CN-121132623-B - Singular point induction-based mechanical arm planning anti-attack method

CN121132623BCN 121132623 BCN121132623 BCN 121132623BCN-121132623-B

Abstract

The invention belongs to the field of mechanical arm control safety, and discloses a mechanical arm planning attack resistance method based on singular point induction, which comprises the following steps of S1, acquiring joint angle vectors and tail end pose of a mechanical arm, and generating thermodynamic diagrams based on the joint angle vectors; S2, acquiring a singular region based on a thermodynamic diagram, performing multi-objective optimization based on the singular region, acquiring the position of an obstacle, S3, judging whether the position of the center of the obstacle is located in an accessible region, if so, moving the physical obstacle to the position of the center of the obstacle, S4, controlling the mechanical arm to perform servo loop execution according to the original production track, and recording the curve and the tail end deviation of the minimum singular value. According to the invention, singular probability modeling is superimposed on the necessary region of the grabbing path, and small obstacles are arranged through a genetic algorithm, so that the kinematic margin of the picking and placing cycle is specially weakened, the degradation can be triggered by modifying the external environment, the clamping jaw control logic or the visual identification model is not required to be modified, and the compatibility risk to the production line software is avoided.

Inventors

TANG KEKE
HAO TIANYU
CAO ZHENXING
PENG WEILONG
LI SHUDONG
TIAN ZHIHONG

Assignees

广州大学

Dates

Publication Date: 20260508
Application Date: 20250808

Claims (9)

1. The mechanical arm planning attack resistance method based on singular point induction is characterized by comprising the following steps of: s1, acquiring joint angle vectors and tail end pose of a mechanical arm, and generating a thermodynamic diagram based on the joint angle vectors; s2, acquiring a singular region based on thermodynamic diagram, and performing multi-objective optimization based on the singular region to acquire the position of an obstacle; s3, judging whether the position of the center of the obstacle is located in an accessible area, if so, moving the physical obstacle to the position of the center of the obstacle; S4, controlling the mechanical arm to perform servo circulation execution according to the original production track, and recording a curve of the minimum singular value and terminal deviation; performing multi-objective optimization based on the singular zones to obtain the positions of the obstacles, including: each obstacle is regarded as a radius Spherical influence domain of (2) on voxels The method comprises the following steps: In the middle of For the original reachable target score, Is a voxel Is defined by the center point coordinate vector of (c), Is the center point coordinate vector of the i-th obstacle, Representation of And Euclidean distance between them; To influence the weights; n represents the total number of obstacles in the optimization of the round; an influence radius for the ith obstacle; Representing voxels A reachable target score after considering the obstacle; modeling the induced robot arm arrival singular zone as multi-objective optimization, maximizing end path integral While minimizing the number of obstructions; Representing the joint angle vector measured at the sampling instant t; and constructing a fitness function by adopting the GA to code the center and the radius of the obstacle.
2. The method for planning challenge based on singular point induction of a mechanical arm according to claim 1, wherein S1 comprises: S10, the upper computer adopts a robot operating system to perform kinematic analysis on the mechanical arm and outputs a joint angle vector q of the mechanical arm in real time; s11, calling KDL under preset frequency, wherein JacobianSolver calculates Jacobian matrix Obtaining six singular values And with minimum singular value Normalized construction singular index ; S12, dividing a Cartesian space into voxels with preset length by adopting octree, obtaining all feasible joint solutions by inverse kinematics at the center position of the voxels, and marking the largest singular degree as the singular value of the voxels; and S13, coloring and displaying the voxel by using an Open3D graphical interface, and enabling a vision sensor to scan the pose of the tail end of the mechanical arm in real time and highlight the current position in the thermodynamic diagram.
3. A method of robot arm planning challenge based on singular point induction according to claim 2, characterized in that the preset frequency comprises 1kHz.
4. A method of robotic arm planning challenge based on singular point induction according to claim 2, wherein the preset length comprises 30mm.
5. A method of robot arm planning challenge based on singular point induction according to claim 2, characterized in that the singular values are minimized Normalized construction singular index Comprising: Wherein the method comprises the steps of The maximum singular value obtained in the global search before the experiment.
6. The method for robot arm planning challenge based on singular point induction according to claim 5, wherein obtaining singular regions based on thermodynamic diagrams comprises: And when S (q) is more than or equal to T, marking the corresponding voxel as a singular voxel, wherein all the odd voxels form a singular region.
7. A method of robot arm planning challenge based on singular point induction according to claim 1, characterized in that the physical barrier comprises a cylindrical stop.
8. A method of robot arm planning challenge based on singular point induction according to claim 1, characterized by moving a physical obstacle to a position in the center of the obstacle, comprising: and the physical barrier is moved to the position of the center of the barrier by the movable chassis, so that the modification of the operation environment of the mechanical arm is completed.
9. The method for planning challenge based on singular point induction of a robot arm according to claim 1, wherein the controlling the robot arm to perform servo loop execution according to an original production track comprises: the upper computer issues the original production track to the controller through the API opened by the robot manufacturer, and the controller does not make any software modification and only executes according to the conventional servo cycle.

Description

Singular point induction-based mechanical arm planning anti-attack method Technical Field The invention relates to the field of mechanical arm control safety, in particular to a mechanical arm planning anti-attack method based on singular point induction. Background Industry 4.0 wave promotes the manufacturing industry to accelerate to evolve to flexibility, intelligence and high reliability, and the serial mechanical arm with six degrees of freedom and above is widely applied to scenes such as automobile welding, 3C assembly, cooperation logistics, additive manufacturing and the like by virtue of high precision, high load and programmability. Among these tasks, the robotic arm is required to perform a continuous, smooth, and highly rigid trajectory in a confined space. However, due to the geometric constraint and nonlinear coupling characteristics of the connecting rod structure, the mechanical arm can be in a singular state of jacobian matrix determinant zeroing under certain joint combinations, wherein the terminal speed mapping is highly sensitive to joint speed, so that the degree of freedom is lost or the joint speed tends to be infinite theoretically, and if the controller still outputs according to the conventional gain, serious jitter, out-of-tolerance and even sudden stop are caused, so that the production line beat and equipment safety are directly threatened. Aiming at the identification and alleviation of singular risks, the academic world firstly gives a quantifiable index. Yoshikawa proposed operability metrics in 1985This value is positively correlated with the jacobian outlier and is reduced to a "minimum singular value" threshold in the industrial controller, which is used as an online singular pre-warning. The following ISO10218 safety standards also suggest to slow down or re-program when operability is below a threshold. With the improvement of computing capability, the controller can calculate the Jacobian singular value in millisecond level and look forward for track fragments above 20ms in real time, so that soft scram is realized instead of hardware power off, and the shutdown reliability is improved. In the aspect of commercial implementation, the FANUC controller integrates a singular point deceleration/attitude switching algorithm, wherein the minimum singular value is monitored in real time in a track interpolation look-ahead module, and when the singular risk is detected, the terminal speed is switched or dynamically reduced through a redundant attitude library, and the related principles and implementations are protected by US 6845295B 2. Siemens SINUMERIK moves the singular region offline analysis to the CAM stage aiming at CN 113305881B of the five-axis machine tool, samples the G code track and calculates the characteristic value of the included angle formed by the joint axis pair, alerts the programmer through color labeling, and can also start an online compensation function on the controller if the workpiece clamping of the production site is offset, so that the singular avoidance is doubly ensured. For possessing redundant degree of freedom) The coordinated robots CHIAVERINI of (a) proposed a singular robust task first (SR-TP) strategy in 1997, which projected manipulability gradients to jacobian zero space, guiding redundant joints to move in the "maximize operability" direction, while maintaining task accuracy. The Boeing company further reconstructs the jacobian structure through the concept of a virtual joint, so that the controller automatically introduces damping or replaces failure degrees of freedom nearby the singular, thereby ensuring terminal continuity and joint rate constraint. In addition to on-line control, production safety monitoring systems begin to present risks via "visual thermodynamic diagrams". Battelle ENERGY ALLIANCE proposes to use mobile robots to carry multi-modal sensors, collect the hazardous concentration in real time and generate a "discrete grid-hazardous intensity" thermodynamic diagram at the GUI, facilitating the operator to visually see the spatial risk gradients. The concept is to pave the road visually with 'singular degree', namely, allocate risk weight for each pose in a discrete working space, so that maintenance personnel can perceive singular risk distribution without analyzing complex mathematics. Meanwhile, the robot path planning research is conducted to deeply explore dynamic environment interference. GM Global Technology adopts 3D vision to estimate the state of the moving obstacle in real time, and combines the harmonic planner to output a safe speed vector, so as to realize avoidance of a fast moving human body or forklift. The academic world verifies the antagonistic characteristics of 'slight obstacle disturbance can obviously deteriorate the path' on a sampling-optimizing planner, but related schemes still aim at 'avoiding collision and ensuring accessibility', and lack of a system thought for actively inducing a mec