CN-121670690-B - Distributed self-adaptive control method and system for multi-mechanical arm cooperation

CN121670690BCN 121670690 BCN121670690 BCN 121670690BCN-121670690-B

Abstract

The invention belongs to the technical field of cooperative control of robots, and particularly relates to a distributed self-adaptive control method and system for cooperation of multiple mechanical arms, wherein the method comprises the following steps: decoupling a pure internal force vector from terminal six-dimensional force sensor data by utilizing pre-recognized dynamic parameters, calculating a cooperation countermeasure index according to the pure internal force vector and a relative position vector, calculating self-adaptive internal force penalty weight through nonlinear mapping based on the cooperation countermeasure index, establishing a state space equation of a system based on a double integral model, introducing an equivalent environment stiffness matrix, mapping a prediction error vector in a prediction time domain into a prediction internal force vector, constructing a cost function containing the product of the self-adaptive internal force penalty weight and the prediction internal force vector, obtaining an optimal control increment sequence through minimizing the cost function, and updating a speed instruction to drive a mechanical arm to move. The invention effectively improves the robustness and execution stability of the system and ensures the cooperation safety.

Inventors

ZHU PEI
LAN WEI
YU YI
ZHAO LIMING

Assignees

武汉孚锐利自动化设备有限公司

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (3)

1. A distributed adaptive control method for multi-robot collaboration, comprising: acquiring a state measured value of the mechanical arm in real time, and decoupling a pure internal force vector from the six-dimensional force sensor data at the tail end of the state measured value by utilizing a pre-identified dynamic parameter; The method comprises the steps of calculating a cooperative countermeasure index according to a pure internal force vector and a relative position vector, wherein the cooperative countermeasure index is obtained by multiplying a projection item and an activation item, the projection item is obtained by dividing the absolute value of a dot product of the decoupled pure internal force vector and a neighbor mechanical arm acquired by local and communication by the modular length of the relative position vector, the activation item is constructed based on a Sigmoid function, an independent variable of the activation item is the difference value between the modular length of the decoupled pure internal force vector and a safe internal force dead zone threshold value, the safe internal force dead zone threshold value is used for filtering sensor background noise and setting allowed basic internal force, the value range is set to be 5-10% of the maximum allowable stress of a workpiece, and an adaptive internal force penalty weight is calculated based on the cooperative countermeasure index through nonlinear mapping, the adaptive internal force penalty weight is equal to the basic weight and is obtained by multiplying a difference value of a saturation weight and the basic weight by a function item related to the cooperative countermeasure index, and the function item is obtained by a ratio of a natural function item comprising the product of the cooperative countermeasure index and an adjustment sensitivity coefficient and the function item is a mother item: In which, in the process, Is that A time of day collaborative countermeasure index; in order to adjust the sensitivity coefficient, the value range is 0.1, 0.5; The method comprises the steps of establishing a state space equation of a system based on a double integral model, calculating a prediction error vector of a future time in a prediction time domain, wherein the prediction error vector of the future time is equal to the sum of an autonomous evolution part of the system and a control input influence part, the autonomous evolution part of the system is obtained by multiplying a state transition matrix by an error vector containing position errors and speed errors at the current time, the control input influence part is obtained by multiplying an input control matrix by a control increment to be solved, introducing an equivalent environment stiffness matrix representing rigidity characteristics of a workpiece and a clamp, and the equivalent environment stiffness matrix is set by an experimental measurement method, and comprises the steps of controlling a mechanical arm to clamp the workpiece to be in a static state in a system initialization stage, and controlling one arm to respectively generate a tiny and known position displacement in the directions of all axes of a Cartesian space Recording the change in the force sensor readings According to the formula The method comprises the steps of calculating rigidity values of all shafts, constructing a diagonal matrix as a preset equivalent environment rigidity matrix, mapping a prediction error vector in a prediction time domain into a prediction internal force vector, wherein the prediction internal force vector is equal to the preset equivalent environment rigidity matrix multiplied by equivalent position deviation at the moment; Constructing a cost function comprising the product of the self-adaptive internal force penalty weight and the predicted internal force vector, wherein the cost function is the summation of the squares of three weighted norms in a prediction time domain, the first term is the square of the weighted Euclidean norms of the prediction error vector and the state weight matrix, the second term is the product of the square of the Euclidean norms of the predicted internal force vector and the self-adaptive internal force penalty weight at the current moment, the third term is the square of the weighted Euclidean norms of the control increment to be solved and the control increment weight matrix, and the optimal control increment sequence is obtained by minimizing the cost function, and the speed instruction is updated to drive the mechanical arm to move.
2. The method according to claim 1, wherein the pure internal force vector is decoupled from the end six-dimensional force sensor data of the state measurement value by using the pre-recognized dynamic parameters, and the decoupled pure internal force vector is equal to the end six-dimensional force sensor data minus a dynamic resultant force required for maintaining the current motion, wherein the dynamic resultant force is calculated by the pre-recognized dynamic parameters and is equal to the sum of an inertial force term represented by the product of the equivalent inertia of the mechanical arm in the cartesian space and the cartesian acceleration of the mechanical arm, a nonlinear force term represented by the product of the coriolis force of the mechanical arm in the cartesian space and the centrifugal force and the cartesian speed of the mechanical arm, and a gravity compensation term of the mechanical arm in the cartesian space.
3. A distributed adaptive control system for multi-robot collaboration, comprising a processor and a memory storing computer program instructions that, when executed by the processor, implement a distributed adaptive control method for multi-robot collaboration according to any of claims 1-2.

Description

Distributed self-adaptive control method and system for multi-mechanical arm cooperation Technical Field The invention relates to the technical field of cooperative control of robots. More particularly, the invention relates to a distributed adaptive control method and system for multi-mechanical arm cooperation. Background In modern intelligent manufacturing production lines, the cooperative transportation of large-load and rigid objects by multiple mechanical arms has become a core process link, in the system, each mechanical arm controller keeps high synchronization of a motion state through communication so as to ensure stability and safety of tasks, a traditional centralized control architecture faces computational power bottlenecks and single-point fault risks when facing large-scale cooperative nodes, and a Distributed Model Predictive Control (DMPC) algorithm becomes a main technical means for solving the problem of cooperative control of multiple intelligent objects because the algorithm can effectively process multivariable constraint and perform online rolling optimization. Although the DMPC algorithm is excellent in track tracking, the DMPC algorithm still has the remarkable defects that in the aspect of weighing control performance, balance needs to be found between position tracking precision and flexibility in the aspect of multi-mechanical arm cooperation, an existing control method often adopts a fixed weight coefficient to construct a cost function, namely the weight of a position error item and a control increment item is kept unchanged in the whole operation process, however, the fixed weight coefficient is difficult to find an optimal balance point between the position tracking precision and the flexibility, if a higher position weight is set for pursuing high precision, huge internal force is generated when motion errors occur, workpieces are damaged, if lower weight is set for pursuing flexibility, the rigidity of the system is insufficient, the track precision in the carrying process is difficult to ensure, and the cooperation fails, and the control strategy lacking the self-adaptive adjustment capability is difficult to adapt to the dynamically-changed cooperation environment and complex task requirements. Disclosure of Invention In order to solve the technical problem that the multi-mechanical arm control method in the prior art is difficult to balance the position accuracy and the flexibility, the invention provides the following aspects. In a first aspect, the invention provides a distributed adaptive control method for multi-mechanical arm cooperation, which comprises the steps of collecting state measurement values of mechanical arms in real time, decoupling pure internal force vectors from end six-dimensional force sensor data of the state measurement values by utilizing pre-recognized dynamic parameters, calculating cooperation countermeasures according to the pure internal force vectors and relative position vectors, calculating adaptive internal force punishment weights through nonlinear mapping based on the cooperation countermeasures, establishing a state space equation of a system based on a double-integral model, calculating prediction error vectors at future time in a prediction time domain, introducing an equivalent environment stiffness matrix representing rigid characteristics of a workpiece and a clamp, mapping the prediction error vectors in the prediction time domain into the prediction internal force vectors, constructing a cost function comprising products of the adaptive internal force punishment weights and the prediction internal force vectors, obtaining an optimal control increment sequence by minimizing the cost function, and updating a speed instruction to drive the mechanical arms to move. Preferably, the pure internal force vector is decoupled from the end six-dimensional force sensor data of the state measurement value by using the pre-recognized dynamic parameters, and the decoupled pure internal force vector is equal to the end six-dimensional force sensor data minus the dynamic resultant force required for maintaining the current motion, wherein the dynamic resultant force is calculated by the pre-recognized dynamic parameters and is equal to the sum of an inertial force term represented by the product of the equivalent inertia of the mechanical arm in Cartesian space and the Cartesian acceleration of the mechanical arm, a nonlinear force term represented by the product of the coriolis force of the mechanical arm in Cartesian space and the centrifugal force and the Cartesian speed of the mechanical arm, and a gravity compensation term of the mechanical arm in Cartesian space. The collaborative countermeasure index is obtained by multiplying a projection item and an activation item, wherein the projection item is obtained by dividing the absolute value of the dot product of the decoupled pure internal force vector and the relative position vector bet