Multi-robot parallel processing scheduling optimization method for local tasks of large cabin

Abstract

The invention discloses a multi-robot parallel processing scheduling optimization method for local tasks of a large cabin, solves the problems of complex robot solving process and low processing efficiency of the existing multi-robot parallel processing method, and belongs to an optimization scheduling technology in robot application in the manufacturing industry. The method comprises the following steps: constructing a multi-robot parallel processing system of a large cabin; target functions U1 and U2 are established, the target function U1 is a load difference value minimization function of robots provided with the same kind of end effectors in the machining system, and the target function U2 is a minimization function of the maximum time of machining all tasks; based on the U1 and the minimum rotation number K of the cabin body, a cabin body rotation angle set and a unique task set of stations on the two sides of the cabin body at each angle are obtained; and a plane mapping method is introduced to eliminate the collision risk of the robot and the cabin body component, a mapping point corresponding to each task is obtained, the mapping points of the tasks in the unique task set are sorted based on the target function U2, and a collision-free task processing sequence is generated.

Description

Technical Field

[0001] The present invention relates to a multi-robot parallel processing scheduling optimization method for local tasks of large cabins, belonging to the optimization scheduling technology in the application of manufacturing robots. Background Art

[0002] As a key component of spacecraft, the large cylindrical thin-walled cabin has metal brackets of various shapes welded on its outer surface for assembling various instruments and equipment. To ensure the performance and reliability of the spacecraft, the dimensional and positional accuracies of the bracket surfaces must meet strict standards. With the continuous diversification of spacecraft functions, the differences in the number, type, and distribution position of brackets make the processing tasks of different types of cabins highly complex. The traditional split offline manufacturing mode has obvious deficiencies in terms of efficiency and processing quality stability, and it is difficult to meet complex manufacturing requirements. Therefore, as a new manufacturing mode for the future, the multi-robot parallel processing system provides a high-efficiency, high-precision, and safe solution for the combined manufacturing of large cabin bracket tasks. In this mode, the cabin components are finish-machined on the basis of the bracket features that have been rough-machined. By installing rotary deflectors on both sides of the cabin, the cabin components are rotated multiple times to bring the bracket features into the processing range of the robots, thereby achieving precise machining.

[0003] However, the multi-robot parallel processing scheduling planning for large cabin brackets faces many challenges. The rotatability of the cabin, the uneven distribution of brackets, and the omnidirectional movement characteristics of the robot system significantly increase the solution search space for task allocation and sequencing, resulting in an exponential growth in the complexity of the solution process and making it difficult to solve effectively. The existing problems of multi-robot task allocation, path planning, and scheduling sequence are difficult to decouple, with numerous optimization objectives that restrict each other, and it is impossible to effectively ensure the load balance of the robots, resulting in problems such as excessive idle time of the robots and low system efficiency. Therefore, how to reduce the complexity of solving the scheduling problem and find an efficient and accurate multi-robot scheduling planning method has become an important technical problem currently faced.

[0004] To solve this problem, researchers have proposed various solutions based on optimization methods, such as game theory methods, deep reinforcement learning, artificial intelligence, etc. However, most of these methods face challenges such as high computational costs and complex solution processes, and it is difficult to quickly and accurately handle the multi-robot task scheduling problem. Especially for the scheduling planning of multi-robot parallel processing of large cabin brackets, it has numerous constraint conditions, conflicting objectives, high computational resource requirements, and high solution difficulty. Summary of the Invention

[0005] Aiming at the problems of complex robot solution process and low processing efficiency in the existing multi-robot parallel processing method, the present invention provides a multi-robot parallel processing scheduling optimization method for local tasks of large cabins.

[0006] A multi-robot parallel processing scheduling optimization method for local tasks of large cabins according to the present invention includes:

[0007] S1. Construct a multi-robot parallel processing system for a large cabin;

[0008] S2. Establish objective functions U1 and U2. The objective function U1 is a function that minimizes the load difference of robots equipped with the same type of end effector in the multi-robot parallel processing system, and the objective function U2 is a function that minimizes the maximum time for the multi-robot parallel processing system to process all tasks;

[0009] S3. Based on the objective function U1 and the minimum number of cabin rotations K, obtain the set A of cabin rotation angles K ; The cabin rotation angle α k ∈A K , k ∈ [1, K], and obtain the unique task set CS p (α k ) and CS n (αk) of the workstations on both sides of the cabin at each cabin rotation angle;

[0010] S4. On the outer surface of the cabin, with the plane of the task to be processed farthest from the cabin as the reference, expand a set distance outward as the radius and construct a cylinder with the center of the cabin as the center of the circle. The outer surface of the cylinder is used as the collision-free surface; after the workstations at each cabin rotation angle are fixed, a vertical reference plane RP αk is established at the intersection of the collision-free surface and the Xc axis, and the center points cp p (α k ) and CS n (α k ) of the task planes in the unique task set CS i are projected along the normal direction of the task plane to the reference plane RP αk to obtain the mapped point cp i ' of each task; Based on the objective function U2, the mapped points cp p (α k ) and CS n (α k ) in the unique task set CS i ' are sorted to obtain the task processing sequences CS p (α k ) and CS n (αk) of the workstations on both sides at each cabin rotation angle;

[0011] In the cabin coordinate system, the axial direction of the cabin is the Y C axis, the direction perpendicular to the horizontal plane and upward is the Z C axis, and the X CThe axis is in the radial direction of the cabin body.

[0012] Preferably, the objective function U1 is:

[0013]

[0014] In the formula, represents the total time of all tasks corresponding to the processes assigned to the positive side of the X-axis of the end effector in the multi-robot parallel machining system when the cabin body rotates by an angle α k ; C represents the total time of all tasks corresponding to the processes assigned to the negative side of the X-axis of the end effector of the same type when the cabin body rotates by an angle α ; k ; C ;

[0015] Preferably, the objective function U2 is:

[0016]

[0017] In the formula, TC(CS k , α k ) represents the operation time span of all assigned tasks when the cabin body rotates by an angle α k ; TC(WS k ) represents the non-operation time when the cabin body rotates from the cabin body rotation angle α k to the cabin body rotation angle α k+1 ;

[0018] Preferably, the number of rotations K of the cabin body satisfies K≥K min , and the acquisition method of K min is as follows:

[0019] Taking the robot dexterity index D=(1 / k F (J(θ)))×100% as the evaluation index, determine the angle range of each joint of the robot, where J(θ) is the robot Jacobian matrix, and k F (J(θ)) is the condition number of the robot Jacobian matrix;

[0020] Based on the angle range of each joint of the robot, use the Monte Carlo method to solve all the positions of the tip point of the end effector, and define the range enveloped by all these positions as the effective machining space A E of the robot;

[0021] Based on the effective machining space A E of the robot, adjust the horizontal distance d CB between the cabin body and the robot positioning point so that the range of the outer surface of the cabin body covered within the effective machining space of the robot is the largest, and solve the central angle αE of the cabin body corresponding to the maximum coverage range;

[0022] The minimum rotation threshold K of the cabin min = 2*π / α E .

[0023] Preferably, S3 includes:

[0024] Set the step size α of the cabin rotation angle unit Discretize the cabin to determine the initial angle set α of the cabin rotation orig = {1:α unit :π}, assuming the cabin rotates to any angle α u ∈α orig , then the positioning angles of the two sides of the cabin along the axis are α u and α u +π respectively; according to the pose parameters of the task to be processed, obtain the maximum task sets to be processed on both sides of the cabin when the cabin rotates to any angle α u as CS p-max (α u ), CS n-max (α u );

[0025] When the number of cabin rotations is not less than the rotation threshold K min , with the goal of minimizing the number of cabin rotations and U1, all TC u (α p ), TC u (α n ) and the set of △TC(α u ) corresponding to the solution search space, determine the number of cabin rotations K and the angle set α u ;

[0026] TC K p (α u n ) is the total time of all tasks corresponding to the processes assigned to the end effector on the positive side of the X u axis of the cabin when the cabin rotates at the angle α C , TC n (α u ) is the total time of all tasks corresponding to the processes assigned to the same type of end effector on the negative side of the X u axis of the cabin when the cabin rotates at the angle α C ; △TC(α u ) = TC p (α u ) - TC n (α u );

[0027] The cabin rotation angle α k ∈AK , where \(k\in[1, K]\), and the maximum task set to be processed is \(CS\) p-max (\(\alpha\) k ), \(CS\) n-max (\(\alpha\) k );

[0028] Decompose the maximum task set to be processed \(CS\) p-max (\(\alpha\) k ) and \(CS\) n-max (\(\alpha\) k ) into a fixed task set \(CS\) p-fix (\(\alpha\) k ), \(CS\) n-fix (\(\alpha\) k ) and a flexible task set \(CS\) p-flex (\(\alpha\) k ), \(CS\) n-flex (\(\alpha\) k ); Each task in the fixed task set corresponds to a unique cabin rotation angle, and each task in the flexible task set corresponds to multiple cabin rotation angles;

[0029] Obtain the tasks in the flexible task set \(CS\) p-flex (\(\alpha\) k ), \(CS\) n-flex (\(\alpha\) k ) and divide them into the cabin rotation angles with a high density, and obtain the unique task set \(CS\) p (\(\alpha\) k ), \(CS\) n (\(\alpha\) k ) of the workstations on both sides of the cabin at each rotation angle.

[0030] Preferably, the motion mode of the end effector based on the collision-free surface is:

[0031] The robot tool center point moves within the reference plane \(RP\) αk to the mapping point \(cp'\) i corresponding to the center point \(cp\) i of the task plane, and then moves in the reverse direction along the normal direction of the task plane to the center point \(cp\) i of the task plane as the feed position point to start processing. After processing, return from the origin of the end effector coordinate system to within the reference plane \(RP\) αk , and move within the reference plane \(RP\) αk to the mapping point \(cp'\) i corresponding to the center point \(cp\) i+1 of the next task plane to perform the processing of the next task.

[0032] Preferably, in S4, based on the objective function \(U2\), use the genetic algorithm for the unique task set \(CS\) p (\(\alpha\) k)、CS n (α k ) The mapping point cp of the task in i ' is sorted to obtain the task processing sequence of the opposite side station of the cabin: CS p (α k ) = {CS kp,1 , CS kp,2 , …, CS kp,nk1}, CS n (α k ) = {CS kn,1 , CS kn,2 , …, CS kn,nk2}, and then the allocation of all tasks to be processed and the processing order of multiple robots are completed. nk1 and nk2 respectively represent the number of tasks in the unique task sets CS p (α k ) and CS n (α k ) on both sides of the cabin at each rotation angle.

[0033] Preferably, S1 includes:

[0034] Obtain data of the large cabin of the spacecraft and the mobile robot processing system;

[0035] According to the obtained data, construct a multi-robot parallel processing system for the large cabin, and determine the unit composition, multi-robot layout and movement path of the multi-robot parallel processing system.

[0036] Advantages of the present invention: The present invention is a two-step scheduling optimization method. Through hierarchical division, different evaluation indexes are gradually optimized, and the scheduling problem of the multi-machine parallel processing of the large cabin task is simplified. In the task allocation stage of the multi-station and multi-robot, the reachability constraint is eliminated, and a task secondary allocation method that simultaneously considers the minimum rotation threshold and the difference in robot task loads is adopted, which can achieve the maximum balance of the workloads of similar robots. In the task sorting stage, a plane mapping method is adopted to eliminate the collision risk between the robot and the cabin components during the processing. Further, a genetic algorithm is used to optimize the sorting of the mapped tasks with priority constraints, which can significantly improve the production efficiency of the multi-machine parallel processing system. Description of the Drawings

[0037] Figure 1 Is the flow chart of the multi-robot parallel processing scheduling optimization method in the present invention;

[0038] Figure 2 Is the schematic diagram of the unit composition of the multi-robot parallel processing system in the present invention;

[0039] Figure 3Schematic diagram of multi-robot layout and movement path in the present invention;

[0040] Figure 4 Effective processing space of the robot and its positional relationship with the cabin in the present invention;

[0041] Figure 5 Schematic diagram of the total milling and drilling time of all tasks at each discrete rotating station in the present invention;

[0042] Figure 6 Schematic diagram of the total time difference between the two sides of the cabin at the same angle in the present invention;

[0043] Figure 7 Schematic diagram of the multi-task mapping method based on a plane at a single station in the present invention;

[0044] Figure 8 Gantt chart of multi-machine parallel system scheduling and sorting for the cabin case in the present invention. Detailed implementation manners

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0047] Next, the present invention will be further described in conjunction with the accompanying drawings and specific embodiments, but it is not a limitation of the present invention.

[0048] The multi-robot parallel processing scheduling optimization method for local tasks of large cabins in this embodiment includes:

[0049] Step 1: According to the actual scenario, input the three-dimensional models of the large cabin of the spacecraft and the mobile robot processing system, and obtain the cabin data, the bracket task data on the outer surface of the cabin, and the size data of the mobile robot processing system;

[0050] Among them, the cabin data includes the cabin length L C and the cabin diameter size parameter D C , and there are n discrete brackets to be processed distributed on the outer surface of the cabin;

[0051] The bracket task data includes the number and pose information of the brackets, and the set and number are CS = {1, 2, 3,..., n}; the pose information PA of the i-th bracket i = {x i , yi , z i , x i-cen , y i-cen , z i-cen , β i-cen , h i},where \(i = 1, 2, \ldots, n\), and among them \(x\) i , y i , z i , represent the spatial position coordinates of the center of the outer surface of the \(i\)-th bracket to be processed, \(x\) i-cen , y i-cen , z i-cen represent the spatial position coordinates of the center point of the connection surface between the \(i\)-th bracket and the outer surface of the cabin, \(β\) i-cen represents the rotation angle of the normal vector of the outer surface of the \(i\)-th bracket relative to the horizontal plane where the cabin is located, \(h\) i is the height of the \(i\)-th bracket; the pose information of the processed bracket will change with the rotation of the cabin;

[0052] The dimensional data of the mobile robot processing system includes the robot model and the overall length dimension \(L\) along the axial direction of the cabin MR ;

[0053] Step 2, construct a multi-robot parallel processing system for a large cabin:

[0054] Based on the obtained cabin data \(L\) C , \(D\) C and the dimensional data \(L\) of the mobile robot processing system MR , determine the type \(q\) of the robot end effector; the set of the multi-robot processing system is \(R\), \(R = \{R1, R2, \ldots, R\) q},\(R\) q represents the set of robots equipped with the \(q\)-th type of end effector, and the number of each type of robot processing system is \(m\) q , the types of robot end effectors, tool models and specifications in the multi-robot parallel processing system are determined by the technological requirements of the spacecraft cabin bracket task, while the number of each type of robot end effector is determined by the size of the spacecraft cabin and the size of the mobile robot processing system.

[0055] Determine the unit composition of the multi-robot parallel processing system, and set the specific layout and movement paths of the multi-robots;

[0056] Step 3, establish the objective functions \(U1\) and \(U2\). The objective function \(U1\) is the function that minimizes the load difference of the robots equipped with the same type of end effector in the multi-robot parallel processing system, and the objective function \(U2\) is the function that minimizes the maximum time for the multi-robot parallel processing system to process all tasks;

[0057]

[0058] Represents the working time deviation of the same type of robots on both sides of the cabin at the rotation station α k when the cabin rotates, Indicates the total time of all tasks corresponding to the processes assigned to the end effector on the positive side of the X-axis of the cabin in the multi-robot parallel processing system when the rotation angle of the cabin is α k when the cabin rotates, C The total time of all tasks corresponding to the processes assigned to the end effector on the positive side of the X-axis of the cabin in the multi-robot parallel processing system when the rotation angle of the cabin is α Indicates the total time of all tasks corresponding to the processes assigned to the end effector on the negative side of the X-axis of the cabin in the multi-robot parallel processing system when the rotation angle of the cabin is α k when the cabin rotates, C The total time of all tasks corresponding to the processes assigned to the end effector on the negative side of the X-axis of the cabin in the multi-robot parallel processing system when the rotation angle of the cabin is α; In the cabin coordinate system, the axial direction of the cabin is the Y-axis, the direction perpendicular to the horizontal plane upward is the Z-axis, and the X-axis is the radial direction of the cabin. C The axial direction of the cabin is the Y-axis, the direction perpendicular to the horizontal plane upward is the Z-axis, and the X-axis is the radial direction of the cabin. C axis, and the X C axis is the radial direction of the cabin.

[0059] The objective function U2 is:

[0060]

[0061] In the formula, TC(αk) represents the operation time span of the cabin rotation angle αk, and TC(CS k ,α k ) represents the operation time span of all assigned tasks when the cabin rotation angle is α k , and TC(WS k ) represents the non-operation time when the cabin rotates from the cabin rotation angle α k to the cabin rotation angle α k+1 .

[0062] Decouple multi-type constraint conditions hierarchically, and divide various constraint conditions into processing layer constraint conditions and process planning layer constraint conditions; among them, the processing aspect includes robot body performance constraints, speed constraints, and collision-free constraints; while the process planning aspect includes multi-process priority constraints, processing task allocation constraints, envelope constraints, and minimum rotation times constraints.

[0063] Use various constraint conditions in batches for the optimization of the two-step scheduling method to reduce the complexity of the multi-robot parallel processing scheduling problem. In the balanced task allocation under multiple rotation stations, the constraint conditions used in the optimization process include: robot body performance constraints, speed constraints in the processing aspect, and processing task allocation constraints and minimum cabin rotation times constraints in the process planning aspect. In the processing sequence planning of multi-robot multi-tasks, the constraint conditions used in the optimization process include: collision-free constraints in the processing aspect, and multi-process priority constraints and envelope constraints in the process planning aspect.

[0064] Step 4. Balancing task allocation under multiple rotating stations: Based on the optimization objective U1 and minimizing the number of times the cabin rotates K, obtain the set A of cabin rotation angles K ; the cabin rotation angle α k ∈A K , k ∈ [1, K], and obtain the unique task set CS of the workstations on both sides of the cabin at each cabin rotation angle p (α k )、CS n (α k ); specifically including:

[0065] Step 41. Obtain the minimum rotation threshold K min :

[0066] Taking the robot dexterity index D = (1 / k F (J(θ))) × 100% as the evaluation index, determine the angle range of each joint of the robot, where J(θ) is the Jacobian matrix of the robot, and k F (J(θ)) is the condition number of the robot Jacobian matrix;

[0067] Based on the angle range of each joint of the robot, use the Monte Carlo method to solve all the positions of the tool tip point at the end of the robot, and define the enveloped range as the effective processing space A E of the robot, and set the effective processing space of the robot as the maximum processing range of the robot to optimize the flexibility of the multi-robot processing system;

[0068] Based on the effective processing space A E of the robot, adjust the horizontal distance d CB between the cabin and the robot positioning point so that the range of the outer surface of the cabin covered within the effective processing space of the robot is the largest, and solve the central angle α E of the cabin corresponding to the maximum coverage range, and set this central angle to be divided into upper and lower central angles by the horizontal plane passing through the center of the cabin and

[0069] According to the cabin size and the maximum central angle α E , calculate the minimum rotation threshold K min of the cabin, where

[0070] Step 42. Set the cabin rotation angle step α unit Discretize the cabin to determine the initial angle set α orig of the cabin rotation = {1:α unit :π}, assuming the cabin rotates to any angle α u ∈α orig , then the positioning angles of the workstations on both axial sides of the cabin are α uand α u +π; Obtain that when the cabin rotates to any angle α according to the pose parameters of the task to be processed u the maximum task sets of the brackets to be processed on both sides of the cabin are CS p-max (α u ) and CS n-max (α u ); Further calculate the total processing time TC u of all tasks at the opposite workstations of the cabin at each discrete angle α p (α u ) and TC n (α u ), and further calculate the difference △TC(α u ) in the total processing time of the opposite workstations of the cabin at each discrete angle α u .

[0071] When the number of cabin rotations is not less than the rotation threshold K min taking the minimization of the number of cabin rotations and U1 as the optimization objective, and using all α u ∈α org and all TC u corresponding to the corresponding TC p (α u ), TC n (α u ) and △TC(α u ) as the solution search space, determine the optimal set A K of rotation times and angles; Output each rotation angle α K in the set A k and the corresponding maximum task sets CS p-max (α k ) and CS n-max (α k ) of the opposite sides of the cabin. It can be known that at the same time α k belongs to both A K and α org .

[0072] Step 43. After determining the rotation angle set A K it is possible to determine that at each rotation angle of the cabin, the maximum task sets to be processed on both sides of the cabin are CS p-max (α k ) and CS n-max (α k);Since when the cabin body is positioned at adjacent angular workstations, in order to meet the envelope constraint, the machining areas covered by the robots at the same station intersect in the circumferential direction of the cabin body, resulting in overlapping tasks in the maximum set of tasks to be processed at adjacent angular workstations. Therefore, the maximum set of tasks to be processed can be decomposed into a fixed task set and a flexible task set. The next step is to reallocate the bracket tasks in the flexible task set to ensure the uniqueness of the task allocation when the cabin body is at each rotating workstation. Decompose the maximum set of tasks to be processed CS p-max (α k ) and CS n-max (α k ) into a fixed task set CS p-fix (α k ) and CS n-fix (α k ) and a flexible task set CS p-flex (α k ) and CS n-flex (α k );Each task in the fixed task set corresponds to a unique cabin body rotation angle, and each task in the flexible task set corresponds to multiple cabin body rotation angles;

[0073] Step 44: Taking the brackets in the flexible task set as objects, calculate the arc length swept by each bracket positioning point and the center points of all workstations, and evaluate the density of the flexible task sets on both sides of the cabin body at each rotation angle under different workstations. Allocate each bracket task in the flexible task set to the workstation with a higher density to obtain the unique task sets CS p (α k ) and CS n (α k ) on both sides of the cabin body at each rotation angle.

[0074] Step 5: Machining sequence planning for multi-robot multi-task: On the outer surface of the cabin body, taking the plane of the task to be processed farthest from the cabin body as a reference, expand a set distance h c as the radius, and construct a cylinder with the center of the cabin body as the center of the circle. The outer surface of the cylinder is used as a collision-free surface; after the workstations at each cabin body rotation angle are fixed, establish a vertical reference plane RP αk at the intersection of the collision-free surface and the Xc axis, and project the center points cp p (α k ) and CS n (α k ) of the task planes in the unique task sets along the normal direction of the task plane to the reference plane RP i to obtain the mapped point cp αk ' corresponding to each task; i ';

[0075] The motion mode of the end effector based on the collision-free conflict surface is as follows:

[0076] The robot tool center point moves within the reference plane RP αk to the center point cp of the task plane i at the corresponding mapped point cp i ', and then moves in the reverse direction along the normal direction of the task plane to the center point cp of the task plane i as the feed position point to start processing. After processing is completed, it returns from the origin of the end effector coordinate system to the reference plane RP αk inside, and within the reference plane RP αk moves to the center point cp of the next task plane i at the corresponding mapped point cp' i+1 to perform the processing of the next task, avoiding the risk of collision between the robot and the cabin during operation;

[0077] Based on the objective function U2, the mapped points cp p (α k ) and cp of the tasks in CS n (α k ) are sorted to obtain the task processing sequences CS i (α p ) and CS k ) of the two stations on both sides at each cabin rotation angle, and then complete the allocation of all tasks to be processed and the processing order of multiple robots. In this embodiment, the genetic algorithm is used to sort the mapped points cp'i of the tasks in the unique task set CS n (α k ) and CS p (α k ) to obtain the task processing sequences of the opposite stations of the cabin: CS n (α k ) = {CS p (α k ), CS kp,1 , CS kp,2 , …, CS kp,nk1}, CS n (α k ) = {CS kn,1 , CS kn,2 , …, CS kn,nk2}, and then complete the allocation of all tasks to be processed and the processing order of multiple robots. nk1 and nk2 respectively represent the number of tasks in the unique task sets CS p (α k ) and CS n (α k ) of the two stations on both sides of the cabin at each rotation angle.

[0078] Example: As Figure 1 shown, in order to enable those skilled in the art to better understand this embodiment, the following will clearly and completely elaborate on this embodiment in conjunction with the accompanying drawings. It includes:

[0079] Taking an example cabin with a length of 5.6 m and a diameter of 3.3 m as the processing object, 60 brackets are unevenly distributed on the outer surface of the cabin. The bracket tasks are numbered from 1 to 60 respectively. All the features of the surfaces to be processed are planes. Among them, the number of bracket tasks that only include milling and drilling processes is 40, while the number of bracket tasks that include three processes of milling, drilling, and grinding is 20; since the shapes and precisions of the feature surfaces of each bracket are different, the processing times of each task are also different, which are set by CAM software. The process tasks and times included in each bracket are shown in Table 1.

[0080] Table 1

[0081]

[0082]

[0083] According to the cabin size, 3 robots of model KUKA_KR500 are selected as the processing main bodies, and 3 AGV mobile trolleys are selected as the mobile devices for carrying the robots. According to the task types of the brackets, two end effectors with integrated milling and drilling functions are selected, and the other one is a grinding end effector. The maximum moving speed of the robot end is set to 0.1 m / s. The above equipment constitutes a multi-robot parallel processing system for local tasks of large cabins. The composition diagram of the processing system unit is as Figure 2 shown.

[0084] According to the cabin size and the distribution of bracket tasks, and considering the existence of the power supply cables of the robot system at the same time, in order to quickly complete all the tasks in the axial direction of the cabin, two milling and drilling robots are respectively distributed on both sides of the axial direction of the cabin, and each milling and drilling robot can only move around on one side of the cabin and is not allowed to cross between the two sides of the cabin. The milling and drilling robot positioned on the positive side of the X C axis in the axial direction of the cabin is set as R1, while the milling and drilling robot positioned on the negative side of the X C axis in the axial direction of the cabin is set as R2; and the mobile grinding robot system, based on the distribution of processing features and on the premise of ensuring that the cables between the systems do not cross, adopts a one-way movement method to enter both sides of the cabin in turn to complete the brackets with grinding tasks. The grinding robot is set as R3. The layout and movement path of the multi-robot parallel system are as Figure 3 shown.

[0085] A simplified method is used to determine the axial positioning positions of the robots along the cabin. The cabin is in the Y CThe 1 / 4 and 3 / 4 positions in the axial direction are set as the positioning positions in the axial direction of the robot; while the positioning of the robot in the X C axis direction is determined by maximizing the intersection arc of the cabin and the effective processing space of the robot, as shown Figure 4 below.

[0086] Balanced task allocation under multiple rotating workstations: First, taking the stiffness performance as the evaluation index, set the robot joint angle range as shown in Table 2. Use the Monte Carlo method to solve all positions of the robot end tcp. According to the dimensions of the cabin and the bracket, calculate the maximum arc length of each rotation envelope of the cabin within the effective processing range, and obtain the minimum rotation threshold K min- = 4.

[0087] Table 2

[0088]

[0089] Set the unit angle α unit = 1° to discretize the cabin, extract the initial angle set of the rotating workstations = [1, 180°], use the traversal method to extract the maximum processable task sets on both sides of the cabin at each discrete angle workstation, and calculate the total milling and drilling time of the processable bracket sets at all discrete workstations and the total time difference between the two sides of the cabin at the same angle when the cabin rotates one week according to the objective function U1, that is: obtain the load difference, as shown Figure 5 and Figure 6 below.

[0090] When the number of rotations K of the cabin satisfies K ≥ K min , to minimize the number of rotations K of the cabin and the load difference, determine the rotation angle set A K = {29°, 67°, 120°, 155°}, and obtain the maximum task sets to be processed and flexible task sets on both axial sides of the cabin at each rotation workstation according to the bracket pose characteristics, as shown in Table 3 and Table 4.

[0091] Table 3

[0092]

[0093] Table 4

[0094]

[0095]

[0096] Comprehensively considering the density of flexible tasks, complete the final allocation of flexible tasks, output the unique task set for each workstation, and complete the final task allocation for each workstation, as shown in Table 5.

[0097] Table 5

[0098]

[0099] Based on the obtained set of cabin rotation workstations and the task allocation results on both axial sides of the cabin at each workstation, multi-robot multi-task processing sequence planning is carried out: Since the number of tasks to be processed is large, to ensure that there is no collision interference between the robot and the cabin during the working process, a multi-task mapping method based on a plane is adopted. The specific schematic is shown in Figure 7 .

[0100] Taking the objective function U2 as the optimization objective, the GA algorithm is used to sort the mapping points in the reference plane to obtain the processing sequences of the brackets on both sides of the cabin at each workstation, and then they are sequentially assigned to the corresponding side and robot to obtain the task sorting of each robot, and finally a processing scheduling plan for the multi-robot parallel processing system is formed.

[0101] The processing scheduling plan of the multi-robot parallel processing system in this embodiment refers to taking time as the horizontal axis, determining the rotation angle of the cabin, the distribution of robots, and the task allocation and sorting of robots at different workstations at the same time node, and is represented by a Gantt chart of the task time sequence planning of the multi-robot parallel processing system, as shown in Figure 8 . It can be clearly observed from the Gantt chart of the processing task time sequence planning the working status and completion time of each robot when the cabin is at each rotation workstation. This two-step scheduling method for multi-robot parallel system processing of large cabin components can provide scientific guidance for the process of multi-robot processing of large components.

[0102] This embodiment provides a two-step scheduling optimization method for multi-robot parallel processing of local tasks of large cabins, and specific solution steps are implemented for a specific case. There are many types of components that can be solved by this technical solution. The above is only a special case selected by the present invention. It should be noted that for those skilled in the art of this technology, the method of this embodiment can be used to solve the multi-robot parallel processing scheduling problems of various different types of large components with multiple discrete features.

[0103] In the first step of this embodiment, by establishing a multi-station division model and combining the support distribution characteristics and the reachability constraints of the robot, the optimal allocation of processing tasks is achieved, ensuring balanced loads at each station and minimizing the number of rotations of the cabin. In the second step, an optimization model for task sequencing under spatio-temporal constraints is constructed. The plane mapping method is introduced to eliminate the collision risk between the robot and the cabin components, and the genetic algorithm is used to generate a collision-free task processing sequence. By decoupling the coupling relationship between task allocation and sequencing, this embodiment decomposes the NP-hard problem into a two-level optimization structure, significantly reducing the solution complexity and achieving the technical effects of shortening the processing time and improving the robot utilization rate. Compared with the traditional collaborative planning method, it effectively solves the problem of system efficiency optimization under multi-objective constraints while ensuring the position accuracy, and is particularly suitable for the high-precision parallel processing scenario of large components such as spacecraft cabins.

[0104] Although the present invention has been described herein with reference to particular embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. It should thus be understood that numerous modifications may be made to the exemplary embodiments and that other arrangements may be devised, provided that they do not depart from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and the features described herein may be combined in ways different from those described in the original claims. It should also be understood that features described in connection with a separate embodiment may be used in other described embodiments.

Claims

1. A multi-robot parallel machining scheduling optimization method for local tasks of large cabins, characterized in that Including: S1. Construct a multi-robot parallel processing system for large cabin bodies; S2. Establish objective functions U1 and U2. The objective function U1 is a function to minimize the load difference of robots equipped with the same type of end effector in the multi-robot parallel processing system, and the objective function U2 is a function to minimize the maximum time for the multi-robot parallel processing system to process all tasks; S3. Obtain the set A of cabin rotation angles based on the objective function U1 and minimizing the number of cabin rotations K K ; The cabin rotation angle α k ∈A K , k ∈ [1, K], obtain the unique task set CS of the workstations on both sides of the cabin at each cabin rotation angle p (α k ), CS n (α k ); S4. On the outer surface of the cabin, take the plane of the task to be processed farthest from the cabin as the reference, expand outward with the set distance as the radius, and build a cylinder with the center of the cabin as the center. The outer surface of the cylinder is used as the non-collision surface; after the workstations at each cabin rotation angle are fixed, a vertical reference plane RP is established at the intersection of the non-collision surface and the Xc axis. αk , and the unique task set CS p (α k ), CS n (α k ) in the center point of the task plane cp i Projection along the normal direction of the task plane to the reference plane RP αk Get the mapping point cp corresponding to each task i '; Based on the objective function U2, the unique task set CS p (α k ), CS n (α k ) in the task mapping point cp i ' to obtain the task processing sequence CS of the two sides of the workstation under each cabin rotation angle p (α k ), CS n (α k ); The cabin coordinate system takes the axial direction of the cabin as the Y C axis, the direction perpendicular to the horizontal plane and upward as the Z C axis, and the X C axis is the radial direction of the cabin.

2. The multi-robot parallel machining scheduling optimization method for local tasks of large cabins according to claim 1, wherein The objective function U1 is: U1 = min{|max{TC pos (CS(R q1 ), a k )} - min{TC neg (CS(R q2 ), a k )}|} where TC pos (CS(R q1 ), a k ) represents the total time of all tasks corresponding processes assigned to the end effector on the positive side of the XC axis of the cabin when the cabin rotation angle is αk, and TC neg (CS(R q2 ), a k ) represents the total time of all tasks corresponding processes assigned to the same type of end effector on the negative side of the X k axis of the cabin when the cabin rotation angle is α C .

3. The multi-robot parallel processing scheduling optimization method for local tasks of large cabin bodies according to claim 1, wherein The objective function U2 is: Wherein, TC(CS k , α k ) represents the operation time span of all assigned tasks at the cabin rotation angle α k , and TC(WS k ) represents the non-operation time when the cabin rotates from the cabin rotation angle α k to the cabin rotation angle α k+1 .

4. The multi-robot parallel machining scheduling optimization method for local tasks of large cabins according to claim 1, characterized in that The number of times K that the cabin rotates satisfies K≥K min , K min Obtaining method: With the robot dexterity index D = (1 / k F (J(θ))) × 100% as the evaluation index, determine the angular range of each joint of the robot, where J(θ) is the Jacobian matrix of the robot, and k F (J(θ)) is the condition number of the Jacobian matrix of the robot; Based on the angular ranges of the robot's joints, the Monte Carlo method is used to solve for all the positions of the tip point of the end effector, and the range enveloped by all these positions is defined as the robot's effective machining space A E ; Based on the effective machining space A of the robot E , adjust the horizontal distance d between the cabin and the positioning point of the robot CB , so that the range of the outer surface of the cabin covered within the effective machining space of the robot is the largest, and solve the central angle α of the cabin corresponding to the maximum coverage range E ; The minimum rotation threshold K of the cabin min = 2*π / α E .

5. The multi-robot parallel machining scheduling optimization method for local tasks of large cabins according to claim 4, characterized in that S3 Including: Set the step size α of the cabin rotation angle unit Discretize the cabin to determine the initial angle set α of the cabin rotation orig ={1:α unit :π}, assuming the cabin rotates to any angle α u ∈α orig , then the working position positioning angles on both axial sides of the cabin are α u and α u +π; according to the pose parameters of the task to be processed, obtain the maximum task sets to be processed on both sides of the cabin when the cabin rotates to any angle α u , which are CS p-max (α u ), CS n-max (α u ) respectively; When the number of times the cabin rotates is not less than the rotation threshold K min , with the goal of minimizing the number of times the cabin rotates and U1, for any angle α u , all the corresponding TCs p (α u ), TC n (α u ) and the set of △TC(α u ) are the solution search space, to determine the number of times the cabin rotates K and the angle set α K ; TC p (α u ) is the rotation angle α of the cabin u The end effector of the multi-robot parallel processing system is located in the cabin X C The total process time of all tasks assigned to the positive side of the axis, TC n (α u ) is the rotation angle α of the cabin u When the same end effector is in the cabin X C The total process time of all tasks assigned to the negative side of the axis; △TC (α u )=TC p (α u )-TC n (α u ); The rotation angle α of the cabin k ∈A K , k ∈ [1, K], and the maximum task set to be processed is CS p-max (α k ), CS n-max (α k ); Decompose the maximum task set CS to be processed p-max (α k ) and CS n-max (α k ) into a fixed task set CS p-fix (α k ), CS n-fix (α k ) and a flexible task set CS p-flex (α k ), CS n-flex (α k ); Each task in the fixed task set corresponds to a unique cabin rotation angle, and each task in the flexible task set corresponds to multiple cabin rotation angles; Obtain the flexible task set CS p-flex (α k )、CS n-flex (α k ) Divide the tasks in CS into the rotation angles of the cabins with high density, and obtain the unique task set CS for the workstations on both sides of the cabin at each rotation angle p (α k )、CS n (α k ).

6. The multi-robot parallel machining scheduling optimization method for local tasks of large cabin bodies according to claim 1, wherein The motion mode of the end effector based on the collision-free surface is: The robot tool center point moves within the reference plane RP αk to the center point cp of the task plane i and the corresponding mapped point cp i '. Then it moves in the reverse direction along the normal direction of the task plane to the center point cp i of the task plane as the feed position point to start machining. After machining is completed, it returns from the origin of the end effector coordinate system to the reference plane RP αk within the reference plane RP αk and moves to the center point cp of the next task plane i and the corresponding mapped point cp i ' +1 to perform the machining of the next task.

7. The multi-robot parallel machining scheduling optimization method for local tasks of large cabins according to claim 6, characterized in that In S4, based on the objective function U2, the genetic algorithm is used to process the unique task set CS p (α k ) and the mapping points cp n (α k ) of the tasks in CS i ' are sorted to obtain the task processing sequence for the opposite-side workstations of the cabin: CS p (α k ) = {CS kp,1 , CS kp,2 , …, CS kp,nk1}, CS n (α k ) = {CS kn,1 , CS kn,2 , …, CS kn,nk2}, thereby completing the allocation of all tasks to be processed and the processing sequence of multiple robots. nk1 and nk2 respectively represent the number of tasks in the unique task sets CS p (α k ) and CS n (α k ) on both sides of the cabin at each rotation angle.

8. The multi-robot parallel machining scheduling optimization method for local tasks of large cabins according to claim 1, characterized in that S1 includes: Obtain data of the large cabin body of the spacecraft and the mobile robot processing system; According to the obtained data, construct a multi-robot parallel processing system for the large cabin body, and determine the unit composition, multi-robot layout and movement path of the multi-robot parallel processing system.

9. A multi-robot parallel processing scheduling optimization device for local tasks of large cabins, comprising a storage device, a processor, and a computer program stored in the storage device and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the multi-robot parallel processing scheduling optimization method for local tasks of large cabin bodies as described in any one of claims 1 to 8.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-robot parallel processing scheduling optimization method for local tasks of large cabin bodies as described in any one of claims 1 to 8.

Images