, , , Represents matrix parameters, The moment when the neural network update is triggered. To account for measurement error, and To adjust the sensitivity parameters, by... By taking the derivative and combining it with sampling and updating iterations, the optimal combination of joint velocities can be obtained.
