Paper bag production line cooperative control method and system

Abstract

The invention belongs to the technical field of intelligent control, and discloses a paper bag production line cooperative control method and system. The method comprises the steps of synchronously collecting paper bag data and process parameters, and performing feature extraction on the paper bag data to obtain paper bag features; splicing the paper bag features and the process parameters to obtain a multi-dimensional feature vector; according to the multi-dimensional feature vector, generating a collision-free path at the edge node based on a path planning algorithm; different modes in the dual-mode speed planning method are switched through a dynamic switching strategy to optimize the collision-free path to obtain an optimized track; splitting the optimized track into discrete control instructions, and distributing the discrete control instructions to each control device of the production line for execution through a communication network; according to the method, the paper bag carrying breakage rate can be effectively reduced, the path planning time consumption and the production line beat synchronization error are reduced, the stability and efficiency of flexible material production are remarkably improved, and meanwhile process waste caused by neglect of material characteristics is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, and more particularly, relates to a paper bag production line cooperative control method and system. BACKGROUND

[0002] With the promotion of "using paper instead of plastic" and the surge in demand for environmentally friendly paper bags in the consumer market, the traditional production mode has been unable to meet the requirements of large-scale and diversified production.

[0003] Chinese patent application with publication number CN115922716A discloses a two-way RRT-connect algorithm that realizes rapid path planning of an industrial robot by fusing process knowledge: S101: based on process knowledge, determine the initial pose of the mechanical arm and the feasible solution set of the target attitude, obstacle information and parameter information of the industrial mechanical arm; S102: based on a dynamic collision detection algorithm, perform collision detection on the two-way attitude in the initial and final point pose set of the mechanical arm, if there is a conflict-free trajectory, select the shortest attitude as the final initial and final attitude, if there is no conflict-free trajectory, go to S103; S103: first, take the traditional RRT-connect algorithm as the basis, set the search step as the basis for each step extension, and set the termination threshold as the final connection range of the two random trees of the starting point and the ending point; then, based on the expert mechanism, quickly filter out new nodes in the search space; S104: the industrial robot starts from the initial position and the end position according to the above strategy, generates two random trees, and then determines whether the two random trees are within the range of direct connection according to the termination threshold, if the termination threshold is lower, stop searching, connect, otherwise the mechanical arm still executes according to the above strategy until the termination threshold is lower; S105: output the position and attitude of each path point in order from the initial position as a collision-free path. The application fuses process knowledge and performs motion planning for the industrial robot. Specifically, based on the industrial environment of the industrial robot, the corresponding process knowledge is applied to determine the initial pose set of the robot. According to the initial pose set, two-way dynamic collision is performed on the space nodes, and in the process of two-way RRT sampling, the expert mechanism is introduced, so that the industrial robot finds a collision-free path in the work space while improving the search efficiency and ensuring the path quality.

[0004] Although the above method can meet most scenarios, research and actual application of the above method and prior art have found that the above method and prior art at least have the following defects:

[0005] The above method is usually only concerned about the workpiece geometry in the industrial robot path planning process, however, as a flexible material, the morphology of the paper bag is easily deformed by environmental factors such as temperature and humidity; However, the existing path planning relies on fixed process rules such as "detour when encountering obstacles" and "priority shortest path", which cannot adjust the strategy according to the flexible characteristics of the material, for example, when carrying workpieces of different materials (such as iron blocks and paper bags), the same speed and path curvature are used, which is easy to cause the paper bag to be thrown and broken during the carrying process or damaged due to uneven stress, thereby causing unnecessary material waste.

[0006] In view of this, the paper bag production line cooperative control method and system are proposed to solve the above problems. SUMMARY

[0007] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a paper bag production line cooperative control method, comprising the following steps:

[0008] Synchronously collecting paper bag data and process parameters, extracting features from the paper bag data to obtain paper bag features; splicing the paper bag features and process parameters to obtain a multi-dimensional feature vector;

[0009] According to the multi-dimensional feature vector, a collision-free path is generated at the edge node based on a path planning algorithm;

[0010] The different modes in the dual-mode speed planning method are switched by a dynamic switching strategy to optimize the collision-free path to obtain an optimized trajectory;

[0011] The optimized trajectory is split into discrete control instructions, which are distributed to each control device in the production line through a communication network for execution.

[0012] Further, the paper bag data includes paper bag point cloud data and thermal imaging data; the process parameters include tension sensor data, paper feeding roller speed and mechanical arm joint angle; a temperature-material mapping relationship is established according to the thermal imaging data, the thermal imaging data is converted to obtain the temperature;

[0013] The method for obtaining the temperature comprises:

[0014] The blackbody calibration temperature detector is used to establish a fitting relationship between the pixel output value and the blackbody temperature through Planck's law; for the measured material, the same radiation power is maintained, the real-time pixel output value of the thermal imaging data is measured, the blackbody temperature is calculated based on the fitting relationship, and the material emissivity is obtained according to the ratio of the blackbody temperature to the real temperature of the material;

[0015] According to the Planck law, a relationship model between radiation brightness and temperature is established, and an inversion model about radiation brightness is built according to received radiation brightness, environmental reflected radiation and material emissivity; the inversion model is solved to obtain the temperature.

[0016] Further, the paper bag features include geometric features and material features;

[0017] The method for obtaining the geometric features includes:

[0018] The point cloud data is subjected to key point detection to obtain ROI point cloud;

[0019] The ROI point cloud is analyzed to obtain the length, width and height of the paper bag, the crease line position and the handle hole coordinates, and the geometric features are spliced;

[0020] The method for obtaining the ROI point cloud includes:

[0021] The k-neighborhood of each point is obtained using k-d tree search;

[0022] The density distribution of the local region is described by integral invariants;

[0023] The second derivative of the integral invariants is solved to obtain the Hessian matrix;

[0024] The Hessian matrix is subjected to eigenvalue decomposition to obtain principal curvatures k1, k2 and k3, wherein k1 < k2 < k3;

[0025] The principal curvature k3 is subjected to normalization processing to obtain a response function;

[0026] The paper bag point cloud data is filtered according to a preset screening threshold, only the points with a response function value greater than the preset screening threshold are reserved, and the ROI point cloud is obtained;

[0027] The method for obtaining the length, width and height of the paper bag includes:

[0028] The ROI point cloud is subjected to axis-aligned bounding box fitting to calculate the length, width and height of the paper bag.

[0029] Further, the method for obtaining the crease line position includes:

[0030] The k-neighborhood of the ith point p i is obtained using k-d tree search;

[0031] The mean value and the covariance matrix C of the neighborhood points of p i are calculated, C is subjected to eigenvalue decomposition, and the eigenvector corresponding to the minimum eigenvalue is taken as the normal vector of p i ;

[0032] The normal vector of p ithe normal vector of the nearest neighbor point p i+1 The points satisfying the angle greater than the preset angle threshold are marked as crease line candidate points, and the crease line candidate points are collected into a set P.

[0033] Two points p a and p b are randomly selected from P to build a candidate straight line L, the distances of all points in P to L are calculated, and the points with distances less than a preset distance threshold are divided into inner points, and the straight line with the most inner points is retained as the crease line position.

[0034] Further, the method for obtaining the handle hole coordinates comprises:

[0035] The actual number of points in the neighborhood of each point and the maximum number of points in the neighborhood are counted, and the neighborhood occupancy rate of each point is calculated; the points with a neighborhood occupancy rate lower than a preset occupancy rate threshold are judged as hole edge points.

[0036] An implicit surface is fitted to the hole edge points to generate a continuous surface model, and virtual points are filled in the hole interior;

[0037] The holes satisfying the closure rule are determined according to the closure rule, and a closed region is obtained;

[0038] The closure rule includes that the perimeter to diameter ratio of the point cloud ring is within a preset variation range, and the point cloud occupancy rate in the ring is lower than a preset point cloud occupancy rate.

[0039] The centroid of the point cloud in the closed region is calculated as an initial estimate of the center of the circle;

[0040] The hole edge point cloud is fitted by the least square method, the optimal circle parameters including the center and the radius are solved by minimizing the sum of squares of the distances from the points to the circle, and the center is taken as the handle hole coordinates.

[0041] Further, the method for obtaining the material feature comprises:

[0042] The global statistics of the thermal imaging data, including the mean and the variance, are calculated;

[0043] The temperature gradient of a preset local region is calculated;

[0044] The point cloud normal vector of each point is obtained;

[0045] The local height standard deviation of the preset local region of the point cloud is calculated;

[0046] The mean, the variance, the temperature gradient, the point cloud normal vector and the local height standard deviation are standardized and spliced into a feature vector, the feature vector is taken as the input of the material classification model, and the material feature is obtained.

[0047] Further, the method for obtaining the collision-free path comprises:

[0048] Obtaining the start point and the end point of the production line robot arm, wherein the end point is obtained according to the geometric feature and a reserved safety distance;

[0049] Selecting a control mode according to the material feature;

[0050] Defining a state space of the robot arm, the state space being composed of a state vector, the state vector including a robot arm pose and a multi-dimensional feature vector;

[0051] Uniformly sampling within a working space of the robot arm, the sampling points satisfying feature constraints, including geometric constraints and process constraints;

[0052] Respectively creating a tree expanded from the start point and a tree expanded from the end point, marked as Ts and Tf, each tree storing node information, including coordinates, a parent node and a path cost;

[0053] In each iteration, first expanding Ts and then expanding Tf, until the node distance of the two trees is less than a preset connection threshold;

[0054] When the distance between the nearest nodes of Ts and Tf is less than the preset connection threshold, checking whether the path between the two points is collision-free, if so, merging the two trees to obtain an initial path, otherwise, re-building Ts and Tf.

[0055] Further, the dual-mode velocity planning method includes a high-speed mode and a high-precision mode;

[0056] The method for switching different modes in the dual-mode velocity planning method through a dynamic switching strategy includes:

[0057] Calculating the path curvature, and when the path curvature exceeds a preset curvature threshold, switching to the high-precision mode;

[0058] When detecting that the paper roll rotation speed changes in real time, comparing the paper roll rotation speed with a preset rotation speed threshold, when the paper roll rotation speed is greater than the preset rotation speed threshold, selecting the high-speed mode; otherwise, selecting the high-precision mode;

[0059] When detecting that the paper bag mass is greater than a preset mass threshold, forcibly enabling the high-precision mode.

[0060] Further, the method for optimizing the collision-free path through the high-speed mode includes:

[0061] Obtaining the collision-free path, taking the start point, the end point and the midpoint of each path, and the midpoints of the previous and subsequent segments, to generate four control points of a cubic Bezier curve;

[0062] Building a Bezier curve equation according to the four control points;

[0063] When the robot arm end moves along the curve, the centripetal acceleration satisfies a preset high-speed constraint condition;

[0064] The movement speed of the mechanical arm is controlled by a trapezoidal speed method; trapezoidal speeds of acceleration segments, constant speed segments and deceleration segments are preset, and proportions of the acceleration segments, the constant speed segments and the deceleration segments satisfy preset proportions; wherein the acceleration segments are obtained according to acceleration calculation, and the acceleration is not higher than the maximum acceleration of the mechanical arm; the constant speed segments are obtained according to maximum centripetal acceleration calculation; and the deceleration segments are obtained according to constant speed segments speed, acceleration, acceleration time and constant speed time calculation;

[0065] The method for optimizing the collision-free path by the high-precision mode comprises:

[0066] The speed is divided into M stages by an S-shaped curve;

[0067] The total time of the mechanical arm operation, i.e., the sum of the times of the M stages, is minimized, the preset high-precision constraint condition is satisfied, and the optimized trajectory is obtained by solving by a natural heuristic optimization algorithm.

[0068] The paper bag production line collaborative control system implements the paper bag production line collaborative control method, comprising:

[0069] The acquisition and processing module: synchronously acquires paper bag data and process parameters, extracts features of the paper bag data to obtain paper bag features, and splices the paper bag features and the process parameters to obtain a multi-dimensional feature vector;

[0070] The trajectory planning module: generates a collision-free path at an edge node based on an improved RRT*-Connect algorithm according to the multi-dimensional feature vector;

[0071] The path optimization module: optimizes the collision-free path by switching different modes in the dual-mode speed planning method through a dynamic switching strategy;

[0072] The instruction execution module: splits the optimized trajectory into discrete control instructions, and distributes the instructions to each control device of the production line through a TSN network for execution.

[0073] The paper bag production line collaborative control method and system of the present application have the following technical effects and advantages:

[0074] The present application synchronously acquires paper bag point cloud, thermal imaging data and process parameters, extracts a multi-dimensional feature vector containing geometric features and material features, generates a collision-free path considering obstacle avoidance and material constraints based on an improved RRT*-Connect algorithm, dynamically switches high-speed mode and high-precision mode to optimize the trajectory according to material characteristics, and finally realizes microsecond-level instruction distribution and multi-device collaborative execution through a TSN network; the present application can effectively reduce paper bag damage rate, reduce path planning time consumption and production line beat synchronization error, significantly improve the stability and efficiency of flexible material production, and reduce process waste caused by neglecting material characteristics. BRIEF DESCRIPTION OF DRAWINGS

[0075] Figure 1 The paper bag production line collaborative control method flowchart of the embodiment 1 of the present application;

[0076] Figure 2 The data flow direction schematic diagram of the embodiment 1 of the present application;

[0077] Figure 3 The equipment task conflict priority dynamic allocation method schematic diagram of the embodiment 2 of the present application;

[0078] Figure 4 The paper bag production line collaborative control system structure diagram of the embodiment 3 of the present application. DETAILED DESCRIPTION

[0079] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0080] Embodiment 1

[0081] Please refer to Figure 1 、 Figure 2 , the embodiment provides a paper bag production line collaborative control method, including the following steps:

[0082] Synchronously collecting paper bag data and process parameters, extracting features of the paper bag data to obtain paper bag features; splicing the paper bag features and the process parameters to obtain a multi-dimensional feature vector;

[0083] Synchronously collecting paper bag data and process parameters, and extracting features of the paper bag data, the flexible material properties of the paper bag (such as temperature sensitive deformation, material tensile strength) and process constraints (such as grabbing force, motion acceleration limit) can be converted into a multi-dimensional feature vector recognizable by a computer. The vector is used as the input of the path planning algorithm, so that the improved RRT*-Connect algorithm can break through the fixed rule limit of the traditional geometric shape priority, dynamically adjust the path curvature, motion speed and control mode (such as high speed mode or high precision mode) according to the real-time state of the paper bag (such as high temperature softening area, low grammage easy to break characteristic), and avoid the rigid planning strategy of "one size fits all". For example, the temperature sensitive laminated paper bag automatically reduces the motion speed and increases the path buffer segment, and the high grammage kraft paper allows faster transportation speed, thereby significantly reducing the damage problems such as paper bag breakage and uneven stress caused by unreasonable path planning, reducing material waste rate, and improving the self-adaptive ability of the production line to flexible materials.

[0084] The paper bag data includes paper bag point cloud data and thermal imaging data.

[0085] The paper bag point cloud data is obtained by a binocular camera (such as Basler ace 2) to acquire RGB-D images, is generated using an Open3D library, and is applied with a bilateral filter to remove noise and retain key geometric features such as crease lines and handle holes; the thermal imaging data is obtained by an infrared thermal imager (such as FLIR A35), a temperature-material mapping relationship is established according to the thermal imaging data (such as an infrared image), the thermal imaging data is converted, and temperature (such as kraft paper: surface temperature gradient less than 0.3℃ / cm 2 ; PLA coated paper: temperature gradient greater than 0.8℃ / cm 2 (in view of the poor thermal conductivity of PLA)): the binocular camera and the infrared thermal imager can be installed 50 cm above the paper roller inlet, with a visual angle covering the full-width paper strip (width less than or equal to 1.2 m).

[0086] The temperature detector (i.e. the infrared thermal imager) is calibrated by a black body, a fitting relationship between pixel output value and black body temperature is established by Planck's law; for the measured material, the real-time pixel output value of the thermal imaging data is measured while maintaining the same radiation power, the black body temperature is calculated based on the fitting relationship, and the material emissivity is obtained according to the ratio of the black body temperature to the real temperature of the material (which can be obtained by a contact type temperature detector);

[0087] According to Planck's law, a relationship model between radiance and temperature is established, an inversion model about radiance is built according to the received radiance, the environmental reflected radiation and the material emissivity; the inversion model is solved to obtain the temperature; wherein the environmental reflected radiation is calculated according to the material emissivity, the real temperature of the paper bag and the environmental temperature.

[0088] The point cloud data of the paper bag is collected by a binocular camera and processed by bilateral filtering using the Open3D library, which can effectively retain key geometric features such as crease lines and handle holes, and avoid damage caused by traditional path planning due to neglecting the flexible structure features of the paper bag (such as weak parts) during grabbing or carrying. At the same time, the thermal imaging data collected by the infrared thermal imager is processed by blackbody calibration, emissivity calculation and radiation brightness inversion model, which can accurately establish the temperature-material mapping relationship (such as the temperature gradient difference between kraft paper and PLA coated paper), so that the path planning algorithm can dynamically adjust the motion parameters (such as reducing the moving speed in the high temperature area and optimizing the path to avoid the temperature gradient mutation area) according to the real-time temperature state of the paper bag (such as the softening degree of the high temperature area). This dual-data source scheme that combines geometric features and temperature features breaks through the limitations of traditional path planning relying only on geometric shapes, realizes the prediction and avoidance of the flexible deformation risk of the paper bag (such as thermal stress concentration of temperature-sensitive materials), reduces the damage rate of the paper bag caused by neglecting the material characteristics, and provides data support for dynamic adjustment of control strategies (such as soft grabbing force and acceleration limit), which significantly improves the adaptability of the production line to flexible workpieces of multiple materials.

[0089] The process parameters include tension sensor data, paper feed roller speed, and mechanical arm joint angle.

[0090] The paper bag features include geometric features and material features.

[0091] The method for obtaining geometric features includes:

[0092] Detecting key points on the point cloud data to obtain ROI point cloud.

[0093] Analyzing the ROI point cloud to obtain the length, width, height, crease line position, and handle hole coordinates of the paper bag, and splicing to obtain the geometric features.

[0094] The method for obtaining ROI point cloud includes:

[0095] Using k-d tree search to obtain the k-neighborhood of each point.

[0096] Describing the density distribution of the local area by integral invariant.

[0097] Solving the second derivative of the integral invariant to obtain the Hessian matrix.

[0098] Performing eigenvalue decomposition on the Hessian matrix to obtain principal curvatures k1, k2, and k3, where k1 < k2 < k3.

[0099] Normalizing the principal curvature k3 to obtain the response function.

[0100] According to a preset filtering threshold (adjusted according to actual point cloud noise), the paper bag point cloud data is filtered, only points with a response function value greater than the preset filtering threshold are reserved, and an ROI point cloud is obtained;

[0101] The method for obtaining the length, width and height of the paper bag comprises:

[0102] The ROI point cloud is subjected to axis-aligned bounding box fitting, and the length (such as along the paper feeding direction), width (such as along the width perpendicular to the paper feeding direction) and height (such as the vertical direction) of the paper bag are calculated.

[0103] The method for obtaining the crease line position comprises:

[0104] The k-neighborhood of the ith point p i is obtained using k-d tree search;

[0105] The mean value and covariance matrix C of the neighborhood points of p i are calculated, C is subjected to eigenvalue decomposition, and the eigenvector corresponding to the minimum eigenvalue is obtained as the normal vector of p i ;

[0106] The normal vector of p i is compared with the normal vector of the nearest neighbor point p i+1 by using the included angle formula, points satisfying the condition that the included angle is greater than a preset included angle threshold (such as 60) are marked as crease line candidate points (generally, the included angle of the normal vectors of the flat area is less than 30, and the included angle of the normal vectors of the crease area is greater than 60), and the crease line candidate points are collected into a set P;

[0107] Two points p a and p b are randomly selected from P, a candidate straight line L is built, the distances of all points in P to L are calculated, points with a distance less than a preset distance threshold are divided into inner points, and the straight line with the most inner points is reserved as the crease line position.

[0108] The method for obtaining the handle hole coordinates comprises:

[0109] The actual number of points in the neighborhood and the maximum number of points in the neighborhood of each point are counted, the neighborhood occupancy rate of each point is calculated, and points with a neighborhood occupancy rate lower than a preset occupancy rate threshold are judged as hole edge points;

[0110] The hole edge points are fitted by using an implicit surface to generate a continuous surface model, and virtual point clouds inside the hole are filled (only used for auxiliary detection, and the original point cloud is not modified);

[0111] According to a closedness rule, holes satisfying the closedness are determined, and a closed region is obtained.

[0112] The closure rule includes that the circumference-to-diameter ratio of the point cloud ring is within a preset variation range (e.g., the circumference-to-diameter ratio of the point cloud ring is close to π (e.g., within a variation range of ±0.1)); and the point cloud occupancy rate in the ring is lower than a preset point cloud occupancy rate (e.g., 0.1);

[0113] The centroid of the point cloud in the closed region is calculated as an initial estimation of the center of the circle.

[0114] The hole edge point cloud is fitted by the least square method, the optimal circle parameters including the center and the radius are solved by minimizing the sum of squares of distances from points to the circle, and the center is taken as the handle hole coordinate.

[0115] The ROI point cloud is obtained through steps such as k-d tree search, integral invariant analysis and Hessian matrix eigenvalue decomposition, which can effectively filter noise and retain key geometric features such as crease lines and handle holes, avoid feature misjudgment caused by point cloud noise (e.g., misjudging the crease area as a plane), and provide accurate structural information (e.g., the location of the weak part of the paper bag) for path planning. The length, width and height data obtained by fitting the axis-aligned bounding box make the algorithm able to dynamically adjust the safety distance (e.g., the distance between the end of the mechanical arm and the edge of the paper bag is 5 mm) according to the actual size of the paper bag, avoiding collision or grabbing deviation caused by size estimation deviation. The crease line detection filters candidate points by the angle between normal vectors and fits a straight line, which can accurately locate the structure boundary prone to tearing (e.g., the angle between normal vectors in the crease area is greater than 60°), prompting the path planning to actively avoid the area or reduce the motion acceleration (e.g., the speed near the crease line is reduced by 30%). The acquisition of the handle hole coordinate is realized through hole edge point detection, implicit surface fitting and least square circle fitting, which ensures that the grabbing point of the mechanical arm is located at the center of the handle hole (the positioning error is less than or equal to 0.3 mm), avoiding uneven stress on the paper bag caused by grabbing position deviation (e.g., stress concentration tearing at the edge of the handle hole). The above-mentioned geometric feature extraction methods provide high-precision structural constraints (e.g., handle hole safety radius, crease line avoidance area) for the path planning algorithm, enabling the algorithm to break through the limitations of traditional rough matching of geometric shapes, implement differentiated planning strategies (e.g., deceleration, force control mode switching) for weak links (e.g., creases, handle hole edges) of the paper bag flexible structure, reduce the paper bag damage rate caused by insufficient recognition of geometric features, and significantly improve the adaptability of path planning to complex structures of flexible materials.

[0116] The method for obtaining material characteristics includes:

[0117] Global statistics of the thermal imaging data are calculated, including mean and variance;

[0118] The temperature gradient of a preset local area is calculated;

[0119] The point cloud normal vector of each point is obtained;

[0120] The local height standard deviation of a preset local area of the point cloud is calculated;

[0121] The mean value, variance, temperature gradient, point cloud normal vector and local height standard deviation are standardized and spliced into a feature vector, and the feature vector is taken as an input of a material classification model to obtain material characteristics.

[0122] By calculating the mean value, variance and temperature gradient of the thermal imaging data, the uniformity of the paper bag surface temperature distribution and the mutation area (such as the heat sealing high temperature area, the temperature gradient greater than 0.8℃ / cm 2 of the PLA film area) can be quantified, combined with the point cloud normal vector (reflecting the surface flatness) and the local height standard deviation (characterizing the roughness), the material characteristics (such as the low temperature gradient and high surface roughness of kraft paper, the high temperature gradient and low roughness of PLA film paper) can be accurately identified. After standardizing these features and splicing them into a feature vector and inputting them into a material classification model, the automatic identification of the paper bag material type (such as kraft paper, PLA film paper) and its state (such as high temperature softening, low temperature embrittlement) can be realized. The identification result directly acts on the path planning strategy: for the film paper with large temperature gradient, the high-precision mode (S-shaped curve, acceleration less than or equal to 1m / s 2 ) is automatically enabled to avoid damage caused by heat stress concentration due to high-speed motion; for the kraft paper with high surface roughness, higher grabbing force (3-5N) and motion speed (1.2m / s) are allowed, and the path curvature is dynamically adjusted according to the local height standard deviation (such as avoiding the crease area with a height mutation greater than 2mm). This classification and strategy adaptation based on multi-dimensional material characteristics upgrades the path planning from geometric shape unified processing to material characteristic differentiated control, reduces the damage rate caused by material misjudgment, and significantly improves the intelligent perception and precise control ability of the production line for multi-material flexible workpieces.

[0123] According to the multi-dimensional feature vector, a collision-free path is generated at the edge node based on the improved RRT*-Connect algorithm;

[0124] The method for obtaining a collision-free path comprises:

[0125] The starting point and the ending point of the production line robot arm are obtained, wherein the ending point is obtained according to the geometric characteristics and the reserved safety distance, such as superimposing a reserved safety distance of 0.02m on the basis of the handle hole coordinates;

[0126] A control mode is selected according to the material characteristics; for example, for non-film materials (such as kraft paper), position control is selected as the main mode, i.e. rigid control; for PLA / PE film materials, impedance control is selected as the main mode, i.e. compliant control (to avoid crushing the surface);

[0127] The state space of the robot arm is defined, and the state space is composed of a state vector, and the state vector includes the pose of the robot arm and a multi-dimensional feature vector;

[0128] Uniformly sample within the workspace of the robot arm, and the sample points satisfy feature constraints, including geometric constraints and process constraints, such as the sample points must be located within a preset height range (such as 0.1-0.3m) above the paper bag; if the speed of the paper roller V>1m / s, prefer to sample straight paths (reduce turns); if the tension T>30N, increase the distance between sample points (reduce path complexity);

[0129] Create a tree expanding from the start point and a tree expanding from the end point, marked as Ts and Tf, and each tree stores node information, including coordinates, parent nodes, and path cost;

[0130] In each iteration, first expand Ts, then expand Tf, until the distance between the nodes of the two trees is less than a preset connection threshold (such as 0.1m);

[0131] When the distance between the nearest nodes of Ts and Tf is less than the preset connection threshold, check if the path between the two points is collision-free, if so, merge the two trees to obtain the initial path, otherwise, rebuild Ts and Tf;

[0132] The method of expanding Ts includes:

[0133] Simultaneously expand from Ts and Tf, and each time select the nearest neighbor node q near , move a step length from the sample point q sam to generate a new node q new ;

[0134] Check if the path from q near to q new collides with the robot arm itself, static obstacles, or dynamic obstacles;

[0135] Combine the path length, motion smoothness, energy consumption, and feature constraint cost into a cost function, and calculate the cost of q near according to the cost function;

[0136] Add q near to Ts to obtain an updated Ts.

[0137] The improved RRT*-Connect algorithm based on multi-dimensional feature vectors realizes dynamic path planning of flexible paper bags at the edge node by fusing geometric features (such as handle hole coordinates + safety distance) and material features (such as rigid-flexible control mode driven by film type): according to the real-time state of the paper bag (such as temperature gradient, tension size), the sampling point range is constrained (such as being limited to the height of 0.1-0.3m above the paper bag, and the straight path is preferred to adapt to high-speed working conditions), and through the synchronous expansion of bidirectional trees (Ts / Tf) and collision detection (robotic arm / obstacle), it is ensured that the path avoids the weak area of the paper bag (such as the crease line, the high-temperature heat sealing area); the feature constraint cost (such as the acceleration penalty term of low grammage paper) is introduced into the cost function, so that the path meets the collision-free, motion-smooth and process-synchronous at the same time. This method upgrades the traditional geometric hard collision avoidance to multi-dimensional soft constraint optimization of material-geometry-process, so that the robotic arm automatically switches to soft control (reduces the stiffness of impedance control) when grabbing the film-coated paper bag, increases the path turning radius to reduce the centrifugal force damage, and finally reduces the damage rate of the paper bag caused by unreasonable path planning, while improving the switching efficiency of the production line for multiple types of paper bags.

[0138] The different modes in the dual-mode speed planning method are switched by a dynamic switching strategy to optimize the collision-free path to obtain an optimized trajectory.

[0139] The dual-mode speed planning method includes a high-speed mode and a high-precision mode.

[0140] The method for switching different modes in the dual-mode speed planning method by a dynamic switching strategy includes:

[0141] The path curvature is calculated, and when the path curvature exceeds a preset curvature threshold, the high-precision mode is switched to;

[0142] When the real-time change of the paper feed roller speed is detected, the relationship between the paper feed roller speed and the preset speed threshold is compared, and when the paper feed roller speed is greater than the preset speed threshold, the high-speed mode is selected; otherwise, the high-precision mode is selected;

[0143] When it is detected that the paper bag grammage is greater than a preset grammage threshold, the high-precision mode is forcibly enabled.

[0144] The method for optimizing the collision-free path by the high-speed mode includes:

[0145] Obtain a collision-free path, and obtain the first and last points and the midpoint of each path, as well as the midpoints of the previous and subsequent segments, to generate four control points of a cubic Bezier curve;

[0146] Build a Bezier curve equation according to the four control points;

[0147] When the robotic arm end moves along the curve, the centripetal acceleration satisfies a preset high-speed constraint condition;

[0148] The movement speed of the mechanical arm is controlled by a trapezoidal speed method; trapezoidal speeds of an acceleration section, a constant speed section and a deceleration section are preset, and proportions of the acceleration section, the constant speed section and the deceleration section satisfy preset proportions; wherein the acceleration section is obtained according to acceleration calculation, and the acceleration is not higher than the maximum acceleration of the mechanical arm; the constant speed section is obtained according to maximum centripetal acceleration calculation; and the deceleration section is obtained according to constant speed section speed, acceleration, acceleration time and constant speed time calculation.

[0149] The method for optimizing the collision-free path by the high-precision mode comprises:

[0150] The speed is divided into M stages by an S-shaped curve, as shown in Table 1, the speed is divided into seven stages, in turn, variable acceleration, uniform acceleration, deceleration, uniform speed, acceleration and deceleration, uniform deceleration, and deceleration;

[0151] Table 1 S-shaped curve planning

[0152] Phase Physical meaning Jerk (change of acceleration) Acceleration a Velocity v 1 Change of acceleration jmax Linear increase from 0 to amax Rising convex curve 2 Uniform acceleration 0 Hold amax Straight line rising 3 Deceleration - jmax Linear decrease from amax to 0 Rising convex curve 4 Uniform speed 0 0 Hold vmax 5 Acceleration and deceleration - jmax Linear decrease from 0 to -amax Falling convex curve 6 Uniform deceleration 0 Hold -amax Straight line falling 7 Deceleration and acceleration jmax Linear increase from -amax to 0 Falling convex curve

[0153] The total time of the mechanical arm is minimized, that is, the sum of the times of the M stages, and the preset high-precision constraint conditions, such as jerk constraint, acceleration constraint and displacement constraint, are satisfied, and the optimized trajectory is obtained by a natural heuristic optimization algorithm.

[0154] The dual-mode speed planning method realized by the dynamic switching strategy constructs an efficiency-precision adaptive adjustment mechanism for the flexible material characteristics of the paper bag and the dynamic working conditions of the production line: when the speed of the paper feeding roller is greater than a preset speed threshold or the path curvature is small, the high-speed mode (Bezier curve + trapezoidal speed) is enabled, the acceleration section, the constant speed section and the deceleration section are distributed by a preset proportion, the centripetal acceleration is maximized, the carrying efficiency is improved, and the rapid transportation demand of the high-gra paper bag is adapted; and when it is detected that the paper bag weight is greater than a preset weight threshold, the path curvature exceeds a preset curvature threshold, or the speed of the paper feeding roller is not greater than a preset speed threshold, the high-precision mode is automatically switched to, and the variable acceleration and the acceleration can also be optimized by a natural heuristic algorithm, the vibration amplitude of the end of the mechanical arm is reduced, and tearing or surface crushing of the low-gra paper or the film-coated paper caused by sudden speed change is avoided. The strategy breaks through the limitations of the traditional “one-size-fits-all” speed control, automatically prolongs the acceleration and deceleration time when the mechanical arm grabs the PLA film-coated paper, controls the path smoothness error within a low range, cooperates with the compliant control mode, reduces the damage rate of the flexible paper bag caused by movement impact, reduces the production line beat synchronization error through the speed-mode linkage mechanism, and significantly improves the flexible adaptation capability of multi-variety paper bag production.

[0155] If it is detected that the paper bag offset is greater than a preset offset threshold, a dynamic re-planning mechanism of the dynamic switching strategy is triggered, the path is corrected, and the grabbing force is adjusted in real time through the joint torque feedback of the mechanical arm.

[0156] The optimized trajectory is split into discrete control instructions and distributed to each control device in the production line (such as a robot controller, a paper feeding roller servo driver, and a die cutting machine punch timing controller, etc.) through a TSN network for execution.

[0157] Embodiment 2

[0158] Referring to Figure 3 The embodiment provides a priority dynamic allocation method for device task conflict in the device of embodiment 1, comprising the following steps:

[0159] Obtain the real-time state of the task, including the task progress, the remaining time, the estimated total time consumption, and the task deadline;

[0160] Calculate the remaining time and the urgency of the conflicting device task, the remaining time being the difference between the task deadline and the current time; the urgency being the ratio of the product of the remaining time and the task importance factor to the product of the estimated total time consumption and the device priority factor; wherein the task importance factor and the device priority factor are preset values according to the task type priority (such as packaging > carrying > sorting, then the packaging task importance factor value is 1.2; the carrying task importance factor value is 1.0; the sorting task importance factor value is 0.8) or the device type priority (such as robot > AGV > conveyor belt, then the robot device priority factor value is 0.9; the AGV device priority factor value is 1.0; the conveyor belt device priority factor value is 1.1); the lower the urgency, the more urgent the task, which needs to be executed first; by introducing the task importance factor and the device priority factor, the situation of equal urgency of different types of tasks can be avoided, and the most urgent task can be directly selected for execution according to the urgency, further improving the control efficiency of the production line.

[0161] When the device task conflicts, the device with the smallest urgency is selected for task execution according to the urgency of the device.

[0162] Embodiment 3

[0163] Referring to Figure 4 The embodiment provides a paper bag production line collaborative control system, comprising:

[0164] The acquisition and processing module synchronously acquires paper bag data and process parameters, extracts features from the paper bag data to obtain paper bag features, and splices the paper bag features and the process parameters to obtain a multi-dimensional feature vector;

[0165] The trajectory planning module generates a collision-free path at the edge node based on the improved RRT*-Connect algorithm according to the multi-dimensional feature vector;

[0166] A path optimization module: different modes in the dual-mode speed planning method are switched through a dynamic switching strategy to optimize a collision-free path to obtain an optimized trajectory;

[0167] An instruction execution module: the optimized trajectory is split into discrete control instructions, which are distributed to each control device in the production line through a TSN network for execution.

[0168] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0169] Finally: the above is only a preferred embodiment of the present application and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included within the protection scope of the present application.

Claims

1. A collaborative control method for a paper bag production line, characterized in that, Includes the following steps: Simultaneously collect paper bag data and process parameters, extract features from the paper bag data to obtain paper bag features; then concatenate the paper bag features and process parameters to obtain a multi-dimensional feature vector; Based on multidimensional feature vectors and a path planning algorithm, collision-free paths are generated at edge nodes. The optimized trajectory is obtained by switching between different modes in the dual-mode velocity planning method using a dynamic switching strategy to optimize the collision-free path. The optimized trajectory is broken down into discrete control commands, which are then distributed to various control devices on the production line for execution via a communication network.

2. The collaborative control method for a paper bag production line according to claim 1, characterized in that, Paper bag data includes paper bag point cloud data and thermal imaging data; process parameters include tension sensor data, paper feed roller speed and robotic arm joint angle; A temperature-material mapping relationship is established based on thermal imaging data, and the thermal imaging data is converted to obtain the temperature. Methods for obtaining temperature include: By calibrating a temperature detector with a blackbody, a fitting relationship between pixel output value and blackbody temperature is established using Planck's law. For the material under test, the same radiation power is maintained, and the real-time pixel output value of thermal imaging data is measured. The blackbody temperature is calculated based on the fitting relationship, and the emissivity of the material is obtained according to the ratio of the blackbody temperature to the material's true temperature. Based on Planck's law, a model relating radiance to temperature is established. An inversion model of radiance is constructed based on received radiance, ambient reflected radiation, and material emissivity. The temperature is obtained by solving the inversion model.

3. The collaborative control method for a paper bag production line according to claim 1, characterized in that, Paper bag characteristics include geometric features and material features; Methods for obtaining geometric features include: Perform key point detection on point cloud data to obtain ROI point cloud; The point cloud of the ROI is analyzed to obtain the length, width, height, crease line position and handle hole coordinates of the paper bag, and the geometric features are obtained by splicing them together. Methods for obtaining ROI point clouds include: Use a kd-tree search to obtain the k-neighborhood of each point; The density distribution of a local region is described by integral invariants; Solve for the second derivative of the integral invariant to obtain the Hessian matrix; Eigenvalue decomposition of the Hessian matrix yields principal curvatures k1, k2, and k3, where k1 <k2<k3; The principal curvature k3 is normalized to obtain the response function; The paper bag point cloud data is filtered according to a preset screening threshold, and only points with response function values ​​greater than the preset screening threshold are retained to obtain the ROI point cloud; Methods for obtaining the length, width, and height of a paper bag include: Axial-aligned bounding box fitting is performed on the ROI point cloud to calculate the length, width, and height of the paper bag.

4. The collaborative control method for a paper bag production line according to claim 3, characterized in that, Methods for obtaining the location of crease lines include: Use kd-tree search to obtain the i-th point p. i k-neighborhood; Calculate p i Given the mean and covariance matrix C of the neighborhood points, perform eigenvalue decomposition on C and obtain the eigenvector corresponding to the smallest eigenvalue as p. i The normal vector; Compare p using the included angle formula i With nearest neighbor p i+1 The normal vector is used to mark points that satisfy the angle greater than a preset angle threshold as candidate points of crease lines, and the candidate points of crease lines are summarized into set P. Randomly select two points p from P. a p b Construct candidate straight lines L, calculate the distance from all points in P to L, divide points whose distance is less than a preset distance threshold into inliers, and retain the straight line with the most inliers as the crease line position.

5. The collaborative control method for a paper bag production line according to claim 3, characterized in that, Methods for obtaining the coordinates of the handle hole include: Count the actual number of points in the neighborhood of each point and the maximum number of points in the neighborhood, and calculate the neighborhood occupancy rate of each point; points with a neighborhood occupancy rate lower than a preset occupancy rate threshold are identified as hole edge points; By implicitly fitting the edge points of the hole, a continuous surface model is generated to fill the virtual point cloud inside the hole; Based on the closure rule, determine the holes that satisfy the closure property, and obtain the closed region; The closure rule includes that the ratio of the perimeter to the diameter of the point cloud ring is within a preset range; and the point cloud occupancy rate within the ring is lower than the preset point cloud occupancy rate. Calculate the centroid of the point cloud of the closed region as an initial estimate of the circle's center; The point cloud of the hole edge is fitted by least squares method, and the optimal circle parameters, including the center and radius, are solved by minimizing the sum of squared distances from the point to the circle. The center of the circle is then used as the coordinates of the handle hole.

6. The collaborative control method for a paper bag production line according to claim 1, characterized in that, Methods for obtaining material characteristics include: Calculate global statistics for thermal imaging data, including mean and variance; Calculate the temperature gradient of a preset local region; Obtain the point cloud normal vector for each point; Calculate the local height standard deviation of a preset local region in a point cloud; The mean, variance, temperature gradient, point cloud normal vector, and local height standard deviation are standardized and then concatenated into a feature vector. This feature vector is then used as input to the material classification model to obtain material features.

7. The collaborative control method for a paper bag production line according to claim 1, characterized in that, Methods for obtaining collision-free paths include: Obtain the start and end points of the robotic arm on the production line, where the end point is determined based on geometric features and a reserved safety distance; Select the control mode based on the material characteristics; Define the state space of the robotic arm, which consists of state vectors, including the robotic arm pose and multi-dimensional feature vectors. Uniform sampling is performed within the workspace of the robotic arm, and the sampling points satisfy feature constraints, including geometric constraints and process constraints. Create trees that expand from the starting point and the ending point, respectively, labeled Ts and Tf. Each tree stores node information, including coordinates, parent node, and path cost. In each iteration, Ts is expanded first, then Tf is expanded, until the distance between the nodes of the two trees is less than the preset connection threshold. If the distance between the nearest nodes of Ts and Tf is less than the preset connection threshold, check if there is no collision in the path between the two points. If so, merge the two trees to obtain the initial path; otherwise, reconstruct Ts and Tf.

8. The collaborative control method for a paper bag production line according to claim 1, characterized in that, The dual-mode speed planning method includes a high-speed mode and a high-precision mode; Methods for switching between different modes in a dual-mode velocity planning approach using dynamic switching strategies include: Calculate the path curvature, and switch to high-precision mode when the path curvature exceeds the preset curvature threshold; When a real-time change in the paper feed roller speed is detected, the relationship between the paper feed roller speed and the preset speed threshold is compared. If the paper feed roller speed is greater than the preset speed threshold, the high-speed mode is selected; otherwise, the high-precision mode is selected. When the detected weight of the paper bag exceeds the preset weight threshold, the high-precision mode is forcibly activated.

9. The collaborative control method for a paper bag production line according to claim 8, characterized in that, Methods for optimizing collision-free paths using high-speed mode include: Obtain a collision-free path, and for each path segment, take the first and last points, the midpoint, and the midpoints of the preceding and following segments to generate four control points for the cubic Bézier curve. The equation of the Bézier curve is constructed based on four control points; When the end effector of the robotic arm moves along the curve, the centripetal acceleration satisfies the preset high-speed constraint condition; The movement speed of the robotic arm is controlled by the trapezoidal velocity method; the trapezoidal velocities of the acceleration segment, constant speed segment, and deceleration segment are preset, and the proportions of the acceleration segment, constant speed segment, and deceleration segment meet the preset ratio; among them, the acceleration segment is obtained based on the acceleration, and the acceleration does not exceed the maximum acceleration of the robotic arm; the constant speed segment is obtained based on the maximum centripetal acceleration; the deceleration segment is obtained based on the constant speed segment speed, acceleration, acceleration time, and constant speed time. Methods for optimizing collision-free paths using high-precision models include: The speed is divided into M stages using an S-shaped curve; Minimize the total running time of the robotic arm, i.e. the sum of the times of M stages, while satisfying the preset high-precision constraints, and obtain the optimized trajectory by solving the problem through a natural heuristic optimization algorithm.

10. A collaborative control system for a paper bag production line, implementing the collaborative control method for a paper bag production line according to any one of claims 1-9, characterized in that, include: Data Acquisition and Processing Module: Synchronously acquires paper bag data and process parameters, extracts features from the paper bag data, and obtains paper bag features; By splicing together the characteristics of paper bags and process parameters, a multi-dimensional feature vector is obtained; Trajectory planning module: Based on multi-dimensional feature vectors and path planning algorithms, it generates collision-free paths at edge nodes; Path optimization module: Optimizes collision-free paths to obtain optimized trajectories by dynamically switching between different modes in the dual-mode speed planning method using a switching strategy; Instruction execution module: It breaks down the optimized trajectory into discrete control instructions, which are then distributed to various control devices on the production line for execution via a communication network.

Images