JP2004243461A - Teaching system of robot - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a teaching system restraining an increase in work time even when the travel distance of a robot in work is long and even when most of operating areas are shared by robots. <P>SOLUTION: This teaching system of the robot which is adapted to plan an operation program to a plurality of robots includes: a task distribution planning part 111 for distributing tasks to two or more robots; a task processing sequence planning part 112 for planning a task processing sequence in each robot; a robot operation planning part 113 for planning the operation program and calculating the operation time according to a task distribution plan and the task processing sequence; a robot-to-robot collision avoiding operation planning part 115 for altering the operation program to avoid a collision between the robots and calculating the operation time; and evaluation reference calculating parts 111a, 112a for calculating an evaluation reference on the basis of the operation time of the robot operation planning part 113 and the operation time of the robot-to-robot collision avoiding operation planning part 115. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a teaching system that determines work sharing and a work order to a plurality of robots installed such that their operation areas overlap with each other, and creates an operation program for each robot. Rather, the present invention relates to a system that determines work sharing and work order such that the overall work time is shortened.
[0002]
[Prior art]
Automobile production lines use multiple industrial robot manipulators to automate operations such as spot welding. Here, a large number of robots are arranged so close that their operation areas overlap. Conventionally, the teaching work of a plurality of robots close to each other has conventionally been performed by the instructor while actually operating the operation of each robot one step at a time, and the operation trajectory of each robot so that interference does not occur between the plurality of robots. The execution timing (interlock) between each motion trajectory was taught by trial and error. However, there is a problem that it is very difficult to operate a plurality of robots safely and efficiently by trial and error teaching by a teacher. When trying to operate a plurality of robots efficiently, the teaching operation becomes very complicated, and it takes a very long time to teach and check the operation. In particular, in high-mix low-volume production where the type of product changes frequently, the teacher will have to redo all the teaching work every time the type of product is changed. In addition, the ratio of the time of the indirect process such as the teaching operation is increased, and the productivity is extremely reduced. Also, since the ratio of the instructor directly involved in the teaching operation is large, a rise in labor costs is also a problem.
In recent years, instead of direct teaching by a teacher, a method and an apparatus have been proposed which increase the efficiency of teaching work and operation control of a plurality of robots using a computer.
[0003]
In Conventional Example 1, when simulating a work time, which is an evaluation criterion at the time of repetitive search, if there is a possibility that a collision occurs when a plurality of robots perform work at the same time, a priority is set for the robots, and a higher priority is set. Collisions are avoided by prioritizing the robot's work and waiting for lower-priority robots to eliminate the possibility of collision. FIG. 12 shows an example of an initial solution of work assignment and work order. In FIG. 12, R1 to R3 indicate the robot name, G1 to G14 indicate the spot welding spot groups, and reverse / order indicates from which direction the spot groups are processed. When a collision occurs between the robots when simulating the work sharing and the work order from R1 to R3 shown in FIG. 12, if a collision occurs between the robots, the point group of the robot with a higher priority is determined according to the initial priority R1 → R2 → R3. The robot is executed first, and the execution of the dot group of the robot having the lower priority is waited until the processing of the dot group is completed (for example, Patent Document 1).
[0004]
In Conventional Example 2, when simulating a work time, which is an evaluation criterion at the time of repetitive search, if there is a possibility that a collision will occur when a plurality of robots perform work at the same time, the work progress state of the currently operating robot is monitored. The collision is avoided by making another robot wait until the possibility of the collision is eliminated. FIG. 13 shows a procedure for avoiding collision between robots. In FIG. 13, reference numerals 1 to 9 indicate welding lines. The upper part of the figure shows that the robot 1 is currently processing the work of the welding line 1, the area where the robot 1 passes from the current position of the robot 1 to the end position of the welding line 1 (interference area), and the robot 2 is now processing. This indicates that the area where the robot 2 passes from the start position to the end position of the welding line 4 interferes (overlaps). In this state, the operation of the robot 2 is put on standby. In the lower part of the figure, the passing area of the robot 1 is updated according to the work progress status of the robot 1, and the robot 2 starts working when the overlapping area with the passing area of the robot 2 disappears (for example, Patent Document 2). .
[0005]
In Conventional Example 3, when teaching data specifying work assignment and work order is given to a plurality of robots, collision between the robots is avoided, and the robots are efficiently operated with a minimum waiting time of the robots. A control device is disclosed. FIG. 14 illustrates the operation principle of the robot control device. In the same figure, (A) shows a work point between the robot A and the robot B and an operation path between the work points. (B) shows an interference table between the work points of the robots A and B calculated by the interference extraction means based on the work point data of (A). (C) shows the priority order table of the robot at the interfering work point calculated by the non-interference passing order determination means. FIG. 14B shows that interference occurs when the robots A and B work simultaneously at the work points 2a and 2b, 3a and 3b, and 4a and 4b. FIG. 14C shows that the tasks 2a, 3a, and 4a of the robot A are executed prior to the tasks 2b, 3b, and 4b of the robot B. When creating the priority order table, the sum of the remaining work time of each robot and the waiting time at the work point is compared, and the longer one is selected to determine the priority order (for example, Patent Document 3). .
[0006]
In Conventional Example 4, when a plurality of machines (welding robots) are assigned work (welding line), a control device and a control method for preferentially determining a work route of each machine from a machine with a large amount of work are disclosed. It has been disclosed. FIG. 15 shows a configuration diagram of the control device. In the figure, a CAD / CAM system 2 creates information necessary for work assigned to a plurality of machines. The work information storage unit 4 stores the work information created by the CAD / CAM system 2. The work amount evaluation unit 6 evaluates the work amount of each machine. The machine selection unit 8 preferentially selects a predetermined number (here, N) of machines with a large amount of work. The work route selection unit 10 creates a plurality of work routes for each of the N machines selected by the machine selection unit 8 and evaluates the work time. In the work route selection unit 12, for each of the N machines selected by the machine selection unit 8, a predetermined number (here, a short work time) of a plurality of routes created by the work route evaluation unit 10 with a shorter work time is given priority. M). The whole work route evaluation unit 14 creates a plurality of work routes for all the machines that work in parallel with the work for each of the N × M work routes selected by the work route selection unit 12, and for all of them, Evaluate working time. The all-work-path selecting unit 16 selects a work path with the shortest working time from the N × M work paths, and the all-work-path evaluating unit 14 performs processing on all the machines other than the machine in charge of the selected work. A work route with the shortest work time is selected from the created plurality of work routes. The machine control unit 22 controls a plurality of machines according to the work paths of all the machines selected by the all work path selection unit 16. When the operations along the work (welding line) path of a plurality of welding robots collide, the welding line is divided into a plurality of parts so as to reduce the interference state, and is processed by all the work path evaluation units (for example, Patent Document 4).
[0007]
[Patent Document 1]
Japanese Patent No. 2984218
[Patent Document 2]
Japanese Patent No. 2778922
[Patent Document 3]
JP-A-11-347984
[Patent Document 4]
JP 2002-116816 A
[0008]
[Problems to be solved by the invention]
In the methods disclosed in Conventional Example 1 and Conventional Example 2, the execution of robot operations is waited for in units of a welding line or an entire group of spots (a plurality of spots of spot welding), thereby avoiding collision between the robots. The work time of the work is simulated, and the work sharing and the work order are determined based on the simulation value as an evaluation criterion. Therefore, if the moving distance within one operation (welding line or dot group) becomes longer, the passage area of the robot for the operation becomes wider and the operation time becomes longer, so that the waiting time of the robot increases and the total operation time becomes longer. There is a problem that (calculated value) becomes longer. Further, when a plurality of robots are arranged closer to each other and most of the operation area is shared between the robots, there is a problem that the possibility of collision frequently occurs, which similarly leads to an increase in working time.
[0009]
In the method disclosed in Conventional Example 3, referring to FIG. 14, the robot B waits at the work point 1b until the robot A finishes the work at the work point 2a and passes through the interference area 2a. However, actually, since there is a time for the robot B to move from the work point 1b to the work point 2b, it is not always necessary to keep the robot B at the work point 1b until the robot A passes the interference area 2a. In other words, there is ample room for further reducing the overall work time, but there is a problem that no further consideration can be given by this method since no consideration is given to the movement trajectory between the work points.
[0010]
In the method disclosed in the conventional example 4, when the robots collide with each other, the corresponding welding line is divided into a plurality of parts, and the calculation is performed in all the work route evaluation units (14 in FIG. 15). It does not mean that a direct calculation for avoiding a collision such as setting is performed. Since the work path is created in consideration of the processing order of the weld line, the welding direction, and the movement path to the welding start point, an enormous number of combinations occur by dividing the weld line. Processing is performed to select one that does not cause a collision from among the enormous combinations of work routes. Therefore, there is a problem that an explosive increase in the amount of computation is caused and efficient collision avoidance cannot be performed. Further, since the movement trajectory from the welding line to the welding line is not subject to division, there is also a problem that a large waiting time is required in moving between the welding lines.
[0011]
The present invention has been made in view of the above problems, and does not consider only one entire operation or an operation point, but considers a movement trajectory between operations (points) and each trajectory in the operation in detail. A robot teaching system that minimizes the increase in work time even when the moving distance of the robots inside is long or when most of the operating area is shared between robots, and that does not cause deadlock between robots. It is intended to provide.
It is still another object of the present invention to provide a robot teaching system having an efficient robot collision avoidance function without causing a sudden increase in the amount of computation even if the trajectory is carefully considered.
[0012]
[Means for Solving the Problems]
In order to solve the above-mentioned problem, the robot teaching system according to claim 1 evaluates each of the robots from a combination of task sharing and task processing order so as to perform a predetermined task on a plurality of robots. In a robot teaching system that preferentially searches for a reference that minimizes or maximizes a reference and generates an operation program, a task distribution planning unit that distributes the work to the plurality of robots, and a robot for each of the distributed tasks A task processing order planning unit that plans a task processing order within, avoiding collision between each robot and an obstacle based on the task distribution plan and the processing order of the tasks, and an operation start position of each robot. A robot operation planning unit that generates an operation program and calculates an operation time in which the end position is set to an area where the robots do not collide with each other; If there is a trajectory in which a collision occurs between robots in a gram, the operation program is changed and the operation time is changed so that the trajectory is divided into a plurality of parts and the divided trajectory avoids collision between robots. And an evaluation criterion calculation unit that calculates the evaluation criterion based on the operation time of the robot operation planning unit and the operation time of the inter-robot collision avoidance operation planning unit. It is characterized by the following.
[0013]
3. The robot teaching system according to claim 2, wherein the inter-robot collision avoidance operation planning unit includes an optimization processing unit that changes an operation plan of each robot so that the operation time of the plurality of robots is minimized. It is assumed that.
According to a third aspect of the present invention, there is provided a robot teaching system including an operation program merging unit that merges a plurality of the operation programs of the same robot into one operation program in an execution order.
5. The robot teaching system according to claim 4, wherein the operation program merging unit deletes a trajectory from a last work point to an end position of the operation program and a trajectory from a start position of the next process operation program to a start work point. And a trajectory adding unit that adds a trajectory from the final working point to the starting working point between the two operation programs.
[0014]
6. The robot teaching system according to claim 5, wherein the task processing order planning unit sets a plurality of tasks that can be executed in parallel as one task set based on a task graph describing a predecessor-relationship between the tasks, and The task distribution planning unit has a function of searching for a combination of robots that execute each task from the first task set in the task set list. It is assumed that.
7. The robot teaching system according to claim 6, further comprising: a development rule information unit that associates the front-back relationship between the tasks with the operation front-back relationship between the robots, wherein the robot collision avoidance operation planning unit determines the front-back relationship between the tasks. The present invention is characterized in that a preceding / rear relationship consideration unit for changing an operation program based on the first and second relations is provided.
[0015]
8. The robot teaching system according to claim 7, wherein the task distribution planning unit, the task processing order planning unit, and the inter-robot collision avoidance operation planning unit are software objects that can be independently and separately processed. To do.
The robot teaching system according to claim 8, wherein the robot operation planning unit is an independent software object for each robot.
[0016]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a system configuration diagram according to an embodiment of the present invention. Reference numeral 101 in FIG. 1 denotes a task input unit for inputting a task to the present system. The task input by the task input unit 101 is stored in the task information 102. Reference numeral 103 denotes a condition input unit for inputting processing conditions related to a workpiece or a component processed by a task. The information input by the condition input unit 103 is stored in the work / part information 104. Reference numeral 105 denotes a model input unit for inputting a geometric model of an environment such as a robot, a part or a work. Information input by the model input unit 105 is stored in the geometric model 106. A robot registration unit 107 registers a target robot and sets the information. Information registered by the robot registration unit 107 is stored in the robot information 108. Reference numeral 109 denotes a rule registration unit for registering a rule relating to the movement development of the robot. The information registered by the rule registration unit is stored in the deployment rule information 110. The database of the work / part information 104, the geometric model 106, the robot information 108, and the development rule information 110 needs to be set before a task is input. These databases are all stored in association with each other.
[0017]
In FIG. 1, reference numeral 111 denotes a task distribution planning unit that accesses the task information 102 and determines distribution (assignment) of tasks to a plurality of robots by a best priority search. A task processing order planning unit 112 accesses the task information 102 and determines the processing order of the tasks by the best priority search. Each of the task distribution planning unit 111 and the task processing order planning unit 112 includes an evaluation criterion calculation unit (111a, 112a) that calculates an evaluation criterion in the priority search process. Reference numeral 113 denotes a robot operation planning unit that generates a robot operation procedure and a trajectory for a task assigned to each robot. The trajectory generated by the robot operation planning unit 113 is stored in the operation program 114. Reference numeral 115 denotes an inter-robot collision avoidance operation planning unit that corrects the operation program 114 so that the robots do not collide with each other when there is a risk of collision between a plurality of robots.
[0018]
FIG. 2 shows an assembly robot cell for assembling parts A to E onto a workpiece W by three robots 201a, 201b, 201c (R1, R2, R3). The work W before the components are assembled is supplied to the position shown in the figure by the belt conveyor 202, and after the assembly, the work W is similarly unloaded by the belt conveyor 202. 203a to 203d are component supply devices. The components A to E are supplied from any of the component supply devices 203a to 203d. 204a to 204c are tool holders. Holds several types of hands (grippers) necessary for assembling parts. The robots 201a to 201c perform work while exchanging hand parts according to parts to be assembled.
[0019]
FIG. 3 shows work information of the assembly work shown in FIG. FIG. 3A describes the task of the assembly work. In the drawing, T0-0 to T0-5 indicate tasks for supplying the workpiece W and the parts A to E to predetermined positions. T1 to T5 indicate tasks for assembling the parts A to E at the designated positions of the work W, respectively. T6 indicates a task of unloading the assembled work W. Here, tasks that can be executed by the robots R1 to R3 are T1 to T5. Also, the tasks from T0-0 to T6 are not executed in this order, but are executed so as to satisfy the precedent-rear relationship between tasks restricted in the assembly process. FIG. 3B shows a task graph for specifying the precedent and succeeding relationship between tasks. Arrows in the figure indicate the front-back relationship. For example, T3 (assembly of component C) can be executed only after T1 (assembly of component A) and T2 (assembly of component B).
[0020]
First, a first embodiment according to the present invention will be described.
The task description of FIG. 3A and the task graph of FIG. 3B (before and after) are input from the task input unit 101 of FIG. When the task description and the task graph are input, the task processing order planning unit 112 groups a plurality of tasks that can be processed in parallel into one task set as shown in FIG. Generate a task set list indicating the order. That is, the processing is performed in the order of TaskSet0 → TaskSet1 → TaskSet2 → TaskSet3 → TaskSet4. Here, TaskSet0 and TaskSet4 indicated by dashed lines are task sets not assigned to the robot, and will be omitted from the following description (assuming that they have been deleted).
When the task set list shown in FIG. 3C is generated by the task processing order planning unit 112, the task distribution planning unit 111 of FIG. 1 generates a set of assigned robots (hereinafter referred to as assigned robot set) for the task set. The task distribution (allocation) to the robot is determined by repeatedly searching for the best one from the candidates (best priority search).
[0021]
FIG. 4 shows a data structure of a search tree used in the best priority search. As shown in FIG. 4A, the search tree is represented by a tree structure, and n0 to n6 in the figure indicate nodes of the tree. Each node of the search tree indicates a state of the search process, and the parent-child relationship of the nodes indicates a transition process of the search state. In the present invention, each node of the search tree is made to correspond to a candidate of a robot set in charge of a task set. Further, each node is actually stored in two linked lists 401 and 402 shown in FIG. 4B. Reference numeral 401 denotes a ClosedList that stores expanded nodes (having child nodes). An OpenList 402 stores an expansion candidate (unexpanded: has no child node) node. Reference numerals 403a to 403f denote pointers to parent nodes, which correspond to the branches of the tree shown in FIG. FIG. 4C shows data held by each node other than the link (broken arrow) indicating the next node in the linked list and the pointer to the parent node. Details of the data shown in (c) will be described later.
[0022]
FIG. 5 shows a flowchart of the task distribution planning unit 111.
In the initialization of the search tree in FIG. 5 (S5-1), a process of generating only the root node of the search tree shown in FIG. 4 is performed. That is, ClosedList is empty, and one node is generated in OpenList. In this initial node, the current task set ptr_CurTaskSet points to Null, and the next task set ptr_NextTaskSet is set to point to the pointer to the first node TaskSet1 in the task set list shown in FIG. Also, the total work time is set to zero. In S5-2, a node having the smallest evaluation value is selected from unexpanded nodes (OpenList) in the search tree (the selected node is set to N), and in S5-3, the node N is moved to ClosedList and changed. In the present invention, nodes are selected using the total work time held by each node as an evaluation criterion. Immediately after the initialization in S5-1, the root node is selected. In S5-4, it is checked whether or not ptr_NextTaskSet held by the node N is Null. If the value is not Null, the process proceeds to a loop for obtaining a candidate for the assigned robot set for TaskSet indicated by ptr_NextTaskSet held by the node N.
[0023]
For example, for TaskSet1 (T1, T2) in FIG. 3 (c), when all the candidates for the assigned robot set are obtained from the robots R1 to R3 in FIG. 2, (R1, R1), (R1, R2), (R1, R3), (R2, R1), (R2, R2), (R2, R3), (R3, R1), (R3, R2), and (R3, R3). At the beginning of the loop of S5-5a, each of the candidates for the assigned robot set shown in the example is generated one by one every time the loop is performed. In S5-5b, it is determined whether or not each robot can be executed with respect to the generated candidates. If possible, an operation program is generated for each task candidate assigned to each robot in S5-5c. The processing in S5-5b and S5-5c is requested to the robot operation planning unit 113 in FIG.
[0024]
The robot operation planning unit 113 uses the work / part information 104, the geometric model 106, the robot information 108, and the development rule information 110 to individually determine whether or not to execute the assigned task candidate and generate an operation program. The technique of generating a robot operation program from a task description (work program) using various work databases, such as work part information and deployment rules, is described in the document "Arai, Inoko, Nogo: Simple teaching of industrial robots. System, Journal of the Robotics Society of Japan, Vol. 17, No. 2, pp. 186-189, 1999 ”and the like, and are known technologies. Here, a specific example will be described using the assembly work shown in FIG. 2 as an example.
FIG. 6 shows an example of the work / part information 104. In FIG. 6, reference numerals 601 and 602 denote a component A and a component supply device, respectively. Tool_type in the figure indicates the type of hand that is suitable for gripping the component A. The position indicates the position (including the posture) of the origin of the coordinate system 603 fixed to the component A. The position data is described as a value based on a coordinate system 604 fixed to the component supply device 602 to which the component is supplied. “approach_pos” indicates a position at which the TCP of the robot reaches when the robot approaches the part A. This position data is expressed as the origin position of the coordinate system 605 based on the coordinate system 603 of the part A. catch_pos indicates the position at which the robot's TCP reaches when the robot grips the part A. This position data is expressed as the origin position of the coordinate system 606 based on the coordinate system 603 of the part A. leave_pos indicates the position at which the robot's TCP reaches when the robot grips the component A and then leaves the component supply device by a certain distance. This position data is expressed as the origin position of the coordinate system 607 based on the coordinate system 603 of the part A.
[0025]
In the execution possibility determination process of the robot operation planning unit 113, referring to the component information shown in FIG. 6, whether the matching tool is held by the robot itself or the tool holder, or the origin position of the coordinate system 605 to 607 is determined by the robot operation. Check if it is within the range. If it is determined that the assigned task candidate is executable, the robot operation planning unit 113 expands the task description into subtasks as much as possible according to the expansion rule information 110. FIG. 7 shows an example of the deployment rule information 110.
In FIG. 7, reference numeral 701 denotes a deployment rule of the Place task. This indicates that the task description “Place X (position) on Y” is expanded to a row of subtasks after the arrow (⇒). “IF Robot.tool_type! = X.tool_type then ChangeTool X” in the column 701 indicates that the tool is replaced (ChangeTool X) when the tool type currently mounted on the robot does not match the tool type of the part X. Is shown. Reference numeral 702 denotes an expansion rule of “ChangeTool X”. “Approach X” in 701 is a subtask description that causes the TCP of the robot to approach the approach_pos of the part X. “PickUp X” is a subtask description that causes the robot to pick up the part X. Reference numeral 703 denotes an expansion rule of “PickUp X”. In this way, the tasks are expanded one after another in the sub-task sequence and the sub-tasks are sequentially expanded into the sub-subtask sequence, and finally, the robot controller is broken down to an operation program written in a robot language that can be directly executed. At this time, specific motion path data of the robot is generated with reference to the work / part information 104, the geometric model 106, and the robot information 108.
[0026]
However, for “Approach X”, an obstacle avoidance trajectory calculation process is performed inside the robot motion planning unit 113 instead of task expansion based on the expansion rule. In the obstacle avoidance trajectory calculation process, the TCP of the robot reaches X approach_pos so that the entire robot including the tool and the gripped work does not collide with the obstacle, starting from the end position of the subtask before “Approach X”. The motion trajectory to be calculated is calculated. However, the moving parts of the robot other than the target robot are not included in the obstacle, and the collision avoidance between the target robot itself and the fixed obstacle is considered.
In “Approach Robot.home_pos”, Robot. home_pos is the home position of the robot registered in the robot information 108 of FIG. This home position is set in a safe area where there is no possibility that the robots collide with each other. This subtask is for retracting the robot in a safe area after executing “Put XY.position1” in consideration of avoiding collision with a fixed obstacle. When the robots are registered in the assembly cell of FIG. 2, each robot is assumed to be at each home position. Therefore, when each robot executes the first task, it is assumed that the operation always starts from the home position.
[0027]
As described above, the robot operation planning unit 113 converts the task description shown in FIG. 3A into an operation program described in a robot language as shown in FIG.
When the generation of the operation program of each robot is completed for the robot set in charge of the task set, collision avoidance between robots (S5-5d) shown in FIG. 5 is performed. This processing is requested to the robot collision avoidance operation planning unit 115 in FIG.
The robot-to-robot collision avoidance operation planning unit 115 adds the internal dividing point to the trajectory of each movement command (MOVJ or MOVL) of the operation program of each robot generated in S5-5c or the path of the trajectory to obtain a plurality. For each of the divided trajectories, determine the execution order of the trajectories between different robots so that the robots do not collide with each other and minimize the overall work time, and assign each trajectory of each robot according to the execution order. Modify each operation program so that it can be executed. In FIG. 1, the robot collision avoidance operation planning unit 115 includes an optimization processing unit 115a inside in order to determine the execution order of the trajectory that minimizes the overall work time.
[0028]
FIG. 10 shows an example of the execution order of trajectories between different robots. In FIG. 10, A1 to A3 and A4-0 to A4-3 indicate the trajectories of the movement command of the robot A. B1-0 to B1-3 and B2 to B4 indicate the trajectories of the movement command of the robot B. Here, A4-0 to A4-3 indicate trajectories obtained by dividing the trajectory of the original movement command A4 into four. The same applies to B1-0 to B1-3. Arrows 1105a to 1105c in the figure indicate the order of execution of the trajectories between different robots (only the minimum necessary execution order relationship is described without redundant components). For example, 1105a indicates that the trajectory B-1 of the robot B is executed after the trajectory A3 of the robot A is executed.
FIG. 11 shows an example of an operation program modified in accordance with the execution order of the trajectories shown in FIG. IO instructions (DOUT and WAIT) are inserted at the start and end points of 1105a to 1105c. OT # and IN # in FIG. 11 are assigned numbers with reference to connection information (IO map) between IO input / output channels of the robot controller.
Further, as shown in FIG. 10, by calculating the execution order of the trajectory between different robots, it is possible to calculate the working time (Te-Ts in the figure) of the plurality of robots as a whole.
[0029]
In S5-5e of FIG. 5, for the task set indicated by ptr_NextTaskSet, the work time obtained as a result of performing the robot collision avoidance S5-5d is acquired, and the sum of the work time and the total work time of the parent node (node N) is obtained. It is the total work time including the corresponding task set. This step S5-5e corresponds to the processing of the evaluation criterion calculation unit 111a inside the task distribution planning unit 111.
In S5-5f of FIG. 5, a new node is created in the OpenList of the search tree, the candidate of the assigned robot set selected in S5-5a, the work time of the corresponding task set obtained by avoiding the collision between the robots in S5-5d, S5-5e Save the total work time up to the corresponding task set calculated in. Further, the ptr_NextTaskSet of the node N is stored as the ptr_CurTaskSet of the newly created node, and a pointer to the task set next to the task set corresponding to the ptr_CurTaskSet from the task set list of FIG. As ptr_NextTaskSet.
[0030]
The above processing from S5-5a to S5-5f is performed for all the robot set candidates. When it is determined in S5-5a that the selection of all the candidates has been completed, the process returns to S5-2. When the ptr_NextTaskSet of the node N becomes Null in S5-4, it means that the search process reaches the last task set in the task set list, and that the node N has the minimum total work time. . In that case, in S5-6, the route on the tree from the root node of the search tree to the node N is stored as a solution. Actually, the order is reversed by finding the route from the node N to the root node by following the parent node.
In S5-7, the number of solutions stored in S5-6 is counted, and if a predetermined number of solutions are obtained, all processes are terminated (S5-8a). If the number of solutions is equal to or smaller than the predetermined number, S5 is executed. Return to -2.
If the selection of the node from the OpenList fails in S5-2, the OpenList is empty, which means that the search has been completed for all the assigned robot set candidates, and thus all the processing is ended (S5-8b). .
When the robot on the simulator or the actual robot is operated according to the operation program, the corresponding operation program may be sequentially played back from the first task set in the task set list in FIG.
[0031]
As described above, when creating an operation program of each robot for each candidate in the task distribution / processing order, the operation start position and the operation end position are set in a safe area where the robots do not collide with each other. The collision avoidance between the fixed robot and the corresponding robot except for the robot is considered. If there is a possibility that multiple robots will collide, the trajectory of the motion command that composes the relevant motion program of the relevant robot, or the trajectory obtained by dividing the trajectory into a finite number of individual The collision order between the robots is avoided by determining the execution order of the trajectory between different robots so that the work time of the robot is minimized. Therefore, even when the moving distance of the robot in the work is long or when most of the operation area is shared between the robots, an increase in the work time can be suppressed as much as possible, and a deadlock state does not occur between the robots. In the robot collision avoidance operation plan, the trajectory in the operation program is finely divided, but the execution order in the same operation program is not changed.
[0032]
Next, a second embodiment according to the present invention will be described. In the second embodiment, in the robot collision avoidance processing (S5-5d) in FIG. 5, all tasks from the first task set in the task set list in FIG. 3C to the task set specified by NextTaskSet are performed. In consideration of the set, an operation plan for avoiding collision between robots is performed for a robot operation program related to all the task sets. For example, in the flowchart of FIG. 5, when ptr_NextTaskSet points to TaskSet2 of FIG. 3C, in S5-5d, a robot collision avoidance operation plan including TaskSet1 and TaskSet2 is performed.
When performing a collision avoidance operation plan between robots including a plurality of task sets, it is necessary to satisfy a front-back relationship between tasks. For example, FIG. 3B shows that after the components A and B are assembled on the workpiece W, the component C is assembled on the workpiece W. Therefore, in order to clarify the correspondence between the predecessor relation between tasks and the predecessor relation between robot operations, as shown in FIG. 9, an expansion rule in which a trigger input point (TRG_IN) and a trigger output point (TRG_OUT) are described in advance is defined. , In the deployment rule information section (110 in FIG. 1). Then, the inter-robot collision avoidance operation planning unit (115 in FIG. 1) uses the internal predecessor-relationship consideration unit 115b to perform the inter-robot collision under the constraint that the TRG_OUT of the preceding task and the TRG_IN of the subsequent task operate after TRG_OUT. Make an avoidance action plan. That is, it is possible for the robot to pick up the component C and move it to the vicinity of the work W before the assembly operation of the component A and the component B is completed.
[0033]
Further, when the same robot executes tasks in both task sets in two consecutive task sets, a process of merging those created as separate operation programs in S5-5c into one operation program is performed. Here, as shown in FIG. 1, the operation program merging unit 113a is mounted inside the robot operation planning unit 113. The operation program merging unit 113a includes a trajectory deleting unit 113a1 and a trajectory adding unit 113a2 inside. The trajectory deletion unit 113a1 deletes the retreat operation from the final work point of the operation program to the end position and the approach operation from the start position to the start work point of the next process operation program. Then, the trajectory adding unit 113a2 adds a trajectory that moves between the last work point of the operation program and the start work point of the next process operation program. The two operation programs are merged into one by the operation of the moving trajectory deletion processing and the addition processing.
As described above, when merging a plurality of operation programs into one, unnecessary robot operations are eliminated, and collision between robots is avoided by strictly reflecting the front-back relationship between tasks in the front-back relationship of robot operations. Is performed, the waiting time of the robot is shorter, and the overall working time is further reduced.
[0034]
Next, a third embodiment according to the present invention will be described. In the third embodiment, the task distribution planning unit 111, task processing order planning unit 112, robot operation planning unit 113, and robot collision avoidance operation planning unit 115 shown in FIG. To be implemented on a computer network. These agents perform distributed processing on the computer network via the task information 102. Further, the task information 102 may be one independent agent.
Further, the robot operation planning unit 113 is mounted on a computer network as an independent agent for each robot. These agents proceed in parallel and distributed manner on the computer network via the task information 102.
[0035]
As described above, since the robot operation planning unit only considers collision avoidance between the robot itself and a fixed obstacle (excluding other robots), the operation plan calculation for each robot can be performed independently. Therefore, by implementing the robot operation planning unit on the computer network as an independent agent for each robot, parallel distributed processing becomes possible.
[0036]
【The invention's effect】
As described above, in the present invention, the operation start position and the operation end position are set in a safe area where the robots do not collide with each other, and an operation program for each robot is created. Since the collision avoidance process is performed, there is an effect that a deadlock state does not occur between the robots.
In addition, the robot collision avoidance operation planning unit does not collide with each other on the trajectories of the operation commands constituting the operation program of the plurality of robots or the trajectory obtained by dividing the trajectory into a finite number, and Since the execution order of the trajectories between different robots is determined so that the time is minimized, the work time will increase even when the moving distance of the robot in the work is long or most of the motion area is shared between the robots. There is an effect that it can be suppressed as much as possible.
In addition, the motion planning unit for avoiding motion between robots divides the trajectory in the motion program into small pieces, but does not change the execution order in the same motion program. is there.
In addition, when merging a plurality of operation programs into one, unnecessary operations of the robot are eliminated, and collision between robots is avoided by strictly reflecting the front-back relationship between tasks in the front-back relationship of the robot operation. Therefore, the waiting time of the robot is shorter, and the overall working time is further reduced.
Furthermore, since the robot operation planning unit only considers collision avoidance between the robot itself and a fixed obstacle (excluding other robots), the robot operation planning unit is implemented on the computer network as an independent agent for each robot. This also has the effect of enabling parallel distributed processing.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a robot teaching system according to an embodiment of the present invention.
FIG. 2 is a diagram showing an example of an assembly robot cell;
FIG. 3 is a diagram showing an example of work information;
FIG. 4 is a diagram showing a data structure of a search tree according to the embodiment of the present invention;
FIG. 5 is a flowchart of a task search priority search according to the embodiment of the present invention;
FIG. 6 is a diagram showing an example of work / part information;
FIG. 7 is a diagram showing an example of deployment rule information.
FIG. 8 is a diagram showing an example of a robot operation program.
FIG. 9 is a diagram showing an example of expansion rule information specifying a trigger input / output location;
FIG. 10 is a diagram showing an outline of an operation plan for avoiding collision between robots.
FIG. 11 is a diagram showing an example of an operation program modified by an inter-robot collision avoidance operation plan.
FIG. 12 is a diagram showing an initial solution of Conventional Example 1.
FIG. 13 is a diagram showing the principle of collision avoidance between robots in Conventional Example 2.
FIG. 14 is a diagram showing the operation principle of Conventional Example 3;
FIG. 15 is a diagram showing a configuration of a conventional example 4;
[Explanation of symbols]
101: Task input section
102: Task information
103: Condition input section
104: Work / part information
105: Model input unit
106: Geometric model
107: Robot registration unit
108: Robot information
109: Rule registration unit
110: Deployment rule information
111: Task distribution planning unit
112: task processing order planning unit
111a, 112a: evaluation criterion calculation unit
113: Robot Motion Planning Department
113a: Operation program merging unit
113a1: Track deletion unit
113a2: Track adder
114: Operation program
115: Robot collision avoidance operation planning unit
115a: optimization processing unit
115b: pre / post relationship consideration unit
201a to 201c: Robot
202: Belt conveyor
203a to 203d: component supply device
204a to 204c: Tool holder
401: ClosedList of search tree
402: OpenList of search tree
403a to 403f: pointer to parent node of search tree node
601: Part A
602: Parts supply device
603: Coordinate system ΣA fixed to part A
604: Coordinate system $ pf1 fixed to component supply device
605: Coordinate system Σaprch indicating approach position to component A
606: Coordinate system @catch indicating gripping position of component A
607: Coordinate system $ leave indicating the detached position after gripping part A
701: Place task deployment rules
702: Expansion rule of ChangeTool subtask
703: Expansion rule of PickUp subtask

Claims (8)

In order to perform a predetermined operation on a plurality of robots, the operation program is searched by preferentially searching for the one that minimizes or maximizes the evaluation criterion from the combination of the task allocation and the operation processing order for each robot. In the generated robot teaching system,
A task distribution planning unit that distributes the work to the plurality of robots,
A task processing order planning unit for planning a task processing order in each robot of the distributed tasks,
Generation of an operation program that avoids collision between each robot and an obstacle based on the task distribution plan and the processing order of the tasks, and sets the operation start position and end position of each robot as an area where robots do not collide with each other. A robot operation planning unit for calculating an operation time,
When there is a trajectory in which a collision occurs between robots in the operation program, the operation program is changed so that the trajectory is divided into a plurality of parts, and the divided trajectory is prevented from colliding between robots. An inter-robot collision avoidance operation planning unit for calculating an operation time,
A robot teaching system, comprising: an evaluation criterion calculation unit that calculates the evaluation criterion based on the operation time of the robot operation planning unit and the operation time of the inter-robot collision avoidance operation planning unit. 2. The robot teaching system according to claim 1, wherein the robot collision avoidance operation planning unit includes an optimization processing unit that changes an operation plan of each robot so as to minimize the operation time of the plurality of robots. . The robot teaching system according to claim 1, further comprising an operation program merging unit that merges a plurality of the operation programs of the same robot into one operation program according to an execution order. The operation program merging unit is a trajectory deletion unit that deletes the trajectory from the final work point to the end position of the operation program and the trajectory from the start position of the next process operation program to the start work point,
The robot teaching system according to claim 3, further comprising: a trajectory adding unit that adds a trajectory from the final working point to the start working point between the two operation programs. The task processing order planning unit generates a task set list in which a plurality of tasks that can be executed in parallel are set as one task set based on a task graph describing a predecessor / relationship between the tasks, and the task sets are connected in series. With features,
5. The robot teaching system according to claim 1, wherein the task distribution planning unit has a function of searching for a combination of robots that execute each task from the task set at the head of the task set list. 6. An expansion rule information unit for associating the pre / post relationship between the tasks with the pre / post relationship between the robots,
5. The robot teaching system according to claim 1, wherein the inter-robot collision avoidance operation planning unit includes a back-and-forth relationship consideration unit that changes an operation program based on the back-and-forth relationship between the tasks. 6. 6. The robot teaching device according to claim 1, wherein the task distribution planning unit, the task processing order planning unit, and the inter-robot collision avoidance operation planning unit are software objects that can be independently processed in a distributed manner. system. 7. The robot teaching system according to claim 1, wherein the robot operation planning unit is an independent software object for each robot.

Images