JP5981215B2 - Method and system for automatically preventing deadlock in a multi-robot system - Google Patents

Description

CROSS REFERENCE TO RELATED APPLICATIONS This application is a continuation-in-part of co-pending US patent application Ser. No. 12 / 124,430, filed May 21, 2008.

  This application claims the benefit of US Provisional Application No. 61 / 482,808, filed May 5, 2011.

  The present invention generally relates to a system for controlling a plurality of robots and a method for controlling motion interference avoidance of the plurality of robots.

  The movement of objects in space is a necessary task in a typical manufacturing environment. Robotics are increasingly being used to perform the necessary actions. However, when operating a plurality of objects, interference may occur between the objects. There is interference when at least two objects share the same space simultaneously, such as when the objects have the same coordinates with respect to the same coordinate system.

  Since modern industrial robots operate at considerable speeds, collisions can occur due to interference between robots, which can cause unwanted damage to the robot and workpieces manipulated by the robot. Collisions can cause costly failure times during the manufacturing process. It is therefore desirable to avoid such a collision.

  Conventional systems and methods have been used to minimize interference and collisions. However, there are several disadvantages of conventional systems and methods. Typically, the tool center point (TCP) is only confirmed against a predetermined interference area or “static space”. It is difficult to prevent these collisions or interferences with multiple robots immediately or effectively. Furthermore, it is difficult to specify the interference space related to the stationary coordinate system of a plurality of moving robots. The arbitrary interference space is not only a function of the robot's motion path, but also a function of the motion speed. There are also difficulties in attempting to handle deadlock situations when two or more robots are required to operate simultaneously in a shared space.

  Conventional systems have also attempted to prevent the robot TCP from colliding in a fixed space relative to the world coordinate system. When multiple robots (with multiple controllers) share a common space or “interference” space during task execution, each controller must wait until the robot is no longer in the common space. In this case, the controller can issue an operation control command to allow the robot to operate. This process is also referred to as the “standby and operation” process and generally increases the work cycle time. However, since the interference space is not only a function of the motion path of the robot but also a function of the motion speed, it is difficult to effectively specify the interference space for the fixed coordinate system. When more than one robot requests to operate simultaneously in a common space, the robot creates a deadlock condition in which neither robot can operate because the robot waits for each other.

  Conventional systems also attempt to model robots with spheres and cylinders. Conventional systems predict the future position of the robot during real-time operation. Since conventional systems do not determine the accumulated space occupied by the robot over a predetermined time, conventional systems need to make frequent comparisons during robot operation. Conventional systems compare models of all robots in a work cell for each component. This comparison is very computationally expensive and increases dramatically as the number of components used to build the robot and robot and tool models increases. Because comparisons are made in real time when an impending collision is detected, conventional systems typically need to stop all robots associated with an impending collision and interrupt the automatic program operation . Conventional systems become even more difficult when the robot belongs to a different controller. This is because a conventional system requires a large amount of information in order to perform real-time communication between controllers. Prior systems have also attempted to use an I / O handshake mechanism to avoid interference. In the present invention, there is no need for I / O PLC.

  One known system and method is disclosed in co-pending international application number PCT / US2007 / 066668 by assignee, which is hereby incorporated by reference in its entirety. Such systems and methods have a “dynamic space check” system that maximizes the efficiency of robot operation and minimizes the possibility of interference or collision of multiple robots. The robot controlled by each controller operates only in a user-defined dynamic space, thereby avoiding collisions. However, dynamic space validation systems generally protect TCP only against user-defined rectilinear spaces.

  Another known method of avoiding robotic collisions is reported in US Pat. No. 5,150,452 by Pollack et al. The method includes creating a collision map that includes the desired robot motion. The robot map and the “world” map are combined in the exclusive OR operation, and then the collision map and the “world” map are combined in the exclusive OR operation, followed by the collision map and the “world” in the inclusive OR operation. By combining the "map" with the byte, the original position of the desired robot is removed from the "world" map. A collision is represented by a difference in arbitrary bit positions of the combination of inclusive or exclusive OR. This method provides a two-dimensional x-y projection and a one-dimensional height to detect a collision, but cannot perform a three-dimensional real-time collision detection.

  Yet another known method of detecting a collision between a robot and one or more obstacles before the collision occurs is described in US Pat. No. 5,347,459 by Greenspan et al. Robots are modeled by spheres in a voxelized workspace. Each voxel in the workspace is assigned a value corresponding to the distance from the nearest obstacle. If the voxel value at the center of the sphere is smaller than the radius of the other sphere in the voxel, it is determined that the collision is near. However, this method only protects a single robot arm. Robots are modeled only by spheres and are therefore insufficient to protect the robot's critical process paths.

  There remains a need for systems and methods for controlling motion interference avoidance of multiple robots. Preferably, the system and method provides three-dimensional real-time collision detection, pre-providing robot motion to the robot system, preserving collision-free programmed trajectories, and protecting critical process paths To do.

  A program or task causes interference between one or more robots associated with the program or task as a result of continued sequential execution of either program or task, and proceeds sequentially without interference A deadlock condition occurs when there is no program or task that can perform.

  US Pat. No. 7,114,157 by Chaffee et al. Describes a method for avoiding deadlocks by acquiring resources in a set order. Although this method can avoid deadlocks, this method does not have the ability to allow operation if a set deadlock-free order is not determined. In addition, this method does not allow an operation that can be made deadlock free if the requested instruction is different from the prescribed deadlock free instruction.

  US Patent Application No. 2009/0326711 by Chang et al. Describes a method of using automatic zones to avoid deadlocks automatically or manually. Chang suggests how to use priority values to prevent deadlocks. Chang does not suggest a way to prevent deadlocks when priority values are not used.

  According to the present disclosure, there is provided a system and method for controlling motion interference avoidance of a plurality of robots, which performs three-dimensional real-time collision detection, provides robot operations to the robot system in advance, and provides a collision-free program. A system and method that preserves critical trajectories and protects critical process paths without utilizing an I / O handshake mechanism is a surprising discovery.

  In one embodiment, a method for preventing deadlock of a pair of robots having a common workspace, wherein each of the robots is controlled by an associated program and each of the robots when the program is executed simultaneously Occupying at least a portion of the common workspace during a portion of the execution of the associated program, and identifying a portion of the common workspace that is occupied by the robot during concurrent execution of the program, Identifying at least one interference region that overlaps a portion of the workspace of the computer, analyzing at least one interference region and identifying a location where at least one deadlock condition of the two robots may occur, and at least At least one deadlock prior to the execution of any robot action in a single deadlock state By automatically determining and executing the Kufuri operation statement, comprising the steps of avoiding at least one of the deadlock in the program.

  In another embodiment, a computer readable medium having instructions executable by a computer to perform a method for preventing deadlock of a pair of robots having a common work space, each of the robots comprising: Each of the robots occupies at least a portion of a common work space during a portion of execution of the associated program, and the method is controlled by the associated program and the method is executed simultaneously. A command for identifying a part of the common workspace occupied by the robot, a command for identifying at least one interference area overlapping with the part of the common workspace, and analyzing at least one interference area and two Instructions identifying where at least one deadlock condition of the robot may occur, and at least one Instructions for avoiding at least one deadlock condition during program execution by automatically determining and executing at least one deadlock free action statement prior to execution of any robot motion that is in a deadlock state; .

  In another embodiment, a method for preventing deadlock of a plurality of robots having a common work space, each of the robots being controlled by an associated program and each of the robots when the program is executed simultaneously Occupying at least a portion of a common workspace during a portion of execution of an associated program, and identifying a portion of the common workspace that is occupied by the robot during concurrent execution of the program; and Identifying at least one interference region that overlaps a common workspace portion for at least two robots, analyzing at least one interference region, and causing at least one deadlock condition of at least two robots A step of identifying the location and at least one deadlock condition Avoiding at least one deadlock condition during program execution by automatically determining and executing at least one deadlock free motion statement prior to execution of any motion of at least two robots; Prepare.

  The system and method according to the present invention does not require a known deadlock free sequence. The system and method according to the present invention automatically determines possible deadlock conditions and establishes a method for avoiding possible deadlock conditions. The system and method according to the present invention does not require a priority value to prevent deadlock. The system and method according to the present invention prevent deadlock conditions that may exist at the same time as preventing interference conditions.

  These and other advantages of the present invention will become apparent to those skilled in the art from the detailed description of the preferred embodiment when considered in conjunction with the accompanying drawings.

2 illustrates an exemplary robot system according to the present disclosure having a first robot and a second robot operating in a work cell. FIG. 5 is a process diagram illustrating an automatic interference confirmation classification method according to the present disclosure. The voxelization model of one of the 1st robot and the 2nd robot shown in FIG. 1 is shown. FIG. 4 is an isometric view of the voxelization model shown in FIG. 3. 1 is an enlarged view of one of the first robot and the second robot shown in FIG. 1 having a voxelization model formed by voxelized spheres and cylinders covering one of the first robot and the second robot. Indicates. 6 shows the voxelized sphere and voxelized cylinder shown in FIG. 5 formed from a plurality of voxels. 2 shows a deadlock free matrix and associated program listing according to the present invention. 2 shows a deadlock free matrix and associated program listing according to the present invention. 2 shows a deadlock free matrix and associated program listing according to the present invention. Fig. 4 shows an interference matrix and associated program list according to the present invention. Fig. 4 shows an interference matrix and associated program list according to the present invention. Fig. 4 shows an interference matrix and associated program list according to the present invention. Fig. 4 shows an interference matrix and associated program list according to the present invention. Fig. 4 shows a path and associated table for combining multiple deadlock areas into a single deadlock free area. 2 shows a plurality of robot program execution sequences not performing deadlock prevention according to the present invention and a plurality of robot program execution sequences performing deadlock prevention according to the present invention. 3 is a flowchart of a deadlock avoidance sequence according to the present invention. 4 is a flow chart for determining possible deadlock regions according to the present invention.

  The following description is merely exemplary in nature and is not intended to limit the disclosure, the application, or the use. In the drawings, it should be understood that corresponding reference numerals indicate like or corresponding parts and features. With respect to the disclosed method, the steps shown are exemplary in nature and are therefore not required or important.

  FIG. 1 is a linear diagram showing a robot system 2 that controls collision avoidance between a plurality of robots. Non-limiting examples include robot system 2, body shop robot system having at least two robots, waterjet cutting robot system, laser welding robot system, arc welding robot system and painting. It can be one of the robot systems. Other robot systems 2 having a plurality of robots can be used as desired.

  The robot system 2 has a work cell 4 that defines an envelope configured to operate the first robot 6 and the second robot 8. The first robot 6 and the second robot 8 are configured to selectively occupy at least one common space 10 arranged in the work cell 4. Only the first robot 6 and the second robot 8 are shown, but it will be appreciated that the robot system 2 can have more than two robots without departing from the scope and spirit of the disclosure.

  The first robot 6 can occupy the first portion 12 of the common space 10 during operation of the first robot 6 along a first programmed path. The second robot 8 can occupy the second portion 14 of the common space 10 during operation of the second robot 8 along the second program path. The first portion 12 and the second portion 14 are also known as “automatic zones” or “autozones”. Each program path comprises one or more automatic partitions, each automatic partition being pre-processed and obtained from one or more action statements. The first program path and the second program path can be controlled, for example, by one or more action statements in a series of instructions having a plurality of action statements. A person skilled in the art knows that robot interference may occur because the first portion 12 and the second portion 14 may overlap if the interference avoidance method for controlling the operation of the first robot 6 and the second robot 8 is not used. It should be understood that

  The first robot 6 and the second robot 8 are controlled by at least one controller 16, 18. In the embodiment shown in FIG. 1, the first robot 6 is controlled by a first controller 16, and the second robot 8 is controlled by a second controller 18. At least one controller 16, 18 is adapted to make electrical communication with a power source (not shown). Controllers 16 and 18 can execute a series of instructions such as computer programs residing in controllers 16 and 18. In other embodiments, the sequence of instructions may reside on a computer readable medium or memory that interacts with the controllers 16, 18.

  The robot system 2 is in charge of electrical communication with a network medium (not shown), a programmable logic device (not shown) and at least one controller 16, 18 configured to connect to various system components. It can further comprise other components known in the art, such as at least one of the teaching devices 20. In a particular embodiment, the teaching device 20 can have a monitor and can be used for viewing the first robot 6, the second robot 8, and the common space 10 for observation by an operator of the robot system 2 as desired. The first portion 12 and the second portion 14 of the common space 10 are configured to be represented by an image. The teaching device 20 can also include means for initiating a series of instructions for performing a robot jog operation.

  A typical sequence of instructions according to this disclosure is shown in FIG. The sequence of instructions has an interference check automatic zone method 200. The automatic interference confirmation method 200 first has a start step 202 in which at least one common space 10 is provided in the work cell 4. The start step 202 is followed by a first determination step 204 and a second determination step 206. The first determining step 204 includes determining the first portion 12 of the common space 10 that is occupied during the operation of the first robot 6 along the first program path. The second determining step 206 includes determining the second portion 14 of the common space 10 that is occupied during operation of the second robot 8 along the second program path. Of course, the first part 12 and the second part 14 can be determined automatically, i.e. without requiring user-specified segments of movement of each robot as is known in the art. . Thereafter, the first portion 12 and the second portion 14 are compared in a comparison step 208 to determine if there is an overlap 210 between them.

  As a non-limiting example, the first determination step and the second determination step can be performed substantially in real time during the starting operation of the robot system 2 having the first robot 6 and the second robot 8. The starting action can be performed by the first robot 6 and the second robot 8 in a lockdown mode to avoid any possible collision between the first robot 6 and the second robot 8. it can. In other examples, the first determination step 204 and the second determination step 206 can be occupied by the robots 6, 8 during real-time operation, for example, the first portion 12 and the second portion of the common space 10. 14 can be identified by an offline run of the first program path and the second program path. The first portion 12 and the second portion 14 identified during the first determination step 204 and the second determination step 206 are held in, for example, a memory, and the first robot 6 and the second robot 8 Can be reused during the next operation. As a result, in order to influence the collision between the first robot 6 and the second robot 8, the program path, that is, the trajectory of the operation of the first robot 6 and the second robot 8 is processed and held in advance. Should be understood.

  The first moving step 212 and the second moving step 214 of the first robot 6 and the second robot 8 are selected according to the presence of the overlap 210. If there is an overlap 210 between the first part 12 and the second part 14, there is a possibility of collision of the robots 6, 8, and the first movement step 212 is selected. The first movement step 212 includes movement of only one of the first robot 6 along the first program path and the second robot 8 along the second program path. If there is no overlap 210 between the first part 12 and the second part 14, there is no possibility of a collision of the robots 6 and 8, and the second movement step 214 is selected. The second moving step 214 includes movement of the first robot 6 along the first program path and movement of the second robot 8 along the second program path.

  The first determination step 204 and the second determination step 206 can be performed by any conventionally known means. In one particular embodiment according to the present disclosure, the first determination step 204 and the second determination step 206 comprise at least one voxel model 300 having at least one voxel 302 as illustrated in FIGS. Can be included to represent at least one of the common space 10, the first portion 12, and the second portion 14. The voxel 302 is a volume element that represents a value on a regular grid in a three-dimensional space. The comparing step 208 determines, for example, whether an overlap 210 exists between the first voxel model 300 representing the first portion 12 and the second voxel model 300 representing the second portion 14. Can have. The voxel 300 is generally represented by a plurality of voxels 302 that approximate the shapes of various components such as the bases, arms, and tools of the first robot 6 and the second robot 8.

  Referring to FIGS. 3 and 4, an example voxel model 300 may have a plurality of voxels arranged at coordinates along the X, Y, and Z axes. The voxel model 300 is configured to represent a three-dimensional volume of the common space 10 occupied by at least one of the first robot 6 and the second robot 8. The voxel model 300 can be in the form of a generated data file and can be stored, for example, in the controller 16,18, other computer readable media, or other memory. In certain embodiments, the voxel model 300 is dynamic and has a plurality of models 300 associated with a series of robot motions. The dynamic voxel model 300 can be used to represent a three-dimensional volume of the common space 10 that is occupied during the operation of at least one of the first robot 6 and the second robot 8. Each automatic segment is obtained from the accumulation / superimposition of multiple snapshots of the voxel model 300, and each snapshot is taken from one or more ITP intervals.

  Referring to FIGS. 5 and 6, the voxel model 300 of the first robot 6 and the second robot 8 is voxelized to represent the three-dimensional volume occupied by the first robot 6 and the second robot 8. At least one of a sphere 500 and a voxelized cylinder 502. As will be further described herein, the voxelized sphere 500 and the voxelized cylinder 502 represent various components such as the bases, arms and tools of the first robot 6 and the second robot 8. Can do. A voxelized sphere 500 and a voxelized cylinder 502 are typically a plurality of voxels 302 interleaved to approximate the general shape and boundary of the first robot 6 and the second robot 8. Have

  One skilled in the art should understand that the voxelized sphere 500 and the voxelized cylinder 502 can be generated by any suitable means. For example, the voxelized sphere 500 can be approximated by projecting a typical sphere representing the components of the robots 6 and 8 onto the XY plane. In this case, the smallest cylinder A surrounding the sphere is identified. Thereafter, the same sphere is projected onto the YZ plane, and the smallest cylinder B surrounding the sphere is identified. Thereafter, the same sphere is projected onto the ZX plane, and the smallest cylinder C surrounding the sphere is identified. The voxelized sphere 500 is then approximated by identifying the intersection of cylinder A, cylinder B, and cylinder C.

  In other embodiments, the voxelized sphere 500 can be approximated by first finding the smallest box that encloses the sphere representing the components of the robot 6 and the components of the robot 8. The box has a volume occupancy A. A voxel volume occupancy B that is inside the box A and outside the sphere is identified. The voxel sphere 500 is then approximated by subtracting the value B from the value A.

  The voxelized sphere 500 can also be approximated by identifying the intersection of the typical sphere and the XY plane. The intersection has the center of the sphere along the Z axis and forms an intersecting circular plate A1. An arbitrary height is assigned to the circular plane A1, and then the voxel occupancy of the intersecting circular plates A1 is identified. Furthermore, the circular plates A2, A3,. . . , A (n) are determined by shifting the intersection of the typical sphere and the XY plane along the Z axis, and these volume occupations are similarly identified. The voxelized sphere 500 is then approximated by consolidating each of the identified voxel occupancy against a typical spherical circular plate.

  The voxelized cylinder 502 identifies the left hemisphere / global voxel occupancy A, identifies the right hemisphere / global voxel occupancy B, and determines the voxel occupancy C of the plurality of circular plates between the left hemisphere and the right hemisphere. It can be approximated by identifying. In this case, a circular substrate perpendicular to the base line is formed. The baseline connects the two hemispheres at the ends of the cylinder. Next, a circular surface can be obtained by shifting the circular substrate along the Z axis. It is also possible to calculate the baseline voxel occupancy and then shift the baseline to fill the entire area between the two hemispheres. Next, the voxelized cylinder 502 is approximated by integrating each of voxel occupancy A, voxel occupancy B, and voxel occupancy C.

  Another method of approximating the voxelized cylinder 502 involves first identifying the smallest box that can surround a typical cylinder representing a robot component. The box has volume occupancy A. Thereafter, the voxel volume occupancy B that is inside the box and outside the cylinder is identified. Next, the voxelized cylinder 502 is approximated by subtracting voxel occupancy B from voxel volume occupancy A. Those skilled in the art should understand that other means suitable for approximating the voxelized spheres 500 and cylinders 502 representing the components of the robots 6, 8 can be used as desired.

  Voxelization can be similarly determined from any CAD surface or volume that can represent robots, arm dressouts, and tooling.

  Voxelization is a very effective way to represent the space occupied by robot motion segments. Although the voxelization process has some computational overhead to form a voxelized space, a run-time component is very useful. Once the space is voxelized, the maximum storage capacity required for the voxelized space is fixed regardless of the degree or amount of complexity of the common space occupied by the robot in the motion path. Runtime verification of interference between voxelized spaces is very effective. Although the preferred embodiment provides for interference confirmation automatic segmentation to be represented by voxelized regions, any method of representing the volume or surface of space occupied by the robot in the motion path can be used.

  Other embodiments avoid collisions during programmed operations, operations initiated by jog operations, or operations initiated by other means. Other embodiments protect the operating robot from collisions with stationary robots, other fixed objects, or other defined areas in the work cell. For programmed operations, a voxelization process or other interference verification automatic segmentation modeling process can be performed off-line or during a test run, and the voxelized data is efficient for later exploration. Stored in This process can also be performed in real time when a new motion path is determined from a new program, robot jog operation or motion initiated by other means. For these cases, the sequence of instructions can be determined by other means of initiating use or operation of the teaching device. The actual operation is not allowed to start until the process is complete and there is no interference due to this and other occupied interference confirmation auto-segments.

  In another embodiment, at least one of the first portion 12 and the second portion 14 of the common space 10 is occupied by a tooling attached to at least one of the first robot 6 and the second robot 8. There can be a tooling space and a dressout assembly (eg, an ArcTool wire feeder). The step of determining the overlap 210 includes a third portion of the common space 10 occupied by at least one of the first portion 12 and the second portion 14 and a non-robot obstacle (not shown). A comparison with can also be included. Obstacles can be identified manually by a set of parameters used to define the size, shape and position of the common space 10 as desired. In other examples, the method 200 can further comprise determining a third portion of the common space 10 occupied by the obstacle. It should be understood that the third portion of the common space occupied by obstacles other than robots can be represented by voxels 302 of a similar voxel model 300 as desired.

  The method 200 of the present disclosure includes an unoccupied portion of the common space 10 after at least one of the movement of the first robot 6 along the first program path and the movement of the second robot 8 along the second program path. Can be further included. In this case, an unoccupied portion of the common space 10 is made available for use by other robots following other program paths or trajectories as desired. In other embodiments, the method 200 transfers the coordinates of the first portion 12 and the second portion 14 of the common space 10 to at least one controller 16, 18 for other uses according to the method 200 of the present disclosure. Can have steps. In this case, the controllers 16 and 18 perform at least one of the first determination step 204 and the second determination step 206 described here, and coordinate the voxel model 300 representing the first portion 12 and the second portion 14. Can be used to convert

  A person skilled in the art will determine whether there is an overlap between the first portion 12 and the second portion 14 of the common space 10 and the operation of the first robot 6 and the operation of the second robot 8. It should be understood that each of these can be determined by a priority value. As a non-limiting example, the priority values of the first robot 6 and the second robot 8 can be based on a first-in first-out system. In another example, one of the first robot 6 and the second robot 8 can always have a higher priority value than the other of the first robot 6 and the second robot 8. The high priority value can be based on a predetermined user setting for the priority of the robots 6, 8 of the robot system 2. In other examples, the priority value can be selected based on the availability of unoccupied portions in the common space 10. A person skilled in the art can assign a desired priority value to the first robot 6 and the second robot 8 as desired.

  In other embodiments, the method 200 can further comprise analyzing a plurality of program paths for the presence of an overlap 210 that can be deadlocked. Next, the operation sequence of the plurality of program paths can be adjusted as desired to affect the occurrence of a deadlock condition. The operating sequence can be adjusted manually or automatically based on priority values, for example, as desired.

  Of course, the automatic interference confirmation segmentation system 2 and method 200 of the present disclosure can provide advantages over conventional multi-arm robot systems. System 2 and method 200 provide three-dimensional real-time collision avoidance. System 2 requires minimal programming. The reason is that system 2 affects the need for identifying or teaching interference areas. This also reduces production downtime.

  One skilled in the art should also understand that using voxelized spheres 500 and cylinders 502 typically minimizes processing requirements associated with robot model generation. The operations of the first robot 6 and the second robot 8 are transmitted in advance to the system 2 by, for example, off-line generation of the voxel model 300. Similarly, the program paths or trajectories of the first robot 6 and the second robot 8 are maintained without collision. A process path is a series of operations during an application process such as cutting, welding, grinding, routing, painting, dispensing, or other similar processes. It is generally important to complete the entire process path without interruption once the process is started. Of course, the critical process paths of the first robot 6 and the second robot 8 are similarly protected by the method 200.

  The system 2 supports the avoidance of multiple robot arm collisions within and between the same controllers 16, 18 as desired. In addition, system 2 and method 200 provide a simple robot configuration and affect the need for I / O handshake protocols as performed in certain conventional systems and methods. The automatic interference confirmation sorting system 2 and method 200 simplify the form of multiple arm interference confirmation and affect the deadlock of multiple robot arms.

  Interference prevention and deadlock prevention are closely related. Interference occurs when two robots attempt to occupy the same physical space. One way to prevent interference involves identifying the space occupied by all robots or currently occupied in the current motion command and stopping one or more robots before interference occurs. This type of interference prevention is stopped because one or more of the stopped robots are currently occupying space that is needed by other robots or will be needed in the future and continues to be disturbed by other robots. This has the serious disadvantage that deadlocks are very prone to occur when the robot in question cannot be started.

  Another way to prevent interference is to set it by programmatically inserting start and end zone commands into the robot program. These commands interact with the PLC or other robots to facilitate allowing only one robot to occupy a particular segment at a time. There is little effect on a single segment. The main issue is one of the priorities of who gets the classification when multiple robots are waiting when available. However, deadlock is very likely to occur for a plurality of sections. For example, for adjacent segments, it is easy to know whether each robot currently occupies one segment and the next action is performed in the segment occupied by another robot. In this case, each robot will wait forever for the other robot to give up the desired segment. A deadlock condition occurs.

  The system and method according to the present invention eliminates the deadlock condition in the above case by determining a deadlock free action statement before performing an action with a possible deadlock condition. The determination of the deadlock-free operation statement can be made offline, separately from normal execution or during normal manufacturing execution. If sufficient CPU processing time is available, normal manufacturing in-progress decisions are most adaptive to respond to dynamic conditions such as changes in I / O timing or timing of external events or sequences Give sex. For minimal CPU effects, decisions are made offline, in which case a large number of permutations of the programming sequence can be analyzed to find an optimized sequence of execution. it can.

  The system and method according to the present invention does not require a known deadlock-free sequence. The system and method can automatically determine possible deadlock conditions and identify ways to avoid possible deadlock conditions. The system and method do not require priority values to avoid deadlocks. The system and method prevent interference conditions while at the same time eliminating the possibility of deadlock conditions.

  The “dead lock free matrix” 400 of the line number of the list 401 of “original program A” executed by the first robot and the line number of the list 403 of “original program B” executed by the second robot As shown in FIG. The shaded cells in the matrix 400 represent an upward convex hull that defines a deadlock free region. Most of the shaded cells represent pairs of program lines that have interference during the course of the path from line i-1 to line i of one program and the path from line j-1 to line j of the other program. . The remaining cells are required to complete the upward convex hull. Cells of the matrix 400 can be specified in the form (i, j), where i is the line number of program A and j is the line number of program B. Thus, the interfering cells are (3, 6; 3, 7; 4, 5; 4, 6; 4, 7; 5, 4; 5, 5; 5, 6; 5, 7), and the remaining cells are (3, 4; 3, 5; 4, 4).

  For example, if program A is on line 3 and program B is on line 5, a deadlock condition exists. The reason is that if A attempts to move to row 4, there is interference with B in row 5, and if B attempts to move to row 6, there is interference with A in row 3. . Therefore, none of the robots can advance. By completing the deadlock partition from A row 3 to row 5 and B row 4 to row 7, only one program A or B can occupy this partition at a time. This is identified as ZONE [1] in the example of the program as shown in the list 402 of “new program A” and the list 404 of “new program B”.

  FIG. 8 shows an example where the “interference matrix” 405 can be subdivided into three deadlock-free categories: ZONE [1], ZONE [2], ZONE [3]. The interfering cells are (3,3; 3,4; 3,5; 3,6; 3,7; 4,5; 4,6; 4,7; 5,4; 5,5; 5,6; 5 7: 6, 6; 6, 7). ZONE [1] has cells (3, 3; 3, 4; 3, 5; 3, 6; 3, 7). ZONE [2] has cells (4, 5; 4, 6; 4, 7; 5, 4; 5, 5; 5, 6; 5, 7). ZONE [3] has (6, 6; 6, 7). The upward convex hull is formed through the cells (4, 4) which are virtual interferences so that three convex hulls can be formed and the entire region is convex upward. The three sections are incorporated in the “new program A” list 406 and the “new program B” list 407. It is not possible for at least one program to proceed to the next row without interference in any combination of row pairs without interference.

  FIG. 9 shows an example of a program pair having a plurality of interference regions where two deadlocks may exist and two deadlock regions. “Program A” follows the path 409 from line 3 to line 15. “Program B” follows the path 410 from line 9 to line 21. The program has a path and a first deadlock area 412 and a second deadlock area 411 as shown in Table 413. A simple way to form a single deadlock region is to combine all interfering regions and intervening non-interfering regions into a single region as shown in Table 414, which was originally deadlocked. There is no.

  FIG. 10 shows five sequences of the robots 6 and 8 that execute “program A” and “program B”, respectively. In the first sequence on the far left, the robot 6 executes the “program A” without the robot 8 operating. In the next sequence on the right side, the robot 8 executes the “program B” without the robot 6 operating. In the middle sequence, the robots 6 and 8 execute each program simultaneously, and a collision occurs. The next sequence on the right shows the robots 6 and 8 that execute each program simultaneously while avoiding the conventional interference that causes deadlock. The rightmost sequence shows the robots 6 and 8 that simultaneously execute each program while preventing deadlock according to the present invention, and both programs are completed without interference.

  The systems and methods described above automatically add and / or modify program statements to both prevent collisions and handle possible deadlocks. In some cases, an automatic wait command that is issued increases the overall cycle time for a given job. If an increase in cycle time is not allowed, the system and method according to the present invention recommends or automatically creates a re-sequenced path that allows the robot to execute the program without stopping. That is, if there is a portion of the path that two robots will occupy the same space at one time, the system and method will be different if a new path sequence (if welding is assumed) to eliminate this condition. Automatically recommend spot order). The feature of resetting the program instructions in this way can be used for paint shop application.

  If conditional execution statements occur in the program, the analysis can include all combinations of sequential execution. Alternatively, for the sake of simplicity, the program segments included in the conditional branch area can be treated as independent. Typically, a robot programmer will teach such a conditional region so that the input path to the region and the output path to the region will have no interference with other robots and be able to make independent handling practical. To do. However, if the input path to the conditional area and the output path to the conditional area have interference with other robots, multiple combinations of conditional and non-conditional areas are used to ensure no interference and no deadlock. analyse.

  One advantage of this method is that the same mechanism that can eliminate the deadlock condition can also eliminate interference during normal program operation. Thus, real-time confirmation of interference between robots can be eliminated during normal program operation. This can ensure significant CPU utilization where processing time can be used for other purposes such as higher interpolation speeds, ie shorter interpolation times. In an error state, ie, when the program is not executed, the co-pending US patent application Ser. No. 12/11 entitled “Multi-Arm Robot System Interference Check Via Three Dimensional Automatic Zones”. Interference confirmation such as that proposed in US Pat. No. 124,430, which uses less CPU, can be used to prevent interference.

  The space occupied by a program can be a function of physical space, time, program sequence, line number, percentage of programs or lines, or state changes associated with a robot, program or system. The interference area and the deadlock area are related by the same function as the function defining the occupied space.

  There are various ways to avoid deadlocks. The simplest method is to expand the interference region to prevent deadlock and interference as shown in FIG. Expanding the interference area so that there is only one interference area for the entire program ensures that the system is deadlock free, but if cycle time is important, it is part of the sequence control. Sorting and / or multiple interference regions can be used. Another method is to provide an execution sequence outside the interference area to prevent deadlock. This can be easily done by having the first robot wait for the second robot to complete a portion of the program execution where a deadlock condition may exist before the first robot proceeds.

  There are several ways in which the occupied space can be determined. The simplest method is to make the occupied area purely the cumulative space occupied by the robot during program execution. When a plurality of robots have the occupied space determined in this way, the interference area is simply a space that is an intersection of the accumulated occupied spaces of the robots. By allowing only one robot to enter this space at a time, both interference and deadlock are prevented.

  Another method is to determine the occupied space as a function of time. There is a space occupied by each robot at each point in time. By representing the occupancy space as an integral of time over a specific period, the occupancy space can be represented as a series of discrete spaces related to time. The interference area is a robot intersection at the same time or time interval. Interference can be avoided by allowing only one robot at a time to be in the interference area. Since the relative time varies with sequence control, the time reference needs to be adjusted to relate to the new time after sequence control is performed.

  For the time-based method described above, time can be expressed as a time interval number. In this method, the total elapsed time is not important and the interference area can be represented as a matrix of the number of time intervals. When interference exists in a specific time interval state, it is necessary not to allow the occurrence of the time interval state. This is because a collision can occur in the time interval state.

  Similarly, a program sequence can be represented by line number, completion rate, program state, and the like. For any such sequential display, the specified interference as a function of the selected parameter can be displayed. The purpose of avoiding interference is to avoid the state of the set of functions in which interference exists. For a function that can be expressed as a sequential discrete interval or sequential state, this can be expressed as an interference matrix as shown in FIG. The interference matrix has an interference area. Of course, a matrix is just one way of representing data, but a matrix can be a more visual representation. In general, the matrix is sparse, and any sparse matrix technique can be applied to the display.

  State definitions such as I / O states or processes or other states can be used to define the region. As long as the state has a sequential meaning or can be defined to represent a sequential motion relative to the robot position and occupied space, the state can be used for parameter display of interference and deadlock regions.

  Once the interference matrix is determined, one way to prevent deadlock is to avoid any “trapping” of monotonically increasing sequence numbers where interference interferes with forward sequencing. By making the interference matrix ascendingly convex, both deadlock and interference can be avoided. Such an interference matrix that is convex upward can be referred to as a deadlock matrix. The deadlock matrix has a deadlock region as shown in FIG. Similar to interference, a deadlock matrix is just one way of representing data.

  Once the deadlock matrix is determined, there are a number of ways to avoid deadlocks and interference. The simplest means is to insert instructions into the program to adjust the execution timing so as to avoid deadlock areas of the deadlock matrix. Adding instructions to the program means that the exact nature of execution timing can be determined by mere program observation. Other means can determine the timing separately from the actual program change.

  The parameter definition of the interference area and deadlock area provides a general way to avoid interference and deadlock. The display of such areas allows various program controls with various factors.

  Also, to reduce CPU usage, the actual program can be updated to include partition and sequential information so that all the information necessary to avoid deadlock conditions is included in the program. This eliminates the need to check in real time the program sequence, interference and possible deadlock conditions.

  FIG. 11 is a flow diagram of a deadlock avoidance sequence according to the system and method of the present invention. In step 420, the program is taught to the robot controllers 16 and 18. In step 421, the occupied space is determined. Thereafter, in step 422, an interference area is determined. In step 423, possible deadlock regions are determined. In step 424, a program execution request for avoiding deadlock is determined. Finally, as described above, a mechanism for avoiding deadlock is determined in step 425.

  Step 423 is shown in more detail in FIG. In step 426, the occupied space and all interference areas are identified. In step 427, the combination of each interference area is examined. A check is made at decision step 428 as to whether a deadlock condition exists. If the result is NO, step 423 is terminated at step 429. If the result is yes, step 430 provides a means to prevent the occurrence of deadlock and returns to step 427.

  While certain representative embodiments and descriptions have been presented to illustrate the invention, those skilled in the art will make various modifications without departing from the scope of the disclosure as further described in the appended claims below. Understand that you can.

Claims (13)

A method for preventing deadlock of a pair of robots having a common work space, wherein each of the robots is controlled by an associated program, and when the programs are executed simultaneously to generate a normal program action, Each of the robots occupies at least a portion of the common workspace during a portion of execution of the associated program;
Prior to the normal program operation, the common work occupied by the robot by a start-up operation or off-line execution of a lockdown mode according to the program comprising representing an occupied portion of the common work space with a voxel model Determining a portion of space;
Determining the at least one interference region by comparing the voxel model to determine at least one interference region overlapping a portion of the common workspace occupied by the robot;
Analyzing the at least one interference area and determining a location where at least one deadlock condition of the two robots may occur;
The normal program operation by automatically determining a mechanism for avoiding the at least one deadlock condition prior to execution of the program of the normal program operation and associating the mechanism with at least one of the programs Avoiding the at least one deadlock condition during execution of the program of:
A method comprising the steps of:   The method of claim 1, wherein the mechanism includes at least one deadlock free action statement.   The method of claim 2, wherein the at least one deadlock free action statement is associated by adding to at least one of the programs.   3. The method of claim 2, wherein the at least one deadlock-free operation statement is related by changing at least one instruction in at least one of the programs.   3. The method of claim 2, wherein the at least one deadlock-free operation statement is related by reordering instructions in at least one of the programs.   The method of claim 2, wherein the at least one deadlock free action statement is associated by adding a wait command to at least one of the programs. A computer readable medium having instructions executable by a computer to perform a method for preventing deadlock of a pair of robots having a common work space, each of said robots being controlled by an associated program Each of the robots occupies at least a portion of the common work space during a portion of execution of the associated program when the programs are executed simultaneously to generate normal program operations, and the computer The readable medium is
Prior to the normal program operation, the common work occupied by the robot by a start-up operation or off-line execution of a lockdown mode according to the program comprising representing an occupied portion of the common work space with a voxel model An instruction to determine a portion of space;
Instructions for determining the at least one interference region by comparing the voxel models to determine at least one interference region overlapping a portion of the common workspace;
Instructions for analyzing the at least one interference region and determining where a deadlock condition of at least one of the two robots may occur;
The normal program operation by automatically determining a mechanism for avoiding the at least one deadlock condition prior to execution of the program of the normal program operation and associating the mechanism with at least one of the programs Instructions to avoid the at least one deadlock condition during execution of the program of
A computer readable medium comprising: The computer-readable medium of claim 7, wherein the mechanism includes at least one deadlock free operational statement. The computer-readable medium of claim 8, wherein the at least one deadlock-free operation statement is associated by adding to at least one of the programs. The computer-readable medium of claim 8, wherein the at least one deadlock-free operation statement is associated by modifying at least one instruction in at least one of the programs. The computer-readable medium of claim 8, wherein the at least one deadlock-free operation statement is related by reordering instructions in at least one of the programs. The computer-readable medium of claim 8, wherein the at least one deadlock-free operation statement is associated by adding a wait command to at least one of the programs. A method for preventing deadlock of a plurality of robots having a common work space, wherein each of the robots is controlled by an associated program, and when the program is executed simultaneously to generate a normal program action, Each of the robots occupies at least a portion of the common workspace during a portion of execution of the associated program;
Prior to the normal program operation, the common work occupied by the robot by a start-up operation or off-line execution of a lockdown mode according to the program comprising representing an occupied portion of the common work space with a voxel model Determining a portion of space;
Determining the at least one interference region by comparing the voxel models to determine at least one interference region that overlaps a portion of the common workspace for at least two robots of the robot;
Analyzing the at least one interference region and determining a location where a deadlock condition of at least one of the at least two robots may occur;
The normal program operation by automatically determining a mechanism for avoiding the at least one deadlock condition prior to execution of the program of the normal program operation and associating the mechanism with at least one of the programs Avoiding the at least one deadlock condition during execution of the program of:
A method comprising the steps of:

Images